# Platform disclosure audit

Review date: 11 October 2026 (America/New_York). Scope: YouTube, TikTok, Meta; organic content, ordinary ads, political/social-issue ads, controls, exceptions, enforcement and provenance. Audited `work/site/dist/index.html`; no website edits made by this reviewer. Line references describe the file as read during this audit and may shift during concurrent edits.

## Evidence and currentness

Primary sources below were actually read in browser DOM. Search-tool results were blank or unresolved compressed references and were not treated as evidence. An initial TikTok navigation timeout resolved into a usable page; the content was then read successfully. Current TikTok Community Guidelines explicitly show release 25 August 2026 and effective date 24 September 2026. Meta has a 1 June 2026 ad-policy update and current help pages; its organic help page displayed “Updated: 9 weeks ago,” which is recorded as displayed rather than converted into an invented exact date. Undated live help pages are checked-on sources, not proof there has been no unannounced change.

## Sources and verified findings

### Y1 — YouTube content disclosure

https://support.google.com/youtube/answer/14328491?hl=en

Live title: “Disclosing use of GenAI content.” Required triggers: making a real person appear to say/do something they did not; altering footage of a real event/place; generating a realistic nonexistent scene; AI music that is the video's main focus. The same page's expanded required-examples list broadly includes AI-generated music without that qualification. Consequently, the supporting-soundtrack question is not explicitly resolved by that page: do not declare background music categorically exempt. The upload control is Attributes → AI use. Exempt examples include own-voice cloning, captions, production assistance, minor repair/upscaling, clearly unrealistic scenes and an AI moving-car backdrop example. Automatic labels can follow YouTube tools, C2PA metadata or internal detection. Some automatic/manual-review labels cannot be changed. Disclosure alone does not restrict audience/monetization; repeated nondisclosure may cause removal or Partner Program suspension.

### T1 — TikTok CURRENT Community Guidelines

https://www.tiktok.com/community-guidelines/en/integrity-authenticity

Released 25 August 2026; effective 24 September 2026. Expanded “Edited Media and AI-Generated Content” section requires disclosure for realistic AI content OR significantly edited media. Significant editing includes meaning-changing cuts, rearranged clips and added/removed audio/video; therefore manual compositing is not universally exempt. Clear caption, watermark or sticker can satisfy this disclosure, and AIGC has its own label. Exceptions include minor color/reframing/cropping, artistic anime styles and generic TTS that is not a recognizable known person's voice. Private-figure likeness without consent and harmful false news/crises/public-figure endorsements are prohibited even when labeled. Depending on harm, unlabeled content may be removed, restricted or labeled. Unconfirmed realistic public-importance media can be ineligible for For You. Humor/art examples do not override these restrictions.

### T2 — TikTok organic help and controls

https://support.tiktok.com/en/using-tiktok/creating-videos/ai-generated-content

Current page read successfully after navigation timeout. It encourages labeling wholly generated/significantly edited AI content and requires labels for realistic images/audio/video. Video control: More options → AI-generated content. Own text/hashtag/sticker or descriptive context are also described. Turning the setting on does not affect distribution when content complies with Community Guidelines. TikTok effects may already label the AI use; adding independently generated/edited AI material requires reassessment. Automatic labels can follow TikTok effects or attached Content Credentials and cannot be removed. Misleadingly labeling unaltered content may be removed. Some help hyperlinks retain a 2024 guideline-version parameter; use T1's current guidelines to resolve current thresholds, especially the generic-TTS exception.

### T3 — TikTok ordinary advertising

https://ads.tiktok.com/resources/help/article/tiktok-ads-policy-misleading-and-false-content

Displayed update: April 2026. Fully generated image/video/audio and significant AI edits require AIGC labeling or a clear disclaimer/caption/watermark/sticker. Minor lighting, brightness, saturation, background modifications and denoising are listed as insignificant AI edits. A label does not permit otherwise prohibited misleading claims or identity misuse.

https://ads.tiktok.com/resources/help/article/add-disclaimers-to-ads

Displayed update: September 2025. Non-Spark ads with completely generated or significantly modified image/video/audio use Ads Manager's AI-generated-content disclaimer. Spark ads are organic posts and follow Community Guidelines. Availability may vary. After submission the toggle cannot simply be unselected unless creative changes; duplicated/new campaigns reset the toggle even if the previously labeled asset is reused. An advertiser should check each new campaign rather than assume an earlier disclosure carries over.

### T4 — TikTok political advertising

https://ads.tiktok.com/resources/help/article/tiktok-ads-policy-politics-government-and-elections?lang=en

Displayed update: July 2026. Paid political advertising is prohibited, including compensated creator content and Promote; political figures/entities cannot advertise. Limited certified government/intergovernmental and official election-management entities can advertise in approved categories, with targeting and content limits. Disclosure does not create permission to run an otherwise prohibited political campaign. T1 independently confirms the same basic restriction.

### M1 — Meta LATEST ordinary-ad help

https://www.facebook.com/business/help/1010479435004531

Current title: “About AI info on ads created or edited with generative AI tools.” Ads previously marked Sponsored now display Ad; name/availability can vary by region. Meta automatically labels ads using certain own tools or detectable third-party signals including C2PA. Typical location: three-dot menu → About this ad → AI info. Own-tool examples include Background Generation, Image Generation and Add Animation. Before publication, a creation-flow message indicates when an AI label will appear. Some regions offer advertiser self-disclosure: European Region, California, New York, India and Taiwan. Boosted organic posts may retain their organic labels. The page does not establish a global blanket mandatory self-disclosure rule for every AI-written ordinary ad.

### M2 — Meta dated ad-transparency update

https://about.fb.com/news/2025/02/gen-ai-transparency-metas-ads-products/

Update 1 June 2026: About this ad rollout consolidates information within every ad's three-dot menu, including existing own-tool AI labels and automatic detection of third-party industry signals; region variation is expressly noted. Older February 2025 text describes significance/photorealistic-human tests and label placement beside Sponsored, but should not be presented as the universal latest UI. The June update is relevant to current ordinary ads and supports distinguishing automatic labeling from advertiser duties.

### M3 — Meta current organic help

https://www.meta.com/help/artificial-intelligence/how-ai-generated-content-is-identified-and-labeled-on-meta/

Displayed update: 9 weeks ago. Meta can require creator disclosure for digitally created/modified photorealistic video or realistic-sounding audio, including AI; this requirement DOES NOT apply to images. Images may still be automatically labeled. Examples needing disclosure include realistic AI narration and AI vocals; cartoon landscapes are exempt. AI-created media detected through signals gets a visible label; merely AI-modified content's details appear under Options. Creator profile AI labels are another optional mechanism and should not be described as a substitute for a required individual-content disclosure. Not all AI content contains identifiable signals. Failure to use a required label may have penalties.

### M4 — Meta political/social-issue advertising

https://www.facebook.com/business/help/1486382031937045

Advertisers must disclose photorealistic images/video or realistic audio created/edited with AI that falsely depicts a real person's acts/speech, a realistic nonexistent person/event, altered event footage or an alleged event recording that is not authentic. Inconsequential edits immaterial to the ad's claim/issue—resize, crop, color correction, sharpening—are exempt. Controls: Special Ad Category → Social issues, elections or politics; Ad creative → Includes media created or edited with AI. Disclosure appears near Paid for by and in Ad Library. Automatic detection does not replace the stated self-disclosure duty. Nondisclosure can remove ads; repeats may penalize accounts. These ads cannot run in the EU.

https://about.fb.com/news/2026/02/meta-prepares-for-2026-us-midterms/

February 2026 article, updated 1 June 2026, independently confirms self-disclosure in specified political/social-issue cases, About this ad updates, current organic video/audio requirement and sanctions. This is a current alternative to old organic announcements, not just a retrospective report.

### P1 — Provenance limitations

https://c2pa.org/faqs/

Content Credentials preserve signed, tamper-evident creation/edit history. Metadata can be removed; supported soft bindings can help rediscover it. Credentials concern provenance, not a universal detector or independent guarantee of factual truth. Absence of a credential does not demonstrate that content is human-made. Platform labels can follow signals, self-disclosure, effects/tools or internal detection; do not attribute every automatic label to C2PA. No verified primary source in this audit establishes the site's old “over a billion videos this way” as a C2PA-only count; remove it rather than substitute another unsupported number.

## Exact replacement copy and reach

The website duplicates platform conclusions in its Platforms panel, comparison summaries, FAQ/scenario table, expandable full-tier table (`ROWS`), walkthrough action data (`ACTIONS`), and result renderer. Fixing the panel alone leaves contradictory verdicts. This audit does not call for adding a new selector or rebuilding the app: use accurate conditional summaries wherever the tool lacks platform/format inputs.

| Location in audited file | Exact proposed replacement |
|---|---|
| Platforms review note, line 593 | “Platform guidance checked 11 October 2026 against current YouTube help, TikTok’s guidelines effective 24 September 2026, and current Meta organic and advertising help. Publishing controls and regional availability can change; follow the linked policy for your format.” |
| YouTube card, 594; music `ROWS`, 1044–1047; music `ACTIONS`, 1260 | “YouTube explicitly requires disclosure when AI music is the video's main focus, and also lists AI-generated music as a disclosure example. The page does not clearly resolve every supporting-soundtrack case; check its current guidance rather than assuming background music is exempt.” |
| TikTok organic card, 595 | “Label realistic AI content and significant edits that make people or scenes appear real. The rule can apply to manual edits as well as AI. Minor cleanup, artistic styles and generic TTS narration that is not a recognizable person's voice are named exceptions. Use More options → AI-generated content or a clear disclosure.” |
| Meta card, 596 | “Organic posts: creator disclosure may be required for digitally created/altered photorealistic video and realistic-sounding audio, including AI narration or vocals. Images do not have the same creator-disclosure requirement, but may get automatic labels. Ads: Meta labels certain own-tool and detectable third-party AI uses under About this ad. Regional self-disclosure controls and political/social-issue ad requirements differ.” |
| Manual Photoshop FAQ, 728; summary 586 | “Manual editing can still trigger platform disclosure. TikTok covers realistic, significantly edited media; Meta’s organic requirement covers digitally created or altered realistic video/audio, including non-AI edits. YouTube’s cited AI-use rules concern AI content. Deception, consent and other rules also apply independently.” |
| Summaries 574 and 704 | “A realistic synthetic person can trigger disclosure, depending on platform, organic or paid format, location and context. Labeling does not permit prohibited impersonation or misleading endorsements.” |
| Summary 580 | “Platforms may add labels from technical signals, their own tools or internal detection. Complete the platform's required disclosure process even when automatic labels are possible.” |
| Summary 581 | “YouTube says disclosure alone does not limit audience or monetization. TikTok says its setting does not affect distribution when the content complies with its guidelines. Do not generalize either promise to every platform or prohibited content.” |
| Metadata FAQ, 751 | “Platforms may use embedded provenance alongside other methods. Missing metadata does not prove human authorship, and metadata does not replace a required publishing disclosure.” |
| Sources 837–842 | Link T1 and T2 for current TikTok organic guidance, T3 for paid ads; link M1/M3/M4 for Meta ads/organic/political cases. Retain dated newsroom links as supporting context rather than the only current policy sources. |
| Scenario table, 709 onwards | Use “Depends on platform, format and realism; see Platforms” instead of blanket Required/Not required for synthetic people, background swaps, retouching and unrealistic artwork. Paid-ad thresholds and organic-image treatment differ. |

### Data-verdict corrections (apply in BOTH ROWS and ACTIONS)

- Replace every unqualified platform “Required” for realistic synthetic models/spokespeople, product alteration, removed objects, staged events, invented places, historical-photo restoration or influencers with “Depends on platform and publishing format.” Keep a second sentence naming the applicable concrete example or trigger. The current UI does not know which platform/format was selected and cannot honestly determine a universal obligation.
- Keep YouTube-specific cleanup and own-voice exemptions explicitly scoped to YouTube. Suggested platform field: “YouTube names an exception; check other platforms.” Own-voice cloning is not automatically exempt on Meta when the resulting audio is realistic.
- Production assistance is distinct from final published AI-written text. Suggested wording: “Behind-the-scenes script/title/caption assistance does not itself trigger YouTube’s AI-use disclosure. Published content still needs its applicable advertising, authenticity and legal checks.”
- Fantasy/art exemption: “YouTube and TikTok organic rules exempt clearly unrealistic/artistic content; TikTok paid ads have a separate fully-generated-media disclosure test.” Do not label all paid fantasy artwork universally exempt.
- Product-demo claims that a policy “names it directly” need an exact supporting source or replacement: “Do not misrepresent what the product can do. AI disclosure depends on the platform’s generated/significantly edited media rules.”
- Historical restoration is not automatically significant manipulation. Suggested wording: “Minor repair/color correction may be exempt; realistic changes to depicted facts can require disclosure.”
- Synthetic influencer guidance: “Paid-partnership/endorsement disclosures and AI disclosures are separate. Whether both apply depends on the content, platform and commercial relationship.”
- “Restricted or prohibited on every major platform” and “commonly removed on sight” are unsupported absolutes. Replace with “Check the platform’s impersonation, consent, misinformation and manipulated-media prohibitions; a label does not override them.”
- Fake-review prohibition can remain, but avoid claiming every conceivable platform has an identical rule. TikTok current deceptive-behavior section expressly includes fake reviews; FTC obligations are independently legal.

## Residual limits

### Requested per-media implementation guidance

| Media/action | Suggested platform explanation |
|---|---|
| Own-voice clone | “YouTube explicitly exempts cloning your own voice for narration/dubbing. TikTok’s generic-TTS exception is limited to narration that is not a recognizable known person's voice; a realistic clone of an actual person needs separate assessment. Meta requires labels for realistic digitally generated/altered audio and expressly names AI narration/vocals.” A TikTok automatic effect label can satisfy the effect's labeling, but independently modified material needs reassessment. |
| Image background swap | “A minor TikTok ad background edit can be exempt. TikTok organic realistic misleading additions/removals can need labels. Meta does not require creator AI disclosure for images, but may add automatic labels. YouTube’s named backdrop exemption concerns a moving-car simulation.” |
| Video background swap | “Evaluate whether a realistic scene or event has been changed, not merely whether a background tool was used. Minor edits and YouTube’s moving-car example differ from realistic misleading scene changes. Meta’s realistic digitally created/altered video rule remains relevant.” |
| Colorize/restore | “Minor repair or color correction may be exempt; adding invented realistic details or changing depicted facts can trigger the platform’s edited/generated-media rule.” |
| AI music | “YouTube's main-focus trigger and broader AI-music example leave supporting-music cases unclear. Meta expressly names AI vocals and realistic generated audio. TikTok's generic-TTS narration exception does not itself settle music or cloned singing.” |
| Clearly unreal generated art | “YouTube and TikTok organic artistic/unrealistic media can be exempt; TikTok paid ads separately cover wholly generated media. Meta's cartoon-video exception and lack of image creator-disclosure duty differ from automatic labeling.” |
| Logo | “An abstract logo normally does not depict realistic synthetic people/scenes; assess its use and format rather than declaring every logo outside disclosure rules. A wholly generated logo used as ad creative may fall within TikTok's fully-generated-image ad test. This last point is an application inference from T3, not a separately named logo example. Copyright/trademark questions are independent.” |

## Review limitations

This review covers the named platforms and their published sources, not account-specific experimental interfaces, every jurisdiction, YouTube/Google paid-ad policy or every third-party creation tool. It does not turn uncertain source wording into certainty: YouTube supporting-music cases remain expressly unresolved. Sources were read without logging into an advertiser account, so control labels are verified from official documentation rather than the user's own account UI.