{"id":5287,"date":"2026-02-24T04:14:15","date_gmt":"2026-02-24T04:14:15","guid":{"rendered":"https:\/\/research.undetectable.ai\/?p=5287"},"modified":"2026-02-24T04:14:50","modified_gmt":"2026-02-24T04:14:50","slug":"i-tested-5-ai-image-detectors-these-are-the-best-in-2026","status":"publish","type":"post","link":"https:\/\/research.undetectable.ai\/es\/he-probado-5-detectores-de-imagen-ai-estos-son-los-mejores-en-2026\/","title":{"rendered":"He probado 5 detectores de im\u00e1genes AI. Estos son los mejores en 2026:"},"content":{"rendered":"

Probablemente hayas visto en Internet m\u00e1s fotos generadas por IA de las que crees. <\/strong><\/p>\n\n\n\n

A veces es obvio que una imagen ha sido generada por IA, pero cada vez es m\u00e1s dif\u00edcil saberlo a medida que mejoran las herramientas generativas de imagen y v\u00eddeo. Nuevas herramientas como Google Nano Banana Pro<\/a>y actualizaciones de OpenAI Modelo de imagen ChatGPT<\/a> permiten a los usuarios generar r\u00e1pidamente im\u00e1genes sint\u00e9ticas que reflejan las reales. Investigaciones anteriores encontraron que 85% de los estadounidenses dicen los deepfakes erosionan la confianza en l\u00ednea<\/a>.<\/p>\n\n\n\n

M\u00faltiples herramientas afirman detectar im\u00e1genes generadas por IA, pero \u00bffuncionan? <\/strong><\/h2>\n\n\n\n

He realizado 50 comprobaciones de detecci\u00f3n en cinco de los detectores de im\u00e1genes de IA m\u00e1s populares y he documentado los resultados. No solo presentar\u00e9 todos los datos en este art\u00edculo y los explicar\u00e9, sino que tambi\u00e9n enlazar\u00e9 con la documentaci\u00f3n al final.<\/p>\n\n\n\n

Los cinco detectores que utilic\u00e9 para estas pruebas fueron: TruthScan<\/a>, IA o no<\/a>, Motor de visi\u00f3n<\/a>, WasItAI<\/a>y Winston AI<\/a>. Prob\u00e9 estas herramientas haciendo que cada una analizara dos im\u00e1genes generadas por ChatGPT, seis im\u00e1genes generadas por Nano Banana y dos im\u00e1genes generadas por Midjourney. Utilic\u00e9 varios estilos y t\u00e9cnicas de incitaci\u00f3n (que detallar\u00e9 m\u00e1s adelante, cuando analice cada imagen).<\/p>\n\n\n\n

Los cinco detectores se encargaron de detectar im\u00e1genes creadas por IA en categor\u00edas como fraude, desinformaci\u00f3n, fotograf\u00eda general y deepfakes. Como era de esperar, no todos los detectores obtuvieron buenos resultados. TruthScan fue el \u00fanico detector que clasific\u00f3 sistem\u00e1ticamente todos los contenidos que envi\u00e9 a 97% o superior.<\/strong><\/p>\n\n\n\n\n\n\n\n\nResultados de la prueba del detector de im\u00e1genes AI<\/title>\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .ai-detector-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .ai-detector-wrap h2 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.75rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .ai-detector-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .ai-detector-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.85rem;\n text-transform: uppercase;\n letter-spacing: 0.06em;\n padding: 16px 20px;\n text-align: center;\n border: none;\n }\n\n .ai-detector-table thead th:first-child {\n text-align: left;\n padding-left: 24px;\n }\n\n .ai-detector-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .ai-detector-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .ai-detector-table tbody td {\n padding: 14px 20px;\n text-align: center;\n font-size: 0.95rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .ai-detector-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .ai-detector-table tbody td:first-child {\n text-align: left;\n padding-left: 24px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n \/* Colored cells *\/\n .cell-green {\n background: #22c55e;\n color: #fff !important;\n font-weight: 700 !important;\n border-radius: 6px;\n padding: 6px 12px;\n display: inline-block;\n min-width: 32px;\n text-align: center;\n font-size: 0.9rem;\n }\n\n .cell-red {\n background: #ef4444;\n color: #fff !important;\n font-weight: 700 !important;\n border-radius: 6px;\n padding: 6px 12px;\n display: inline-block;\n min-width: 32px;\n text-align: center;\n font-size: 0.9rem;\n }\n\n \/* Accuracy badges *\/\n .badge-100 { background: #16a34a; color: #fff; }\n .badge-80 { background: #a3e635; color: #1a1a2e; }\n .badge-70 { background: #facc15; color: #1a1a2e; }\n .badge-60 { background: #fb923c; color: #fff; }\n .badge-30 { background: #ef4444; color: #fff; }\n\n .accuracy-badge {\n font-weight: 800;\n font-family: 'Outfit', sans-serif;\n font-size: 0.95rem;\n border-radius: 8px;\n padding: 6px 16px;\n display: inline-block;\n min-width: 60px;\n text-align: center;\n letter-spacing: 0.02em;\n }\n\n \/* Row rank indicator *\/\n .rank-icon {\n margin-right: 6px;\n font-size: 1rem;\n }\n\n \/* Asterisk note *\/\n .table-note {\n margin-top: 0.75rem;\n font-size: 0.8rem;\n color: #888;\n padding-left: 4px;\n }\n\n \/* Responsive *\/\n @media (max-width: 600px) {\n .ai-detector-table thead th,\n .ai-detector-table tbody td {\n padding: 10px 10px;\n font-size: 0.8rem;\n }\n .ai-detector-wrap h2 {\n font-size: 1.3rem;\n }\n }\n<\/style>\n<\/head>\n<body style=\"background:#f8f8fa; padding: 2rem;\">\n\n<div class=\"ai-detector-wrap\">\n <h2>\ud83d\udd0d Resultados de la prueba del detector de im\u00e1genes AI<\/h2>\n\n <table class=\"ai-detector-table\">\n <thead>\n <tr>\n <th>Herramienta de detecci\u00f3n<\/th>\n <th>Total de pruebas<\/th>\n <th>Detectado correctamente<\/th>\n <th>Fallado (Falla)<\/th>\n <th>Precisi\u00f3n<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd47<\/span>TruthScan<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">10<\/span><\/td>\n <td><span class=\"cell-green\">0<\/span><\/td>\n <td><span class=\"accuracy-badge badge-100\">100%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd48<\/span>IA o no<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">8<\/span><\/td>\n <td><span class=\"cell-red\">2*<\/span><\/td>\n <td><span class=\"accuracy-badge badge-80\">80%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">\ud83e\udd49<\/span>Motor de visi\u00f3n<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">7<\/span><\/td>\n <td><span class=\"cell-red\">3<\/span><\/td>\n <td><span class=\"accuracy-badge badge-70\">70%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">   <\/span>WasItAI<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">6<\/span><\/td>\n <td><span class=\"cell-red\">4<\/span><\/td>\n <td><span class=\"accuracy-badge badge-60\">60%<\/span><\/td>\n <\/tr>\n <tr>\n <td><span class=\"rank-icon\">   <\/span>Winston AI<\/td>\n <td>10<\/td>\n <td><span class=\"cell-green\">3<\/span><\/td>\n <td><span class=\"cell-red\">7<\/span><\/td>\n <td><span class=\"accuracy-badge badge-30\">30%<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n\n <p class=\"table-note\">* AI or Not etiquet\u00f3 una generaci\u00f3n AI como 78%, otra como 85% y otra como 89%. Como la prueba requiere al menos 90%, contamos las dos primeras como fallidas pero tratamos la 89% como casi fallida, de ah\u00ed el asterisco...<\/p>\n<\/div>\n\n<\/body>\n<\/html>\n\n\n\n<p style=\"font-size:17px\">AI or Not tambi\u00e9n funcion\u00f3 bastante bien. De hecho, aunque AI or Not clasific\u00f3 m\u00faltiples elementos generados por la IA por debajo de 90% de certeza, fue el segundo detector m\u00e1s preciso y consistente durante mis pruebas, por detr\u00e1s de TruthScan. Incluso cuando AI or Not fall\u00f3, esos fallos no fueron catastr\u00f3ficos, pero a\u00fan hay margen de mejora. <\/p>\n\n\n\n<p style=\"font-size:17px\">El resto de los detectores de im\u00e1genes de IA que prob\u00e9 obtuvieron resultados mucho peores. Por ejemplo, Sight Engine clasific\u00f3 err\u00f3neamente como aut\u00e9nticas 3 im\u00e1genes de IA relacionadas con fraudes.<\/p>\n\n\n\n<p style=\"font-size:17px\">Ahora te mostrar\u00e9 cada imagen (de las 10 que gener\u00e9) te explicar\u00e9 c\u00f3mo la cre\u00e9 y te mostrar\u00e9 c\u00f3mo la puntu\u00f3 cada modelo.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#1. \"Man On a Ledge\" (generado por ChatGPT)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"683\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-683x1024.jpg\" alt=\"\" class=\"wp-image-5289\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-683x1024.jpg 683w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-200x300.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-768x1152.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge-8x12.jpg 8w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT-Generation-1-Man-on-ledge.jpg 1024w\" sizes=\"auto, (max-width: 683px) 100vw, 683px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>ChatGPT <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td>General<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">77.99% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0,98% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p style=\"font-size:17px\">Para m\u00ed (y veo muchos contenidos generados por IA a diario), la imagen parece bastante convincente a primera vista. Si hicieras clic en el perfil de alguien en las redes sociales, te desplazaras por su feed y vieras esta imagen, \u00bfdestacar\u00edas inmediatamente que se trata de una falsificaci\u00f3n generada por IA? Para ser justos, intent\u00e9 ser creativo con la pregunta. \u00c9sta es la que utilic\u00e9 para generar esta imagen:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p class=\"has-text-align-center has-vivid-cyan-blue-color has-text-color has-link-color has-medium-font-size wp-elements-99b76fc78211de13a760aff177dd504c\">“<em><strong>Generame una foto estilo c\u00e1mara instant\u00e1nea de un hombre de pie en lo alto de un tejado mirando hacia la calle, de noche, est\u00e1 iluminado por el flash de la c\u00e1mara, hay rastros de luz de los coches de abajo, una mujer est\u00e1 de pie detr\u00e1s de \u00e9l con la mano en los labios, sonriendo, la est\u00e9tica de la foto es como una foto tomada con una c\u00e1mara instant\u00e1nea en 2009 por amigos de la universidad haciendo el tonto.<\/strong><\/em>“<\/p>\n<\/blockquote>\n<\/blockquote>\n<\/blockquote>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Por muy creativo que fuera o por muchos detalles que incluyera en mi mensaje, TruthScan y Sight Engine segu\u00edan clasificando correctamente el resultado como generado por la IA. El detector AI or Not se acerc\u00f3, pero no dio en el blanco. <\/p>\n\n\n\n<p style=\"font-size:17px\">Winston y WasItAI se equivocaron rotundamente al clasificar la imagen ChatGPT como real.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#2. \"El recibo falso\" (generado por ChatGPT)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"683\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-683x1024.jpg\" alt=\"Una imagen que muestra lo que parece ser un recibo manchado\" class=\"wp-image-5291\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-683x1024.jpg 683w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-200x300.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-768x1152.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest-8x12.jpg 8w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/CGPT2Bestbuyreceipttest.jpg 1024w\" sizes=\"auto, (max-width: 683px) 100vw, 683px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>ChatGPT <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td>Fraude<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">19.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">94.48% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0,04% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Quer\u00eda generar un recibo falso con ChatGPT para ver c\u00f3mo de realista se ver\u00eda. Puede que notes que a la imagen le falta una direcci\u00f3n leg\u00edtima, pero la textura sigue pareciendo ligeramente realista. Dir\u00eda que esta imagen es menos convincente que la primera que gener\u00e9 con ChatGPT, pero la mayor\u00eda de los detectores no funcionaron al analizarla.<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li style=\"font-size:17px\">De todas las comprobaciones que hice en esta imagen, <strong>TruthScan<\/strong> era el m\u00e1s preciso. <\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>Motor de visi\u00f3n<\/strong> realmente fall\u00f3 aqu\u00ed, mostrando mal juicio en una imagen de IA relacionada con documentos.<\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>IA o no<\/strong> obtuvo resultados notablemente mejores en este segundo an\u00e1lisis.<\/li>\n\n\n\n<li style=\"font-size:17px\"><strong>Winston AI <\/strong>obtuvo los peores resultados, y WasItAI tambi\u00e9n fracas\u00f3 por completo.<\/li>\n<\/ul>\n\n\n\n<p style=\"font-size:17px\">Dado que esta imagen est\u00e1 asociada a la categor\u00eda de fraude, es preocupante que la mayor\u00eda de los detectores la clasificaran err\u00f3neamente. <\/p>\n\n\n\n<h3 class=\"wp-block-heading\">El prompt que utilic\u00e9 para generar la imagen<\/h3>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-b58fc16e4a944c203389dba029f958d9\"><strong>“<em>Generar una imagen de un recibo de best buy, que diga que la cantidad total gastada es de mil millones de d\u00f3lares, y que haya una mancha de caf\u00e9 en \u00e9l.<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Est\u00e1 claro que escrib\u00ed un prompt r\u00e1pido y sencillo. A\u00fan as\u00ed, la calidad de la salida ChatGPT es visualmente convincente a simple vista. No tan convincente para dos de los cinco detectores que prob\u00e9, parece.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#3: \"Hazard Package\" (generado con Nano Banana)<\/h2>\n\n\n\n<p><strong><em>En lugar de hacer que Nano Banana generara una imagen original, le di una para que la modificara. <\/em><\/strong><\/p>\n\n\n\n<p style=\"font-size:17px\">La idea era demostrar c\u00f3mo alguien podr\u00eda utilizar f\u00e1cilmente la IA para crear pruebas falsas y afirmar que ha recibido un paquete da\u00f1ado\/peligroso. En primer lugar, tom\u00e9 una foto con un iPhone 15 de un paquete vac\u00edo de Amazon que encontr\u00e9.<\/p>\n\n\n\n<p><strong>Aqu\u00ed est\u00e1 la foto que tom\u00e9:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"759\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-759x1024.jpg\" alt=\"Una foto de un paquete de Amazon\" class=\"wp-image-5293\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-759x1024.jpg 759w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-222x300.jpg 222w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-768x1036.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Amazon-package-9x12.jpg 9w\" sizes=\"auto, (max-width: 759px) 100vw, 759px\" \/><\/figure>\n\n\n\n<p style=\"font-size:17px\">A continuaci\u00f3n, sub\u00ed la foto que tom\u00e9 a Nano Banana y le di la siguiente indicaci\u00f3n:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-41ee9a66efaa63aab6aedf0d2e80cce1\"><strong>“<em>Quiero que edites esta foto no cambies nada excepto donde est\u00e1 la etiqueta a\u00f1\u00e1dele da\u00f1os y a\u00f1ade manchas de lodo negro en la caja que parezcan manchas de grasa reales<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">Salida Nano Bana:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"747\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-747x1024.jpg\" alt=\"Una foto de un paquete\" class=\"wp-image-5292\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-747x1024.jpg 747w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-219x300.jpg 219w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-768x1052.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged-9x12.jpg 9w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Amazon-Package-Damaged.jpg 864w\" sizes=\"auto, (max-width: 747px) 100vw, 747px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edici\u00f3n Deepfake<\/span><\/td>\n <td>Fraude<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">98.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.10% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">31.98% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">El lodo aceitoso que Nano Banana gener\u00f3 en la salida ten\u00eda un convincente brillo mugriento, con un efecto de empapado, nada caricaturesco. Por suerte, la mayor\u00eda de los detectores por los que pas\u00e9 la imagen la etiquetaron correctamente como falsa. El \u00fanico caso at\u00edpico fue el de Winston, que, una vez m\u00e1s, dio un resultado mediocre, clasificando err\u00f3neamente la imagen como aut\u00e9ntica. AI or Not, TruthScan y WasItAI fueron los detectores m\u00e1s precisos a la hora de detectar esta imagen.<\/p>\n\n\n\n<p style=\"font-size:17px\">Es preocupante lo sencillo que parece subir una imagen a un chatbot y que la altere por completo. Me imagino a estafadores y estafadores utilizando herramientas de generaci\u00f3n de im\u00e1genes para intentar realizar reclamaciones de devoluci\u00f3n fraudulentas o cometer fraudes en el mercado.<\/p>\n\n\n\n<p style=\"font-size:17px\">Si las plataformas de comercio electr\u00f3nico s\u00f3lo requieren una imagen como prueba para iniciar una reclamaci\u00f3n por paquete da\u00f1ado, entonces, sin ninguna forma fiable de detectar la manipulaci\u00f3n de IA, est\u00e1n expuestas a un importante vector de ataque. Por ejemplo, el sistema oficial de Amazon <a href=\"https:\/\/sellercentral.amazon.com\/help\/hub\/reference\/external\/GYW2EV9FEMJ3JJGR?locale=en-US\" target=\"_blank\" rel=\"noopener\">pol\u00edtica de devoluci\u00f3n de art\u00edculos<\/a> afirma literalmente<strong>, “<em>Los materiales peligrosos, incluidos los l\u00edquidos o gases inflamables, no son retornables.<\/em>.”<\/strong><\/p>\n\n\n\n<p style=\"font-size:17px\">As\u00ed que si realmente recibes un paquete cubierto de aceite o lodo negro grasiento, no tendr\u00edas que devolverlo, pero podr\u00edas obtener un reembolso. Si, por alguna raz\u00f3n, eso te ocurriera realmente, Amazon probablemente te pedir\u00eda una prueba fotogr\u00e1fica. \u00bfVes el problema que pueden plantear las im\u00e1genes generadas por IA cuando son indistinguibles de las reales?<\/p>\n\n\n\n<h2 class=\"wp-block-heading has-text-align-left\">#4. \"Cockroach Meal\" (generado con Nano Banana)<\/h2>\n\n\n\n<p style=\"font-size:17px\">La siguiente imagen de IA podr\u00eda hacerte perder el apetito. Resulta que hay casos documentados de personas que utilizan IA para recibir <a href=\"https:\/\/dailydot.com\/ai-generated-undercooked-food-doordash\" data-type=\"link\" data-id=\"https:\/\/dailydot.com\/ai-generated-undercooked-food-doordash\" target=\"_blank\" rel=\"noopener\">devoluciones fraudulentas<\/a> en aplicaciones de reparto de comida. Para esta cuarta prueba, segu\u00ed los mismos pasos que en la tercera. Esta vez, tom\u00e9 una foto de mi caja vac\u00eda de comida para llevar despu\u00e9s de comer.<\/p>\n\n\n\n<p><strong>Aqu\u00ed est\u00e1 la imagen real:<\/strong><\/p>\n\n\n\n<figure class=\"wp-block-image size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"768\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-768x1024.jpg\" alt=\"Una foto de una caja vac\u00eda de comida para llevar.\" class=\"wp-image-5296\" style=\"width:462px;height:auto\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-768x1024.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-225x300.jpg 225w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-scaled.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/IMG_6145-9x12.jpg 9w\" sizes=\"auto, (max-width: 768px) 100vw, 768px\" \/><\/figure>\n\n\n\n<p style=\"font-size:17px\">Luego puse la foto real que tom\u00e9 en Nano Banana y le di la siguiente indicaci\u00f3n:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-e9f00ca4e893e7742bce463148a639f5\"><strong><em>\"Edita esta imagen. No cambies nada excepto a\u00f1adir fideos donde est\u00e1 el papel de aluminio y a\u00f1adir un mont\u00f3n de peque\u00f1as cucarachas beb\u00e9 en la comida. Haz que todo parezca realista y no caricaturesco\". <\/em><\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\">La salida Nano Banana:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"747\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal.jpg\" alt=\"Una foto alterada por la IA de la caja de comida para llevar con fideos y bichos dentro de la caja\" class=\"wp-image-5295\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal.jpg 747w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal-219x300.jpg 219w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Cockroach-meal-9x12.jpg 9w\" sizes=\"auto, (max-width: 747px) 100vw, 747px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 12%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edici\u00f3n Deepfake<\/span><\/td>\n <td>Fraude<\/td>\n <td><span class=\"score-pill score-pass\">97.15% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">18.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">85.87% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">72.29% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Una vez m\u00e1s, TruthScan detect\u00f3 con seguridad los medios generados por la IA. Sight Engine y WasitAI fallaron por completo. AI or Not fue el segundo detector m\u00e1s preciso (con una certeza de IA de 85%) al escanear esta imagen de IA, y Winston obtuvo mejores resultados que en la tercera prueba (con una certeza de IA de 72%).<\/p>\n\n\n\n<p style=\"font-size:17px\">En esta imagen, los fideos y las cucarachas fueron generados en su totalidad por la IA; sin embargo, dos de los detectores no parec\u00edan pensar lo mismo, y s\u00f3lo uno (TruthScan) estaba por encima de 90% seguro de la manipulaci\u00f3n por parte de la IA.<\/p>\n\n\n\n<p style=\"font-size:17px\">Un punto de preocupaci\u00f3n aqu\u00ed fue lo sencillo que fue crear esta imagen falsa. Desde subir la foto que tom\u00e9 hasta pedirle a Nano Banana que lo hiciera y recibir el resultado, en total, s\u00f3lo tard\u00e9 unos 2 minutos. Una pesadilla para las aplicaciones de reparto de comida; un sue\u00f1o para los mentirosos hambrientos. Un \"factor fraude\".<\/p>\n\n\n\n<p class=\"has-vivid-red-color has-text-color has-link-color wp-elements-4c89b6ee3afcbdd6b6617c150e5b2e03\"><strong>DIVULGACI\u00d3N: Algunas de las im\u00e1genes generadas por IA incluidas en las siguientes pruebas pretend\u00edan demostrar c\u00f3mo pueden utilizarse las im\u00e1genes de IA para desinformar. Contienen im\u00e1genes controvertidas y temas delicados<\/strong>, <strong>y NO pretenden representar a ning\u00fan partido pol\u00edtico en particular.<\/strong> <strong>opiniones o ideolog\u00edas.<\/strong><\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><strong>#5<\/strong>. \"Autovandalismo\" (generado con Nano Banana)<\/h2>\n\n\n\n<p style=\"font-size:17px\">He encontrado una imagen real de un coche a la venta en l\u00ednea. Bastante mundano. Pero, \u00bfy si alguien quisiera utilizar esta imagen para difundir desinformaci\u00f3n o presentar una reclamaci\u00f3n falsa al seguro del coche? Por desgracia, la IA generativa lo hace f\u00e1cil y r\u00e1pido.<\/p>\n\n\n\n<p style=\"font-size:17px\"><strong>Para esta quinta prueba, tom\u00e9 una imagen real de un coche y le di a Nano Banana la siguiente instrucci\u00f3n:<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-d06074fe9533ce61c08bcf9750cd398d\">“<strong><em>No cambies nada del coche, excepto que quiero que le a\u00f1adas pintura en spray que ponga 'Palestina libre' y que el parabrisas est\u00e9 rajado. Los faros delanteros est\u00e1n destrozados y el retrovisor est\u00e1 roto y colgando y las dos ventanillas del lateral del coche est\u00e1n<\/em><\/strong> <strong>aplastado<\/strong>.”<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h3 class=\"wp-block-heading\"><strong>Aqu\u00ed est\u00e1 la imagen real:<\/strong><\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"940\" height=\"627\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2.webp\" alt=\"Una foto de un coche azul\" class=\"wp-image-5298\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2.webp 940w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-300x200.webp 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-768x512.webp 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/2000_honda_civic-si_2000_honda_civic-si_bbeed9f1-7b50-4314-a504-40be9411b8ab-mXhf1y-2-18x12.webp 18w\" sizes=\"auto, (max-width: 940px) 100vw, 940px\" \/><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">Imagen de Nano Banana AI:<\/h3>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"683\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage.jpg\" alt=\"Una foto del coche azul alterada por la IA, que muestra un parabrisas agrietado y da\u00f1os de pintura en spray en el lateral del coche.\" class=\"wp-image-5299\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-300x200.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-768x512.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Protest-Car-Damage-18x12.jpg 18w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Edici\u00f3n Deepfake<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Fraude \/ Desinformaci\u00f3n<\/td>\n <td><span class=\"score-pill score-pass\">97.48% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">18.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">89.52% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0,02% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Los resultados del detector para esta salida fueron mixtos. En esta prueba, WasItAI (sorprendentemente) obtuvo la puntuaci\u00f3n m\u00e1s alta en detecci\u00f3n de IA (99%). TruthScan obtuvo la segunda mejor puntuaci\u00f3n (97%). Sight Engine fall\u00f3 significativamente a la hora de detectar elementos sint\u00e9ticos en la imagen, y Winston fue el peor de todos los detectores (asignando una probabilidad de implicaci\u00f3n inferior a 1%). AI or Not lo hizo bien, pero la precisi\u00f3n sigui\u00f3 siendo inferior a 90%.<br><br>Hablemos de la imagen. Los dos aspectos m\u00e1s preocupantes de este tipo de contenido alterado o creado con IA son que ambos implican enga\u00f1o. En primer lugar, el uso de IA para alterar im\u00e1genes mundanas con el fin de crear contenido con carga pol\u00edtica es r\u00e1pido, f\u00e1cil y parece real. Tanto si lo que se busca con este tipo de contenido es captar la atenci\u00f3n como si se trata de provocadores pol\u00edticos, el problema es que se trata de informaci\u00f3n falsa presentada como aut\u00e9ntica.<\/p>\n\n\n\n<p style=\"font-size:17px\">La segunda forma de abusar de este tipo de herramientas de generaci\u00f3n de im\u00e1genes es el fraude. Si un estafador quiere presentar una reclamaci\u00f3n falsa al seguro del autom\u00f3vil, tendr\u00e1 que presentar pruebas falsas. En lugar de pasar horas en Photoshop editando im\u00e1genes y manipulando pruebas, pueden utilizar generadores de im\u00e1genes para producir r\u00e1pidamente pruebas falsas.<\/p>\n\n\n\n<p style=\"font-size:17px\">En el ejemplo de la imagen de IA anterior, observe c\u00f3mo un estafador podr\u00eda utilizar un proceso similar para presentar una reclamaci\u00f3n falsa de acuerdo con Progressive's <a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-vandalism\/\" data-type=\"link\" data-id=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-vandalism\/\" target=\"_blank\" rel=\"noopener\">Pol\u00edtica de cobertura global<\/a>que cubre espec\u00edficamente:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Acuchillado o <a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-tire-damage\/\" target=\"_blank\" rel=\"noopener\">neum\u00e1ticos da\u00f1ados<\/a><\/strong><\/li>\n\n\n\n<li><strong>Cristales, faros o pilotos traseros rotos<\/strong><\/li>\n\n\n\n<li><strong>Da\u00f1os por pintura en aerosol<\/strong><\/li>\n\n\n\n<li><strong><a href=\"https:\/\/www.progressive.com\/answers\/does-car-insurance-cover-scratches\/\" target=\"_blank\" rel=\"noopener\">Abolladuras o ara\u00f1azos<\/a> de alguien rayando tu coche<\/strong><\/li>\n\n\n\n<li><strong>Echar az\u00facar u otras sustancias en el dep\u00f3sito de gasolina<\/strong><\/li>\n<\/ul>\n\n\n\n<p>Por supuesto, presentar pruebas falsas no s\u00f3lo es enga\u00f1oso, sino tambi\u00e9n ilegal. En un mundo ideal, nadie infringir\u00eda la ley; en realidad, se cometen delitos todos los d\u00edas. Y sin una detecci\u00f3n fiable, las compa\u00f1\u00edas de seguros de autom\u00f3viles corren el riesgo de ser estafadas por millones a medida que aumentan las reclamaciones de seguros fraudulentas basadas en IA.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#6. \"Soda Musk y Don\" (generado con Nano Banana)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"559\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1024x559.jpeg\" alt=\"Una imagen generada por IA de Elon Musk y Donald Trump\" class=\"wp-image-5300\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1024x559.jpeg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-300x164.jpeg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-768x419.jpeg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-1536x838.jpeg 1536w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-2048x1117.jpeg 1877w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-18x10.jpeg 18w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3-Elon-Spilling-Soda-scaled.jpeg 1878w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Desinfo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0,24% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">86.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Una vez m\u00e1s, utilic\u00e9 Nano Banana para esta prueba, y el objetivo era demostrar im\u00e1genes falsas, pol\u00edticamente cargadas con un trasfondo m\u00e1s tonto. TruthScan, Sight Engine y AIorNot marcaron la imagen generada con una puntuaci\u00f3n de detecci\u00f3n de IA de 99%. WasitAI no se mostr\u00f3 tan seguro, ya que s\u00f3lo detect\u00f3 la imagen generada por la IA con una certeza de 86%. Y Winston fall\u00f3 completamente, asignando una puntuaci\u00f3n de IA de 0.2%. Para generar la imagen, le di a Nano Banana este mensaje:<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p class=\"has-vivid-cyan-blue-color has-text-color has-link-color wp-elements-4f7e002ecaefec24bbebdca3101fff42\"><strong>“<em>Genera una foto de Elon Musk derramando un gran vaso de refresco como un Big Gulp sobre su camisa y flipando mientras Donald Trump est\u00e1 sentado a su lado ri\u00e9ndose. Est\u00e1n en un avi\u00f3n a reacci\u00f3n. Esa es la escena. Y la foto debe parecer realista como si hubiera sido tomada con una c\u00e1mara instant\u00e1nea e incluir artefactos como el flash de la c\u00e1mara reflej\u00e1ndose en la ventanilla del avi\u00f3n y fuera del avi\u00f3n. Es de noche. Debe tener un aspecto hiperrealista.<\/em>“<\/strong><\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p style=\"font-size:17px\">Cada vez que uno de estos programas de software no detecta IA, lo etiquetan como aut\u00e9ntico. Por muy tonto que sea este ejemplo de imagen de la prueba, no hay nada divertido en pensar que un deepfake es real.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#7. \"Still Alive\" (generado por Nano Banana)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"765\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo.jpg\" alt=\"Imagen generada por IA de un hombre con un cartel que dice "ayuda".\" class=\"wp-image-5301\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo.jpg 765w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo-224x300.jpg 224w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/GV3Christian_Ransom_Photo-9x12.jpg 9w\" sizes=\"auto, (max-width: 765px) 100vw, 765px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Deepfake<\/td>\n <td><span class=\"score-pill score-pass\">97.49% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">95.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.27% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">0,15% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">\u00bfTe parece inquietante esta imagen? Para generar este deepfake, utilic\u00e9 Nano Banana y le di una foto m\u00eda. A continuaci\u00f3n, le di un mensaje largo para ponerme en una situaci\u00f3n que diera a entender que hab\u00eda sido secuestrado. Tal vez sea porque esta foto muestra <em>me<\/em> en una posici\u00f3n comprometida. \u00bfQu\u00e9 pasar\u00eda si alguien hiciera algo as\u00ed, se lo enviara a mi madre y amenazara con hacerme da\u00f1o a menos que le enviara dinero? Qu\u00e9 miedo. <\/p>\n\n\n\n<p class=\"has-small-font-size\"><em>(La indicaci\u00f3n completa que utilic\u00e9 para ello se encuentra en el conjunto de datos de prueba completo al final de este art\u00edculo).<\/em><\/p>\n\n\n\n<p style=\"font-size:17px\"><br>Por suerte para esta imagen, a pesar de tener un aspecto muy realista, <strong>tres de cada cinco detectores lo identificaron como generado por IA.<\/strong> <\/p>\n\n\n\n<p style=\"font-size:17px\">AI or Not obtuvo una puntuaci\u00f3n de 99%, TruthScan de 97% y Sight Engine de 95%. Por desgracia, Winston y WasItAI clasificaron la imagen como real.<\/p>\n\n\n\n<p style=\"font-size:17px\">Debo aclarar que estas pruebas no se hicieron en mi cuenta personal de Gemini. Nano Banana nunca me pidi\u00f3 que demostrara que yo era la persona de la imagen. Cualquiera podr\u00eda haber descargado una foto m\u00eda de Internet, meterla en Nano Banana y hacer que creara este tipo de imagen. <\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#8. \"Abuso de un Ben\" (generado con Nano Banana)<\/h2>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full\"><img loading=\"lazy\" decoding=\"async\" width=\"200\" height=\"200\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1.jpg\" alt=\"\" class=\"wp-image-5307\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1.jpg 200w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1-150x150.jpg 150w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/c945d528-730f-48c1-bdca-30e5799fc1aa_thumb-1-12x12.jpg 12w\" sizes=\"auto, (max-width: 200px) 100vw, 200px\" \/><\/figure>\n<\/div>\n\n\n<p style=\"font-size:17px\">Conoce a Ben. Pasa la mayor parte de su tiempo dirigiendo varios departamentos en las empresas en las que trabaja. La imagen de arriba es el (verdadero) elegante retrato corporativo de Ben. Por si no se nota, Ben es un tipo profesional. Se guarda para s\u00ed las opiniones pol\u00edticas que tiene, y sus conversaciones de trabajo son s\u00f3lo sobre trabajo. Pero en un momento dado, la imagen de Ben podr\u00eda utilizarse para difundir propaganda o mensajes pol\u00edticos sin su consentimiento.<\/p>\n\n\n\n<p style=\"font-size:17px\">Recientemente se ha informado mucho sobre c\u00f3mo se han difundido en Internet v\u00eddeos generados por IA en los que aparecen agentes del Servicio de Inmigraci\u00f3n y Control de Aduanas (ICE) y protestas. Bas\u00e1ndome en todas las pruebas realizadas hasta el momento, llegu\u00e9 a la conclusi\u00f3n de que crear propaganda pol\u00edtica falsa con IA era realmente f\u00e1cil. Decid\u00ed hacer otra prueba, esta vez con Ben. <\/p>\n\n\n\n<p style=\"font-size:17px\">Con el permiso de Ben, utilic\u00e9 su retrato para crear un deepfake. <strong>Aqu\u00ed est\u00e1 la imagen generada por IA que cre\u00f3 Nano Banana:<\/strong><\/p>\n\n\n<div class=\"wp-block-image\">\n<figure class=\"aligncenter size-full is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"1024\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x.jpg\" alt=\"Una foto alterada por la IA de un hombre llamado Ben, sosteniendo un cartel que dice Abolish ICE.\" class=\"wp-image-5308\" style=\"width:537px;height:auto\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-300x300.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-150x150.jpg 150w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-768x768.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/Gemini_Generated_Image_qd4x5sqd4x5sqd4x-12x12.jpg 12w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n<\/div>\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>Nano Banana<\/td>\n <td style=\"font-size:0.78rem;\">Deepfake \/ Desinformaci\u00f3n<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">97.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">98.88% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">5.74% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Me llev\u00f3 treinta segundos. Todo lo que tuve que hacer fue arrastrar y soltar su foto de la cabeza en Gemini y le ped\u00ed que \"levantara un cartel que dijera abolir ICE\". El resultado de Nano Banana no es tan fotorrealista como los dem\u00e1s, pero sigue pareciendo lo bastante aut\u00e9ntico como para resultar persuasivo.<\/p>\n\n\n\n<p style=\"font-size:17px\"><span style=\"box-sizing: border-box; margin: 0px; padding: 0px;\">Afortunadamente, cuatro de los detectores identificaron la imagen con precisi\u00f3n:\u00a0<strong>TruthScan<\/strong> (9<\/span>9% AI), Was It AI (99% AI), Sight Engine (97% AI), y AI or Not (97% AI). Winston AI fall\u00f3, asignando s\u00f3lo una puntuaci\u00f3n de 5% AI.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#9. \"Club Void\" (generado por Midjourney)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"771\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-1024x771.jpg\" alt=\"Una mujer sonriente con un vestido blanco de manga corta est\u00e1 de pie en una habitaci\u00f3n poco iluminada y llena de gente. La foto parece tomada con flash, lo que crea una atm\u00f3sfera brumosa y azulada con tenues part\u00edculas visibles en el aire. Otras personas vestidas con ropa informal est\u00e1n de pie en el fondo, bajo un techo da\u00f1ado o de aspecto industrial.\" class=\"wp-image-5309\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-1024x771.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-300x226.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-768x578.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club-16x12.jpg 16w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-Girl-in-club.jpg 1232w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>A mitad de viaje <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td>General<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">89.21% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">1.56% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">70.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Gener\u00e9 la imagen de arriba con Midjourney. Cuanto m\u00e1s la miraba, m\u00e1s inquietante me parec\u00eda. La est\u00e9tica es oscura e inquietante. Una presencia sombr\u00eda parece surgir de detr\u00e1s del sujeto. Para crear esta imagen, utilic\u00e9 la funci\u00f3n de referencia est\u00e9tica de Midjourney y simplemente le ped\u00ed que fuera una \"chica de pie en un club sonriendo\". Mostr\u00e9 la imagen a varias personas y todas pensaron que era real. <br><br>La mayor\u00eda de los detectores marcaron la imagen como generada por IA. TruthScan y Sight Engine la marcaron con una puntuaci\u00f3n de IA de 99%, AI or Not dijo 89%. <\/p>\n\n\n\n<p style=\"font-size:17px\">WasItAI no estaba tan seguro como los detectores, clasificando la imagen con s\u00f3lo una puntuaci\u00f3n AI de 70%. <\/p>\n\n\n\n<p style=\"font-size:17px\">El detector de Winston AI fall\u00f3 completamente, dando a la imagen una puntuaci\u00f3n AI de 1,56%. <\/p>\n\n\n\n<h2 class=\"wp-block-heading\">#10 \"Hielera\"<strong> <\/strong>(generado por Midjourney)<\/h2>\n\n\n\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"771\" src=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-1024x771.jpg\" alt=\"Una foto borrosa que muestra a un hombre con una camisa manchada de rojo, de pie delante de alguien.\" class=\"wp-image-5311\" title=\"\" srcset=\"https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-1024x771.jpg 1024w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-300x226.jpg 300w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-768x578.jpg 768w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE-16x12.jpg 16w, https:\/\/research.undetectable.ai\/wp-content\/uploads\/2026\/02\/MJ-ICE.jpg 1232w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n .detector-scores-wrap {\n max-width: 780px;\n margin: 2rem auto;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n }\n\n .detector-scores-wrap h3 {\n font-family: 'Outfit', sans-serif;\n font-size: 1.4rem;\n font-weight: 800;\n color: #1a1a2e;\n margin-bottom: 1.25rem;\n letter-spacing: -0.02em;\n }\n\n .detector-scores-table {\n width: 100%;\n border-collapse: separate;\n border-spacing: 0;\n border-radius: 14px;\n overflow: hidden;\n table-layout: fixed;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n\n .detector-scores-table thead th {\n background: #1a1a2e;\n color: #ffffff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.7rem;\n text-transform: uppercase;\n letter-spacing: 0.04em;\n padding: 14px 6px;\n text-align: center;\n border: none;\n }\n\n .detector-scores-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 18%;\n }\n\n .detector-scores-table thead th:nth-child(2) {\n width: 14%;\n }\n\n .detector-scores-table tbody tr {\n transition: background 0.2s ease;\n }\n\n .detector-scores-table tbody tr:hover {\n background: rgba(0,0,0,0.02);\n }\n\n .detector-scores-table tbody td {\n padding: 14px 6px;\n text-align: center;\n font-size: 0.9rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #eee;\n }\n\n .detector-scores-table tbody tr:last-child td {\n border-bottom: none;\n }\n\n .detector-scores-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n }\n\n .score-pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.78rem;\n border-radius: 6px;\n padding: 5px 8px;\n display: inline-block;\n min-width: 0;\n text-align: center;\n white-space: nowrap;\n }\n\n .score-pass {\n background: #22c55e;\n color: #fff;\n }\n\n .score-warn {\n background: #fb923c;\n color: #fff;\n }\n\n .score-fail {\n background: #ef4444;\n color: #fff;\n }\n\n @media (max-width: 600px) {\n .detector-scores-table thead th,\n .detector-scores-table tbody td {\n padding: 10px 4px;\n font-size: 0.7rem;\n }\n .score-pill {\n padding: 4px 5px;\n font-size: 0.68rem;\n }\n .detector-scores-wrap h3 {\n font-size: 1.15rem;\n }\n }\n<\/style>\n\n<div class=\"detector-scores-wrap\">\n <h3>Puntuaci\u00f3n de los detectores de im\u00e1genes<\/h3>\n\n <table class=\"detector-scores-table\">\n <thead>\n <tr>\n <th>Modelo de imagen<\/th>\n <th>Categor\u00eda<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n <tr>\n <td>A mitad de viaje <span style=\"font-size:0.75rem; color:#888; display:block;\">Generaci\u00f3n completa<\/span><\/td>\n <td style=\"font-size:0.82rem;\">Desinfo<\/td>\n <td><span class=\"score-pill score-pass\">99.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">92.00% AI<\/span><\/td>\n <td><span class=\"score-pill score-pass\">97.94% AI<\/span><\/td>\n <td><span class=\"score-pill score-fail\">41.20% AI<\/span><\/td>\n <td><span class=\"score-pill score-warn\">72.00% AI<\/span><\/td>\n <\/tr>\n <\/tbody>\n <\/table>\n<\/div>\n\n\n\n<p style=\"font-size:17px\">Cre\u00e9 esta imagen utilizando el mismo m\u00e9todo de medio viaje de la prueba anterior. Esta vez, el objetivo era generar algo un poco m\u00e1s oscuro. TruthScan fue el detector m\u00e1s preciso, etiquetando la imagen como 99% AI. Sight Engine y AI or Not detectaron correctamente la imagen. Winston AI fall\u00f3 por completo, y WasItAI tuvo una puntuaci\u00f3n de confianza mucho m\u00e1s baja que TruthScan, Sight Engine y AI or Not.<\/p>\n\n\n\n<p style=\"font-size:17px\">Debo decir que, en este caso, la imagen que he generado podr\u00eda utilizarse o enmarcarse de dos maneras. Por un lado, alguien podr\u00eda utilizar este tipo de medio (o cualquier otro de tem\u00e1tica pol\u00edtica o de disturbios civiles) en alg\u00fan tipo de proyecto art\u00edstico o creativo para hacer una declaraci\u00f3n. En tal caso, el riesgo parece menor. Lo que m\u00e1s me preocupa de esta categor\u00eda de medios generados por IA es que los utilicen individuos malintencionados que afirmen que son reales.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Recapitulaci\u00f3n y resultados: Los mejores detectores de im\u00e1genes AI<\/h2>\n\n\n\n<p style=\"font-size:17px\">Bien, ahora que hemos visualizado todas las pruebas, voy a recapitular: He generado 10 im\u00e1genes AI<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li style=\"font-size:17px\">Utilic\u00e9 ChatGPT, Nano Banana y Midjourney para generar 10 im\u00e1genes de IA<\/li>\n\n\n\n<li style=\"font-size:17px\">He probado cinco detectores de im\u00e1genes de IA haciendo pasar por ellos todas las im\u00e1genes de IA que he generado.<\/li>\n\n\n\n<li style=\"font-size:17px\">TruthScan super\u00f3 todas las pruebas y fue el detector m\u00e1s preciso. AI or Not super\u00f3 8 de las 10 pruebas, mostrando cierta fiabilidad. Sight Engine fall\u00f3 3 de las 10 pruebas y demostr\u00f3 una precisi\u00f3n cuestionable. Was It AI no super\u00f3 4 de las 10 pruebas y su precisi\u00f3n fue deficiente en general. Winston AI fue el detector de im\u00e1genes menos preciso, s\u00f3lo super\u00f3 3 de 10 pruebas, clasificando err\u00f3neamente las im\u00e1genes.<\/li>\n<\/ul>\n\n\n\n<!DOCTYPE html>\n<html lang=\"en\">\n<head>\n<meta charset=\"UTF-8\">\n<meta name=\"viewport\" content=\"width=device-width, initial-scale=1.0\">\n<title>AI Image Detector - Resultados completos de las pruebas<\/title>\n<style>\n @import url('https:\/\/fonts.googleapis.com\/css2?family=DM+Sans:wght@400;500;600;700&family=Outfit:wght@600;700;800&display=swap');\n\n *,*::before,*::after{box-sizing:border-box;margin:0;padding:0}\n\n body {\n background: #f4f3f0;\n min-height: 100vh;\n font-family: 'DM Sans', sans-serif;\n -webkit-font-smoothing: antialiased;\n color: #2d2d3a;\n }\n\n \/* \u2500\u2500 wrapper \u2500\u2500 *\/\n .results-page {\n max-width: 860px;\n margin: 0 auto;\n padding: 3rem 1.25rem 4rem;\n }\n\n \/* \u2500\u2500 hero header \u2500\u2500 *\/\n .hero {\n text-align: center;\n margin-bottom: 2.75rem;\n }\n .hero-badge {\n display: inline-block;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.65rem;\n text-transform: uppercase;\n letter-spacing: 0.12em;\n color: #fff;\n background: #1a1a2e;\n padding: 6px 16px;\n border-radius: 100px;\n margin-bottom: 1.1rem;\n }\n .hero h1 {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 2.4rem;\n color: #1a1a2e;\n letter-spacing: -0.03em;\n line-height: 1.15;\n margin-bottom: 0.6rem;\n }\n .hero p {\n font-size: 1.02rem;\n color: #6b6b7b;\n max-width: 560px;\n margin: 0 auto;\n line-height: 1.55;\n }\n\n \/* \u2500\u2500 accuracy cards row \u2500\u2500 *\/\n .accuracy-cards {\n display: grid;\n grid-template-columns: repeat(5, 1fr);\n gap: 10px;\n margin-bottom: 2.75rem;\n }\n .acc-card {\n background: #fff;\n border-radius: 14px;\n padding: 1.15rem 0.6rem 1rem;\n text-align: center;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.05),\n 0 6px 20px rgba(0,0,0,0.07),\n 0 0 0 1px rgba(0,0,0,0.03);\n position: relative;\n overflow: hidden;\n transition: transform 0.2s ease, box-shadow 0.2s ease;\n }\n .acc-card:hover {\n transform: translateY(-3px);\n box-shadow:\n 0 2px 6px rgba(0,0,0,0.06),\n 0 12px 32px rgba(0,0,0,0.1),\n 0 0 0 1px rgba(0,0,0,0.04);\n }\n .acc-card::before {\n content: '';\n position: absolute;\n top: 0; left: 0; right: 0;\n height: 4px;\n }\n .acc-card.rank-1::before { background: #16a34a; }\n .acc-card.rank-2::before { background: #a3e635; }\n .acc-card.rank-3::before { background: #facc15; }\n .acc-card.rank-4::before { background: #fb923c; }\n .acc-card.rank-5::before { background: #ef4444; }\n\n .acc-card .rank-num {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 0.62rem;\n text-transform: uppercase;\n letter-spacing: 0.08em;\n color: #999;\n margin-bottom: 0.35rem;\n }\n .acc-card .tool-name {\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.92rem;\n color: #1a1a2e;\n margin-bottom: 0.65rem;\n line-height: 1.2;\n }\n .acc-ring {\n width: 72px; height: 72px;\n margin: 0 auto 0.55rem;\n position: relative;\n }\n .acc-ring svg { transform: rotate(-90deg); }\n .acc-ring .ring-bg { fill: none; stroke: #eee; stroke-width: 5; }\n .acc-ring .ring-fg { fill: none; stroke-width: 5; stroke-linecap: round; transition: stroke-dashoffset 1s ease; }\n .acc-ring .ring-label {\n position: absolute;\n inset: 0;\n display: flex;\n align-items: center;\n justify-content: center;\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.1rem;\n color: #1a1a2e;\n }\n .acc-card .acc-fraction {\n font-size: 0.78rem;\n color: #888;\n font-weight: 500;\n }\n\n \/* \u2500\u2500 section titles \u2500\u2500 *\/\n .section-title {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.45rem;\n color: #1a1a2e;\n letter-spacing: -0.02em;\n margin-bottom: 1rem;\n }\n .section-subtitle {\n font-size: 0.88rem;\n color: #888;\n margin-bottom: 1.25rem;\n line-height: 1.5;\n }\n\n \/* \u2500\u2500 master table \u2500\u2500 *\/\n .master-table-wrap {\n border-radius: 14px;\n overflow: hidden;\n box-shadow:\n 0 1px 3px rgba(0,0,0,0.06),\n 0 8px 24px rgba(0,0,0,0.08),\n 0 0 0 1px rgba(0,0,0,0.04);\n margin-bottom: 2.75rem;\n background: #fff;\n }\n .master-table {\n width: 100%;\n border-collapse: collapse;\n table-layout: fixed;\n }\n .master-table thead th {\n background: #1a1a2e;\n color: #fff;\n font-family: 'Outfit', sans-serif;\n font-weight: 700;\n font-size: 0.62rem;\n text-transform: uppercase;\n letter-spacing: 0.05em;\n padding: 13px 5px;\n text-align: center;\n border: none;\n position: sticky;\n top: 0;\n z-index: 2;\n }\n .master-table thead th:first-child {\n text-align: left;\n padding-left: 14px;\n width: 14%;\n }\n .master-table thead th:nth-child(2) { width: 13%; }\n .master-table thead th:nth-child(3) { width: 9%; }\n\n .master-table tbody tr {\n transition: background 0.18s ease;\n }\n .master-table tbody tr:hover {\n background: rgba(26,26,46,0.025);\n }\n .master-table tbody td {\n padding: 11px 5px;\n text-align: center;\n font-size: 0.82rem;\n font-weight: 500;\n color: #2d2d3a;\n border-bottom: 1px solid #f0f0ec;\n }\n .master-table tbody tr:last-child td { border-bottom: none; }\n .master-table tbody td:first-child {\n text-align: left;\n padding-left: 14px;\n font-weight: 600;\n color: #1a1a2e;\n font-size: 0.78rem;\n line-height: 1.3;\n }\n .test-num {\n display: inline-block;\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 0.6rem;\n background: #1a1a2e;\n color: #fff;\n width: 18px; height: 18px;\n line-height: 18px;\n text-align: center;\n border-radius: 5px;\n margin-right: 5px;\n vertical-align: middle;\n }\n .img-model-sub {\n display: block;\n font-size: 0.68rem;\n color: #999;\n font-weight: 400;\n margin-top: 1px;\n }\n .cat-badge {\n font-family: 'Outfit', sans-serif;\n font-size: 0.6rem;\n font-weight: 700;\n text-transform: uppercase;\n letter-spacing: 0.03em;\n padding: 3px 6px;\n border-radius: 5px;\n display: inline-block;\n white-space: normal;\n line-height: 1.35;\n max-width: 100%;\n }\n .cat-fraud { background: #fef2f2; color: #dc2626; }\n .cat-disinfo { background: #fefce8; color: #a16207; }\n .cat-general { background: #f0fdf4; color: #16a34a; }\n .cat-deepfake { background: #faf5ff; color: #7c3aed; }\n .cat-mixed { background: #fff7ed; color: #c2410c; }\n\n \/* score pills *\/\n .pill {\n font-weight: 700;\n font-family: 'Outfit', sans-serif;\n font-size: 0.7rem;\n border-radius: 6px;\n padding: 4px 7px;\n display: inline-block;\n text-align: center;\n white-space: nowrap;\n min-width: 68px;\n }\n .pill-pass { background: #22c55e; color: #fff; }\n .pill-warn { background: #fb923c; color: #fff; }\n .pill-fail { background: #ef4444; color: #fff; }\n\n \/* \u2500\u2500 legend \u2500\u2500 *\/\n .legend {\n display: flex;\n gap: 1.25rem;\n align-items: center;\n margin-bottom: 1.25rem;\n flex-wrap: wrap;\n }\n .legend-item {\n display: flex;\n align-items: center;\n gap: 6px;\n font-size: 0.78rem;\n font-weight: 500;\n color: #666;\n }\n .legend-dot {\n width: 12px; height: 12px;\n border-radius: 4px;\n }\n .legend-dot.lg-pass { background: #22c55e; }\n .legend-dot.lg-warn { background: #fb923c; }\n .legend-dot.lg-fail { background: #ef4444; }\n\n \/* \u2500\u2500 bottom summary strip \u2500\u2500 *\/\n .summary-strip {\n display: grid;\n grid-template-columns: repeat(3, 1fr);\n gap: 10px;\n margin-bottom: 2.5rem;\n }\n .strip-card {\n background: #1a1a2e;\n border-radius: 12px;\n padding: 1.1rem 1rem;\n color: #fff;\n text-align: center;\n }\n .strip-card .strip-val {\n font-family: 'Outfit', sans-serif;\n font-weight: 800;\n font-size: 1.65rem;\n letter-spacing: -0.02em;\n margin-bottom: 0.15rem;\n }\n .strip-card .strip-label {\n font-size: 0.72rem;\n color: rgba(255,255,255,0.55);\n font-weight: 500;\n text-transform: uppercase;\n letter-spacing: 0.05em;\n }\n .strip-val .hl-green { color: #4ade80; }\n .strip-val .hl-red { color: #f87171; }\n .strip-val .hl-amber { color: #fbbf24; }\n\n \/* \u2500\u2500 footer note \u2500\u2500 *\/\n .foot-note {\n font-size: 0.78rem;\n color: #999;\n line-height: 1.6;\n text-align: center;\n max-width: 640px;\n margin: 0 auto;\n }\n .foot-note strong { color: #666; font-weight: 600; }\n\n \/* \u2500\u2500 responsive \u2500\u2500 *\/\n @media (max-width: 700px) {\n .hero h1 { font-size: 1.6rem; }\n .accuracy-cards { grid-template-columns: repeat(2, 1fr); }\n .accuracy-cards .acc-card:last-child { grid-column: span 2; justify-self: center; max-width: 200px; }\n .summary-strip { grid-template-columns: 1fr; }\n .master-table thead th,\n .master-table tbody td { padding: 9px 3px; font-size: 0.68rem; }\n .pill { font-size: 0.6rem; padding: 3px 4px; min-width: 54px; }\n .test-num { width: 15px; height: 15px; line-height: 15px; font-size: 0.52rem; }\n .section-title { font-size: 1.15rem; }\n }\n\n \/* \u2500\u2500 entrance animations \u2500\u2500 *\/\n @keyframes fadeUp {\n from { opacity: 0; transform: translateY(18px); }\n to { opacity: 1; transform: translateY(0); }\n }\n .anim { animation: fadeUp 0.6s ease both; }\n .anim-d1 { animation-delay: 0.08s; }\n .anim-d2 { animation-delay: 0.16s; }\n .anim-d3 { animation-delay: 0.24s; }\n .anim-d4 { animation-delay: 0.32s; }\n .anim-d5 { animation-delay: 0.4s; }\n<\/style>\n<\/head>\n<body>\n\n<div class=\"results-page\">\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 HERO \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <header class=\"hero anim\">\n <span class=\"hero-badge\">50 controles de detecci\u00f3n - 5 detectores - 10 im\u00e1genes AI<\/span>\n <h1>Resultados completos de las pruebas<\/h1>\n <p>Rendimiento de cinco populares detectores de im\u00e1genes por IA en casos de fraude, desinformaci\u00f3n, deepfakes y fotograf\u00eda en general.<\/p>\n <\/header>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 ACCURACY RING CARDS \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"accuracy-cards anim anim-d1\">\n\n <!-- 1 \u2014 TruthScan 100% -->\n <div class=\"acc-card rank-1\">\n <div class=\"rank-num\">#1<\/div>\n <div class=\"tool-name\">TruthScan<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#16a34a\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"0\"\/>\n <\/svg>\n <span class=\"ring-label\">100%<\/span>\n <\/div>\n <div class=\"acc-fraction\">10 \/ 10 detectados<\/div>\n <\/div>\n\n <!-- 2 \u2014 AI or Not 80% -->\n <div class=\"acc-card rank-2\">\n <div class=\"rank-num\">#2<\/div>\n <div class=\"tool-name\">IA o no<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#a3e635\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"42.72\"\/>\n <\/svg>\n <span class=\"ring-label\">80%<\/span>\n <\/div>\n <div class=\"acc-fraction\">8 \/ 10 detectados<\/div>\n <\/div>\n\n <!-- 3 \u2014 Sight Engine 70% -->\n <div class=\"acc-card rank-3\">\n <div class=\"rank-num\">#3<\/div>\n <div class=\"tool-name\">Motor de visi\u00f3n<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#facc15\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"64.08\"\/>\n <\/svg>\n <span class=\"ring-label\">70%<\/span>\n <\/div>\n <div class=\"acc-fraction\">7 \/ 10 detectados<\/div>\n <\/div>\n\n <!-- 4 \u2014 WasItAI 60% -->\n <div class=\"acc-card rank-4\">\n <div class=\"rank-num\">#4<\/div>\n <div class=\"tool-name\">WasItAI<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#fb923c\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"85.44\"\/>\n <\/svg>\n <span class=\"ring-label\">60%<\/span>\n <\/div>\n <div class=\"acc-fraction\">6 \/ 10 detectados<\/div>\n <\/div>\n\n <!-- 5 \u2014 Winston AI 30% -->\n <div class=\"acc-card rank-5\">\n <div class=\"rank-num\">#5<\/div>\n <div class=\"tool-name\">Winston AI<\/div>\n <div class=\"acc-ring\">\n <svg viewbox=\"0 0 80 80\" width=\"72\" height=\"72\">\n <circle class=\"ring-bg\" cx=\"40\" cy=\"40\" r=\"34\"\/>\n <circle class=\"ring-fg\" cx=\"40\" cy=\"40\" r=\"34\"\n stroke=\"#ef4444\"\n stroke-dasharray=\"213.6\"\n stroke-dashoffset=\"149.52\"\/>\n <\/svg>\n <span class=\"ring-label\">30%<\/span>\n <\/div>\n <div class=\"acc-fraction\">3 \/ 10 detectados<\/div>\n <\/div>\n\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 STATS STRIP \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"summary-strip anim anim-d2\">\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-green\">34<\/span> \/ 50<\/div>\n <div class=\"strip-label\">Total de detecciones correctas<\/div>\n <\/div>\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-red\">16<\/span> \/ 50<\/div>\n <div class=\"strip-label\">Total de detecciones omitidas<\/div>\n <\/div>\n <div class=\"strip-card\">\n <div class=\"strip-val\"><span class=\"hl-amber\">68%<\/span><\/div>\n <div class=\"strip-label\">Precisi\u00f3n media de todos los<\/div>\n <\/div>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 LEGEND \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <h3 class=\"section-title anim anim-d3\">Las 10 pruebas - Desglose completo<\/h3>\n <p class=\"section-subtitle anim anim-d3\">Cada puntuaci\u00f3n representa el \u00edndice de confianza AI del detector. Umbral de aprobaci\u00f3n: \u2265 90% AI.<\/p>\n\n <div class=\"legend anim anim-d3\">\n <div class=\"legend-item\"><span class=\"legend-dot lg-pass\"><\/span> Pase (\u2265 90%)<\/div>\n <div class=\"legend-item\"><span class=\"legend-dot lg-warn\"><\/span> Casi fallo (70-89%)<\/div>\n <div class=\"legend-item\"><span class=\"legend-dot lg-fail\"><\/span> Fallo (< 70%)<\/div>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 MASTER TABLE \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <div class=\"master-table-wrap anim anim-d4\">\n <table class=\"master-table\">\n <thead>\n <tr>\n <th>Prueba<\/th>\n <th>Categor\u00eda<\/th>\n <th>Fuente<\/th>\n <th>TruthScan<\/th>\n <th>Motor de visi\u00f3n<\/th>\n <th>IA o no<\/th>\n <th>Winston AI<\/th>\n <th>WasItAI<\/th>\n <\/tr>\n <\/thead>\n <tbody>\n\n <!-- #1 -->\n <tr>\n <td><span class=\"test-num\">1<\/span> Hombre en la cornisa<\/td>\n <td><span class=\"cat-badge cat-general\">General<\/span><\/td>\n <td><span class=\"img-model-sub\">ChatGPT<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">78.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.98%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #2 -->\n <tr>\n <td><span class=\"test-num\">2<\/span> El recibo falso<\/td>\n <td><span class=\"cat-badge cat-fraud\">Fraude<\/span><\/td>\n <td><span class=\"img-model-sub\">ChatGPT<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">19.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">94.48%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.04%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #3 -->\n <tr>\n <td><span class=\"test-num\">3<\/span> Paquete de peligros<\/td>\n <td><span class=\"cat-badge cat-fraud\">Fraude<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">98.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.10%<\/span><\/td>\n <td><span class=\"pill pill-fail\">31.98%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #4 -->\n <tr>\n <td><span class=\"test-num\">4<\/span> Comida para cucarachas<\/td>\n <td><span class=\"cat-badge cat-fraud\">Fraude<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.15%<\/span><\/td>\n <td><span class=\"pill pill-fail\">18.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">85.87%<\/span><\/td>\n <td><span class=\"pill pill-warn\">72.29%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #5 -->\n <tr>\n <td><span class=\"test-num\">5<\/span> Autovandalismo<\/td>\n <td><span class=\"cat-badge cat-mixed\">Fraude \/ Desinformaci\u00f3n<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.48%<\/span><\/td>\n <td><span class=\"pill pill-fail\">18.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">89.52%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.02%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #6 -->\n <tr>\n <td><span class=\"test-num\">6<\/span> Soda Musk & Don<\/td>\n <td><span class=\"cat-badge cat-disinfo\">Desinfo<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.24%<\/span><\/td>\n <td><span class=\"pill pill-warn\">86.00%<\/span><\/td>\n <\/tr>\n\n <!-- #7 -->\n <tr>\n <td><span class=\"test-num\">7<\/span> A\u00fan con vida<\/td>\n <td><span class=\"cat-badge cat-deepfake\">Deepfake<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.49%<\/span><\/td>\n <td><span class=\"pill pill-pass\">95.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.27%<\/span><\/td>\n <td><span class=\"pill pill-fail\">0.15%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.00%<\/span><\/td>\n <\/tr>\n\n <!-- #8 -->\n <tr>\n <td><span class=\"test-num\">8<\/span> Abuso de un Ben<\/td>\n <td><span class=\"cat-badge cat-mixed\">Deepfake \/ Desinformaci\u00f3n<\/span><\/td>\n <td><span class=\"img-model-sub\">Nano Banana<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">98.88%<\/span><\/td>\n <td><span class=\"pill pill-fail\">5.74%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <\/tr>\n\n <!-- #9 -->\n <tr>\n <td><span class=\"test-num\">9<\/span> Club Void<\/td>\n <td><span class=\"cat-badge cat-general\">General<\/span><\/td>\n <td><span class=\"img-model-sub\">A mitad de viaje<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-warn\">89.21%<\/span><\/td>\n <td><span class=\"pill pill-fail\">1.56%<\/span><\/td>\n <td><span class=\"pill pill-warn\">70.00%<\/span><\/td>\n <\/tr>\n\n <!-- #10 -->\n <tr>\n <td><span class=\"test-num\">10<\/span> La nevera<\/td>\n <td><span class=\"cat-badge cat-disinfo\">Desinfo<\/span><\/td>\n <td><span class=\"img-model-sub\">A mitad de viaje<\/span><\/td>\n <td><span class=\"pill pill-pass\">99.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">92.00%<\/span><\/td>\n <td><span class=\"pill pill-pass\">97.94%<\/span><\/td>\n <td><span class=\"pill pill-fail\">41.20%<\/span><\/td>\n <td><span class=\"pill pill-warn\">72.00%<\/span><\/td>\n <\/tr>\n\n <\/tbody>\n <\/table>\n <\/div>\n\n <!-- \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 FOOTNOTES \u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550\u2550 -->\n <p class=\"foot-note anim anim-d5\">\n <strong>Metodolog\u00eda:<\/strong> Cada imagen se envi\u00f3 una vez a cada detector. Se asign\u00f3 una puntuaci\u00f3n de <strong>\u2265 90% AI<\/strong> se cuenta como una detecci\u00f3n correcta. Las puntuaciones entre 70-89% son \"casi fallos\". Todo lo que est\u00e9 por debajo de 70% es un fallo. AI or Not etiquet\u00f3 una generaci\u00f3n a 78%, una a 85% y una a 89% - la de 89% se trata como casi fallo con un asterisco en la tabla resumen.\n <\/p>\n\n<\/div>\n\n<\/body>\n<\/html>\n\n\n\n<h2 class=\"wp-block-heading\">Observaciones finales y datos<\/h2>\n\n\n\n<p style=\"font-size:17px\">En total, entre la redacci\u00f3n de este art\u00edculo y las rigurosas pruebas, he dedicado 25 horas a elaborar este informe. Las herramientas de detecci\u00f3n multimodal de IA siguen desarroll\u00e1ndose, pero est\u00e1 claro que algunas son m\u00e1s precisas que otras. Tras probar todas las herramientas, TruthScan es la m\u00e1s precisa. <a href=\"https:\/\/truthscan.com\/ai-image-detector\" target=\"_blank\" rel=\"noopener\">Detector de im\u00e1genes AI<\/a>. Las pruebas hablan por s\u00ed solas. <\/p>\n\n\n\n<p style=\"font-size:17px\">Si quieres acceder a una copia en CSV de los datos de las pruebas que hice en este art\u00edculo, puedes encontrarla <a href=\"https:\/\/docs.google.com\/spreadsheets\/d\/1iuapCicNcdG6dZCFd-dCmSRgUJ9GVCS2v9kh4PShRew\/edit?usp=sharing\" data-type=\"link\" data-id=\"https:\/\/docs.google.com\/spreadsheets\/d\/1iuapCicNcdG6dZCFd-dCmSRgUJ9GVCS2v9kh4PShRew\/edit?usp=sharing\" target=\"_blank\" rel=\"noopener\">aqu\u00ed.<\/a> La hoja de c\u00e1lculo de datos que he enlazado contiene todas las indicaciones originales y los resultados de detecci\u00f3n de las pruebas de este art\u00edculo. <\/p>","protected":false},"excerpt":{"rendered":"<p>Probablemente hayas visto en Internet m\u00e1s fotos generadas por IA de las que crees. A veces es obvio que una [...]<\/p>","protected":false},"author":15,"featured_media":5326,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"rank_math_lock_modified_date":false,"_themeisle_gutenberg_block_has_review":false,"footnotes":""},"categories":[34],"tags":[19,96,97],"class_list":["post-5287","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news","tag-ai-detection","tag-ai-images","tag-detecting-deepfakes"],"_links":{"self":[{"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/posts\/5287","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/users\/15"}],"replies":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/comments?post=5287"}],"version-history":[{"count":27,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/posts\/5287\/revisions"}],"predecessor-version":[{"id":5334,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/posts\/5287\/revisions\/5334"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/media\/5326"}],"wp:attachment":[{"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/media?parent=5287"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/categories?post=5287"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/research.undetectable.ai\/es\/wp-json\/wp\/v2\/tags?post=5287"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}