PHOTO/ART indicator at the top of the analysis page expresses how photo-like vs. illustration-like an image is, as a continuous value between 0.00 and 1.00. The AI Probability and Manipulation Probability are computed by continuously blending photo-tuned weights and illustration-tuned weights using this value. There is no separate mode to switch between. However, a real selfie with overlaid stickers, or a photo that contains an embedded illustration or anime poster, can pull the indicator toward the illustration side, and any downstream probabilities will then be skewed by that classification."},"errorCases":{"title":"Cases where results are less reliable","intro":"The following kinds of images can fall outside what the detectors were designed for, or push the classification toward an extreme, so the resulting probabilities should be read with extra caution.","items":{"embeddedArt":"Photos that contain an embedded illustration, anime poster, or painting (the PHOTO/ART indicator can swing toward illustration, which can inflate the AI Probability).","stamps":"Real photos with stickers, emoji, or text overlays added (these can shift the classification and produce manipulation artefacts that raise either probability).","filters":"Images with heavy color filters, skin-smoothing, or stylization applied (the Manipulation Probability is expected to be high in this case).","screenshots":"Screenshots of phone or PC screens (these are outside the training distribution of the detectors, so both probabilities become less reliable).","reencoded":"Images that have been re-saved or re-encoded many times, especially as low-quality JPEG (the compression noise can resemble manipulation artefacts and inflate the Manipulation Probability)."}},"feedSplit":{"title":"How the home / PHOTO / ART feeds split into HUMAN and AI columns","description":"On the home, PHOTO, and ART pages, the HUMAN column and AI column are split purely by the AI Probability 50% threshold (AI Probability < 50% goes into HUMAN, ≥ 50% goes into AI). The Manipulation Probability is not used for this split. As a result, an image with a high Manipulation Probability but a low AI Probability will still appear in the HUMAN column."}},"useCases":{"title":"Use Cases","marker":"→","viralImages":"Verify viral images on social media before sharing or believing them","journalists":"A supplementary signal for journalists and fact-checkers when assessing image authenticity, alongside other verification steps","education":"Educational purposes to understand characteristics of AI-generated images","research":"Personal research and learning about image manipulation detection techniques"},"disclaimers":{"title":"Important Disclaimers","intro":"All detection results are probabilistic, not definitive proof. Our system uses advanced algorithms to estimate the likelihood of AI generation or manipulation, but it cannot provide absolute certainty.","falsePositives":"False positives can occur: Authentic images may be incorrectly flagged as AI-generated","falseNegatives":"False negatives can occur: AI-generated images may be incorrectly classified as authentic","indicators":"Results are indicators, not absolute truth: Use our analysis as one factor in your assessment, not the sole basis for conclusions","technology":"Technology evolves rapidly: As AI generation techniques improve, detection methods must continuously adapt. Detection accuracy may vary.","judgment":"Exercise your own judgment: Always consider multiple sources and use critical thinking when evaluating image authenticity","acknowledgement":"By using TagFake, you acknowledge these limitations and agree that the service provides estimated analysis, not guaranteed verification."},"faq":{"title":"Frequently Asked Questions","accuracy":{"question":"How accurate is the detection?","answer":"TagFake combines multiple detection methods to improve reliability, but no det