AI media · the story behind the screen
Big reaction.
Better questions.
The impossible celebrity photo. The familiar voice selling something unexpected. The image everyone is arguing about. Trace the story, read the sources and learn what you can check.
Inspect an imageCelebrity AI & creative work
Read the production record behind synthetic performances, surprise appearances, endorsements and fan-made media.
Zendaya and Tom Holland wedding photos: separate AI images from public statements
What Zendaya said about AI wedding photos, the later Holland interview context, and how to verify a particular image without guessing private wedding details.
Practical guideGene Wilder’s Wonka AI voice: what the consent disclosure tells you
Check the source, credits and scope of permission behind Gene Wilder’s recreated Wonka voice. An authorized replica is still different from an archive recording.
Documented caseVal Kilmer’s AI film announcement: distinguish plans, permission and footage
How to verify a posthumous AI-casting story without confusing a production announcement with a released performance or an actor’s personal endorsement.
Documented caseMichael Caine’s AI-voiced Odyssey: how to read an audiobook credit
Identify the exact audiobook edition, distinguish a licensed voice replica from a studio reading, and check what the publisher actually discloses.
Documented caseTaylor Swift's fake AI endorsement: verify the statement, not the poster
A verification guide to the 2024 false endorsement imagery: separate a celebrity's own statement from fan posts, campaign shares and generated pictures.
Documented caseTom Hanks's AI dental-ad warning: check who authorized an endorsement
What Hanks's 2023 warning establishes, and how to document an apparent celebrity ad without trusting a familiar face or a screenshot of an account.
Documented caseScarlett Johansson and Sky: resemblance is not proof of voice cloning
Read the May 2024 statements from Johansson and OpenAI separately, and learn what evidence a claim about a synthetic voice actually needs.
Documented caseKaty Perry's 2024 Met Gala AI images: verify attendance before the outfit
A practical way to investigate an apparent red-carpet photo using the event date, the person's account and independently sourced event coverage.
Documented caseThe 2021 Tom Cruise TikTok deepfakes: preserve the creator's disclosure
How creator interviews and production disclosures help verify a convincing impersonation, and why cropped reposts need their context restored.
Documented caseCadbury's Shah Rukh Khan AI ads: verify the authorized campaign scope
The 2021 personalized campaign shows why synthetic media and deceptive endorsement are separate questions. Check the issuer, participation and reuse scope.
Practical guideThat celebrity ad looks convincing. Check the endorsement, too.
A practical checklist for celebrity AI ads: verify the campaign, seller and destination before treating a familiar face as an endorsement.
Practical guideThe celebrity giveaway screenshot is missing the most important part.
Check an alleged celebrity giveaway using its original post, official terms and seller identity instead of relying on a cropped screenshot.
Practical guideBefore the red-carpet photo becomes tomorrow’s headline, check the event.
Verify a surprising celebrity event photo through the event timeline, credited photo series, official attendance and the original image file.
Practical guideAI performance or impersonation? Ask where the permission is documented.
Review a synthetic performance through production credits, authorization statements, disclosure and the exact use being described.
Practical guideThe fan poster is stunning. Is the film actually announced?
Distinguish an AI fan poster or concept trailer from an official film announcement using studio sources, credits and release records.
Viral images & world events
Trace a surprising post to its source, place and date before accepting the story attached to it.
Is Jean Phil AI? Trace the viral character, clips and coin.
What September 2026 reporting establishes about Jean Phil, the AI debate and the linked token—and how to verify an individual clip without guessing.
Documented caseThe Wiener–Chan chatbot: who is speaking when a bot uses a person’s name?
How to attribute a campaign chatbot response: identify its operator, preserve the prompt and verify claims against the represented person’s original statements.
Documented caseThe Fannie Mae AI-voice ad: check permission and repost dates separately
Trace an AI-voice advertisement back to its original disclosure. A recent repost can contain older material, and authorization does not make generated speech live audio.
Documented caseThe Pope puffer-jacket image: check the source before sharing
What first-hand fact-checking established about the 2023 viral image, with a source-tracing workflow for realistic pictures of public figures.
Documented caseHeart on My Sleeve: verify a song's release, voice and credits separately
Use the 2023 Drake and The Weeknd soundalike case to distinguish an official release, a synthetic performance and an unresolved production history.
Practical guideFind the first source of a viral photo—not the loudest repost.
Trace a viral photo through image search, earlier uses and original files. Build a source timeline without confusing first search appearance with capture time.
Practical guideThat impossible concert moment needs more than one camera.
Check a surprising concert clip using the set list, venue timeline, complete footage and independent audience angles.
Practical guideThe photo can be real. The story attached to it can be wrong.
Investigate a misleading image caption by separating the file, event, date, place and allegation into independently checkable claims.
Practical guideBefore a disaster image drives your next move, verify the place and time.
Check a dramatic disaster image through official incident information, source footage and identifiable location details before relying on its caption.
Practical guideA news logo is easy to recognize. Find the actual article.
Verify a viral news screenshot by locating the publisher’s article, headline, timestamp and correction record.
Verification field guides
Work through original files, screenshots, signed records and conflicting results with a clear evidence note.
What is a deepfake? Separate generated media, altered identity and false context
Understand deepfake meaning with practical examples. Separate synthetic people, altered performances, ordinary edits and misleading captions before choosing a check.
Practical guideAI art or human artwork? Request a useful production record
A practical provenance workflow for art commissions: agree on AI use, preserve stages, inspect an original export and record what the evidence supports.
Practical guideSora watermark checks: trace an existing video and its production record
Review a claimed Sora video using its original file, dated production context, visible marks and supported provenance evidence. No universal Sora detector claim.
Practical guideReverse video search: extract frames and trace the original clip
Search for a video’s source using locally extracted frames, manual image searches and an evidence manifest. Compare earlier copies without claiming whole-video identification.
Practical guideAI watermark detector: check visible marks, SynthID and C2PA separately
Choose the right AI watermark check. Learn how to access Google’s SynthID verification, read signed Content Credentials and interpret a missing watermark.
Practical guideHow to verify an AI influencer: identity, operator and advertising disclosures
Check who operates an AI influencer, how its synthetic identity is disclosed and what a sponsored post actually claims. A realistic face is not an identity check.
Practical guideHow to tell if a video is AI: verify the full clip, source and soundtrack
A practical AI video verification workflow: find the full recording, inspect disclosures, compare the soundtrack and context, and record what remains unknown.
Practical guideSamuel Smith’s AI demos: trace the workflow before labeling the final song
A production-history check for AI-assisted music: distinguish a generated arrangement demo from the musicians and audio used in the released master.
Practical guideA signed image still needs a source check.
Read Content Credentials alongside the claim attached to a viral image. Separate signed provenance, editing history and real-world context.
Practical guideIt sounds like them. Verify the message another way.
A practical verification workflow for a suspicious celebrity voice clip or voice message, without treating familiar sound as proof of identity.
Practical guideOne video frame cannot authenticate the whole video.
Use extracted video frames responsibly: match scenes and inspect still-image records without claiming to verify motion, speech or an entire clip.
Practical guideYour screenshot is evidence of a screen—not the original photo.
Understand what a screenshot preserves, what it cannot establish and how to request the original image for a useful verification record.
Practical guideMake an AI-image finding someone else can actually check.
Use a concise evidence-note template with the source, claim, original file, observed records, unresolved questions and a reproducible next step.
Practical guideWill your AI disclosure survive the screenshot?
Plan a clear AI disclosure for original posts, crops and shared copies, with a production note that describes the actual work.
Practical guideTwo AI detectors disagree. What should you check next?
Resolve conflicting AI-image results by comparing inputs, reading evidence and seeking the original file instead of treating votes as proof.
Disclosures, platforms & rules
Follow documented changes to labels, creative-industry policies and platform decisions. Sources and event dates stay visible.
Brazil’s TSE AI-video decision: verify the event before the headline
A practical source-checking workflow for the TSE decision about an AI convention video: distinguish the clip, its distribution and the court’s actual ruling.
Documented caseThe Manhattan deepfake-domain seizure: verify the action without spreading abuse
Read a domain-seizure announcement accurately, distinguish allegations from findings, and document harmful impersonation without recirculating intimate material.
Documented caseThe councillor deepfake decision: an AI label is not the whole moderation question
Read the Oversight Board’s September decision precisely: distinguish manipulated speech, the policy supporting removal and recommendations for future handling.
Documented caseAI harassment of a private person: lessons from the September Board decision
How to document a manipulated interview and report impersonation while avoiding further exposure of the person targeted. Keep the original account and altered copy distinct.
Documented caseARIA’s AI music rules: check the recording’s role, not just an AI label
Read ARIA’s 2026 chart eligibility change accurately. Separate AI-generated and AI-assisted production, the applicable chart week and the evidence behind a claim.
Documented caseSAG-AFTRA’s 2026 agreement: a synthetic performer is not always a digital replica
Read AI performer headlines precisely: distinguish a fictional synthetic performer from a replica of a real person, and check which production agreement applies.
Documented caseYouTube’s AI label changed in 2026. Here is what to check.
Understand YouTube’s May 2026 AI-label update, automatic labels and why a disclosure is only one part of reviewing a clip.
Make the next check count.
Start with the original image and a specific question. Inspect supported saved records, compare copies or work through our controlled examples. A useful result says what was found and what still needs another source.
Cases summarize linked public sources; field guides present isGenAI’s suggested review workflows. Event dates are separate from publication and review dates. We do not reproduce disputed celebrity likenesses. See our corrections process and checking methodology.