AI Content Labels: What LinkedIn, YouTube and Meta Make You Disclose

LinkedIn, YouTube and Meta all label AI generated content, and no 2 of them use the same trigger. YouTube asks creators to flip a toggle. Meta labels most ads automatically and has required self-disclosure on political ads since 11 January 2024. LinkedIn mostly reads C2PA metadata and relies on a one-sentence rule. If your team publishes on all 3, your AI content labels process is really 3 processes.

The short answer. Before you publish, ask one question: would a viewer believe this shows a real person, place or event as it actually happened? If yes, disclose on YouTube with the Studio toggle, on Meta with the disclosure tool for organic video and political ads, and on LinkedIn in the post text. If no, none of the 3 platforms requires you to say anything.

This post is written for the operations lead or head of delivery who owns the publishing checklist, because the failure here is a process gap, not a knowledge gap. The source for every rule below is the platform's own help centre or newsroom, read in October 2026.

The test behind every AI content label: would a viewer think it is real

Strip away the interface differences and the 3 platforms are policing the same thing. Each one cares about synthetic or altered media that a reasonable viewer could mistake for a recording of reality. None of them cares whether AI touched your workflow.

YouTube's help centre says it requires creators to disclose AI-generated or altered content that seems realistic. Its examples are a real person appearing to say or do something they did not, altered footage of a real event or place, and a realistic scene that never happened. Meta's political ad rule lists almost the same cases, adding a realistic-looking person who does not exist.

LinkedIn's version is the narrowest. Its Professional Community Policies say members must not share synthetic or manipulated media that depicts a person saying something they did not say or doing something they did not do without clearly disclosing it. The rule is about people, not scenes.

The practical consequence is a single sorting question for your team. Most B2B creative falls on the safe side of it. A product screenshot, an AI-drafted carousel, an illustrated explainer and an AI-cleaned podcast all pass without a label anywhere.

The cases that fail it are the ones B2B marketers are now actually producing. They include AI avatars of executives, cloned voices reading customer quotes, generated b-roll of offices or cities, and synthetic "customers" in testimonial-style ads. Those are where the 3 rulebooks start to diverge.

AI content labels compared: one table for LinkedIn, YouTube and Meta

This is the table the rest of the post explains. Read each row as "what happens if you publish realistic synthetic media here".

Every cell paraphrases the platform's own page, listed in the references. "Not found" means the pages reviewed for this post did not contain a rule, not that one cannot exist elsewhere in the platform's policy library.

YouTube's altered content disclosure in practice

YouTube introduced the Creator Studio disclosure on 18 March 2024. It is a yes or no question in the upload flow, and the burden sits with whoever uploads the video, which for most B2B teams means the brand channel owner.

What needs the toggle, and what does not

The help centre example that should worry a marketing team is not a deepfake. It is AI generated extra footage of a real place, like a video of a surfer in Maui for a promotional travel video. Generated b-roll of a real city behind your event promo is the same case, and it needs disclosure.

The exemptions are generous and specific. YouTube does not require disclosure for beauty filters, colour and lighting adjustments, background blur, or using generative AI to create or improve a video outline, script, thumbnail, title or infographic. Captions, idea generation and cloning your own voice for your own voice-over are also exempt.

My reading is that the thumbnail exemption matters more than it looks. Your team can use an AI-generated thumbnail on a webinar recording with no label, while 4 seconds of generated b-roll inside the same video triggers one. Build the checklist around the video body, not the packaging.

What YouTube labels without asking you

YouTube may automatically apply an AI label to content made with its own generative tools, content carrying C2PA metadata, and content its internal systems detect. Labels applied after manual review, or triggered by C2PA or YouTube's tools, cannot be removed by the creator.

The label normally sits in the expanded description. The help centre says a label in the video player may also appear for photorealistic content and AI music. The 2024 launch post promised a more prominent on-video label for health, news, elections and finance topics.

YouTube also states that disclosure alone will not limit a video's audience or its eligibility to earn money. That sentence is the best argument to put in front of a nervous brand manager. The penalty attaches to hiding it, not to disclosing it.

The "Captured with a camera" label works in the other direction

YouTube now shows a "Captured with a camera" note for footage filmed on a device with built-in C2PA support. That support must be version 2.1 or higher, and the footage must carry no edits to sound or visuals. It is a proof-of-authenticity badge, separate from the AI disclosure.

For most B2B footage it will never appear, because nearly every corporate video is edited. YouTube says so itself: a missing badge does not mean the content was altered. I would not plan around it unless you publish raw testimony, such as unedited customer interviews where authenticity is the point.

Meta's AI info label: detection for organic posts, automation for ads

Meta runs the most automated system of the 3, and it has also revised it the most. Your team needs the current state, not the 2024 headlines, because the label's name and position have both changed.

Organic posts on Facebook, Instagram and Threads

In February 2024 Meta said it would detect AI images using the C2PA and IPTC metadata standards. It also said it would require people to use the disclosure and label tool when posting organic content with photorealistic video or realistic-sounding audio that was digitally created or altered. It added that it may apply penalties if they do not.

The label began as "Made with AI" in May 2024. Retouched photos were caught in it, so on 1 July 2024 Meta renamed it "AI info". On 12 September 2024 it moved the label into the post menu for content only edited with AI, while keeping it visible for content generated by AI.

Meta also stopped removing content solely under its manipulated video policy in July 2024, preferring labels and context. For a marketer that is mildly good news. A missed disclosure on an organic post is more likely to earn a label than a takedown.

Commercial ads

Meta labels commercial ads automatically and asks nothing of the advertiser. Ads created or significantly edited with Meta's generative creative tools get an "AI info" label. If the tools generate a photorealistic human, the label appears next to the Sponsored label, not behind the menu.

If the edits are not significant and include no photorealistic human, there is no label at all. The 1 June 2026 update extended the system to ads made with third-party tools, detected through industry-standard signals. Meta told Social Media Today in July 2026 that those signals include C2PA, and named Photoshop and DALL-E as examples.

The update also says the experience may vary by region because of legal requirements. Treat that as a hint that EU placements may carry stricter labelling than your US campaigns, and check the preview per market.

Social issue, electoral and political ads

Here Meta reverses the burden. Since 11 January 2024 advertisers must disclose when they use AI to create or edit photorealistic images, video or realistic audio in these ads. A February 2026 update says the self-disclosure label will change from "digitally created" to "AI Info", and that failure to disclose may result in ad removal and account penalties for repeat violations.

B2B teams rarely run political ads deliberately. They do run issue-adjacent campaigns on topics such as immigration policy, energy or labour law, which Meta can classify as social issue ads. If your category touches regulation, assume the stricter rule applies.

The LinkedIn AI label is provenance, not policy

LinkedIn began showing C2PA labels on in-feed images and video in May 2024, according to MediaPost's report of 21 May 2024. Click the icon and you see the AI-use assertion, the app or device used, who created it, who issued the credential and when.

The label appears only when the file carries signed Content Credentials. LinkedIn's help article is candid that it is not yet possible to identify and label all AI-generated and modified content. In select countries members can self-attest that media was made with AI, and LinkedIn generates the credential without changing the original file.

So a LinkedIn label tells your audience which tool you used, not whether you behaved well. An image exported from a C2PA-signing tool gets labelled. The same image screenshotted and reposted does not. That asymmetry is the reason I think the written rule matters more than the icon.

The written rule is the synthetic media sentence quoted earlier. LinkedIn's help article on false or misleading content says it removes photorealistic or audio-realistic content confirmed to depict someone saying or doing something they did not, unless the disclosure is clear. Obvious parody or satire may not need an explicit label.

For ads, LinkedIn's Advertising Policies, last revised 18 November 2025, contain no clause about AI-generated content. They do require ads to comply with the Professional Community Policies, which brings the synthetic media rule into Campaign Manager. They also prohibit political ads entirely, and in the EU they exclude political advertising as defined by Regulation (EU) 2024/900.

The C2PA icon does reach ads, with limits. It appears on single image and single video Sponsored Content when the file is uploaded, whether through Campaign Manager or sponsored from a Company Page. Text ads and other formats are not covered by that sentence.

AI content labels on ads versus organic posts: where the rules split

The single most common mistake I see in briefs is treating "the AI rule" as one rule per platform. On 2 of the 3 platforms, the obligation changes the moment you put money behind a post.

Rows marked "likely" or "by default" are this post's interpretation, not quoted platform text. The synthetic spokesperson row is where the published rules are thinnest.

The pattern across the table is simple. On paid placements, Meta moves the labelling job from you to its detection systems. On YouTube, the paid and organic rule is the same upload toggle in the pages reviewed. On LinkedIn, the written rule follows the content into ads, but the platform adds no AI-specific ad clause.

My recommendation is to stop trying to predict which assets a platform will auto-label. Your team cannot control Meta's detection or another tool's metadata. You can control one decision: whether a realistic depiction gets disclosed in your own words before the platform does it in theirs.

That also covers the legal floor. For EU audiences, the AI Act's deployer duty to disclose deepfakes applies regardless of what any platform labels, which is covered in the EU AI Act transparency checklist. A platform label is not a substitute for that disclosure.

Where this reference is weakest

This post claims to save you reading 3 policy libraries. It can only do that honestly if it says where the summary is likely to be wrong.

Help centres change without dates

YouTube's two help pages used here carry no update date. LinkedIn's showed "last updated 2 weeks ago" on the day it was read. Meta's newsroom posts carry visible update stamps, which is better, but its political ad page was rewritten in February 2026 without a changelog. Any row in the tables above can be overtaken quietly.

The internal launch post and the current help page also disagree in places. YouTube's 2024 announcement tied the prominent player label to health, news, elections and finance. The current help page ties it to photorealistic content and AI music. I have reported both rather than pick one.

Detection is partial, and the gaps are not symmetric

Every automatic label in this post depends on metadata surviving the trip. Exports, screenshots and some editing pipelines strip it. That means 2 identical images can be labelled differently, and the team that disclosed honestly can look worse than the one whose file lost its credentials.

The strongest argument against this post's advice is exactly that: if labels are inconsistent, voluntary disclosure is a unilateral cost. I think that argument loses over time, because platforms penalise non-disclosure and not disclosure. But for a single campaign measured on click-through rate, it may well win. Whether a label hurts performance is covered in the evidence on AI disclosure and customer trust, which is more mixed than either side admits.

One more gap is scope. This post covers LinkedIn, YouTube and Meta only. TikTok, X and Google Ads beyond election advertising each run their own regimes, and none of them were checked for this piece.

Frequently asked questions

How to run the pre-publish check this week

Add one line to your publishing brief, above the platform list. The line is a question: does this asset show a real person, place or event in a way a viewer would believe is genuine, and was any of it generated or altered? If the answer is yes, the asset gets a disclosure in the caption on every platform, plus the YouTube Studio toggle where relevant. This is the same discipline that separates credible volume from the scaled content patterns platforms penalise.

Then pick your 3 most recent paid assets and preview them on Meta in each market you run. If an "AI info" label appears that your team did not expect, trace which tool wrote the metadata. That one exercise will tell you more about your real exposure than this table can, and it will take under an hour. If your brand voice on LinkedIn already leans on generated creative, the analysis of what still works on LinkedIn is the next read.

Related reading