How to agree on an AI disclosure policy with each client
Do you have to label AI content on Instagram, TikTok, or YouTube? What each platform asks for, and a one-page AI policy to agree on with every client.
A client’s new product reel has an AI voiceover. The photo carousel that goes with it was retouched with an AI tool, and the caption started as an AI draft. Which of the three needs a label?
The short answer: on Instagram, TikTok, and YouTube, you have to label AI-made or AI-altered content when it looks or sounds real. Each platform draws that line in its own place. AI help with ideas or scripts, and minor edits, usually don’t need a label. And all three platforms can add a label themselves, whatever your team decided.
That’s why we’d stop deciding post by post. Agree on a one-page policy with each client instead: what gets labeled, who decides, and where it’s recorded.
What each platform asks for
Meta (Instagram, Facebook, and Threads). Meta requires an AI label when you post photorealistic video or realistic-sounding audio that was made or altered with AI, and says there may be penalties for content shared without one. A reel narrated with a realistic AI voiceover is one of its examples. Its help page says the requirement doesn’t apply to images. You add the label yourself when you post.
Meta also labels content on its own when it finds industry-standard signals that it was made with AI. When those signals show content was only edited with AI, the label sits in the post’s menu instead of on the post.
TikTok. TikTok’s Community Guidelines require creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people. TikTok says unlabeled content may be removed, restricted, or labeled by its team. You turn on TikTok’s AI-generated label when you post. TikTok also labels content automatically when it was made with TikTok’s own AI effects, or when it arrives with Content Credentials attached.
YouTube. YouTube requires creators to disclose realistic content made or meaningfully altered with AI: a real person saying or doing something they didn’t, altered footage of a real event or place, a realistic scene that never happened, or AI music that’s the main focus of the video. Content that’s clearly unrealistic doesn’t need it, and neither do minor edits such as beauty or lighting filters, or production help such as AI for ideas, a script, or a title. You disclose it in YouTube Studio, under “AI use” when you upload.
In May, YouTube started labeling videos itself when its systems detect significant photorealistic AI use. Creators can change a label YouTube got wrong in most cases, but not one on content made with YouTube’s own AI tools or carrying C2PA metadata. YouTube says disclosing alone won’t limit a video’s audience or its eligibility to earn money.
Content Credentials, the C2PA standard, is a record attached to a file that shows where it came from and how it was edited. Platforms can read it, so ask the client how every asset was made before it goes into a post.
Ads and branded content
Meta’s help page says labeling works differently for ads. Meta adds an “AI info” label itself to ads made or materially edited with its own generative AI tools, usually in the ad’s menu. And Meta requires ads about social issues, elections, or politics to disclose photorealistic images, video, or realistic audio that was digitally created or altered.
TikTok’s ads policy allows AI-generated or significantly AI-edited content in ads only when it’s disclosed, and says ads with undisclosed AI content will be rejected or restricted.
For a paid partnership post, treat the AI label and the paid partnership label as two separate checks. Both belong in your pre-publish checklist.
Why a policy beats deciding post by post
Labels are the easy part. The hard part is the call behind them, made at 5 p.m. by whoever is scheduling, for one client who wants everything labeled and another who’s never asked. A person should make that call, take responsibility for it, and make it the same way every time. Being open about AI use protects the trust a client’s audience has in them.
In ContentCrew, the policy can live as a page you share with the client by link for their comments, and the client approves each post through a review link without logging in.
The one-page policy to copy
Fill this in with each client. If it runs past one page, it’s too long.
- What counts as AI-made for this client. For example: any image, video, voice, or music generated or altered with AI. AI help with ideas, scripts, or captions is recorded but not labeled.
- When to label. Always when the platform requires it, plus whatever the client adds, such as every AI-generated image.
- Who decides edge cases. One named person on your team, and one at the client when that person is unsure. If nobody can decide, label it.
- How the client signs off. The client sees whether a post will carry an AI label when they approve it, not after it’s live.
- Where each post’s AI use is recorded. Which assets used AI, for what, and whether the post was labeled, kept next to the post’s approval.
- When to review this policy. On a set date, such as every six months, and whenever a platform changes its rules. Check Meta’s, TikTok’s, and YouTube’s pages when you do.
How to agree on it with the client
Keep it to a short conversation. Walk through the six points together, send the result in writing the same day, and treat the client’s reply as their sign-off. Put the review date in both calendars.
Ask whether the client’s industry or market adds rules of its own, and have them check with whoever handles their legal questions. Laws can add rules too. In the EU, for example, the AI Act’s transparency rules have applied since August 2026, and they include labeling deepfakes: images, video, or audio generated or edited with AI. This post isn’t legal advice: what the law asks of a specific client is a question for the client’s own adviser.
FAQ
Do I have to label captions written with AI? The platforms’ labeling rules are about content that looks or sounds real. YouTube names AI help with ideas, scripts, and titles as production help that doesn’t need a disclosure. Your client may still want to know, which is what the policy is for.
Do AI-edited photos need a label on Instagram? Meta’s requirement doesn’t cover images, but Meta may still label a photo itself if it finds AI signals. For content only edited with AI, that information sits in the post’s menu.
What if a platform labels a post we didn’t label? Usually the file carried Content Credentials or other AI signals, or the platform’s systems detected AI. On YouTube, you can change a wrong label in YouTube Studio in most cases. Record what happened in the client’s policy log and tell the client before they find it.