How to Label AI-Generated Images and Video on Social Media

Use each platform's own AI label when you post, plus a plain caption line for anything realistic. Meta requires its label on photorealistic video and realistic-sounding audio, YouTube asks for its "AI use" disclosure on realistic synthetic content, and TikTok requires its "AI-generated content" setting for realistic AI images, audio and video.

Two habits make the rest easy. Keep the metadata your AI tool embeds in files, because platforms read it and may label for you. And if your audience includes people in the EU, remember that the AI Act's transparency rules have applied since 2 August 2026: deepfake-style images, audio and video must be disclosed as artificially generated or manipulated.

Follow me on Instagram@sagnikteaches

Platform rules side by side

Each platform's help pages set out what it requires, and they differ more than you'd expect. The summary as of September 2026:

Connect on LinkedInSagnik Bhattacharya

All four also apply labels themselves when they detect industry-standard signals in the file, so the metadata question in the table's last row matters on every platform.

Subscribe on YouTube@codingliquids

Meta: what Facebook and Instagram require

Meta's help pages say it requires you to label content "that has photorealistic video or realistic-sounding audio that has been digitally generated or altered, including with AI", and that there may be penalties if you don't. Images are not on the mandatory list, but you can label them, and Meta may label them for you.

• How: turn on "Add AI label" in the post composer, or in the menu when creating a reel or story.
• What people see: an AI label on the post. Meta's current Facebook help page calls it an "AI content" label; its older Transparency Centre pages call it "AI Info". Same feature, renamed.
• Automatic labels: if Meta detects AI signals in the file, it may apply the label even if you didn't. Content whose signals show it was AI-modified rather than generated doesn't get the label on the post itself; the information sits in the post's options menu instead.
• Profiles: a creator who regularly shares AI content can add an "AI creator" label to the profile. Most businesses won't need it.
• Ads work differently. Meta labels ads fully created or materially edited with its own generative AI creative features, either in the ad's three-dot menu or next to "Sponsored" when photorealistic people are generated. Ads made with third-party tools don't get that automatic label, and political and election ads must self-disclose digitally created or altered content whatever tool made it.

A caption line is worth adding even when you use the toggle, because the label can be small and the caption is what people read.

YouTube: answering the "AI use" question

YouTube asks the question at upload: in YouTube Studio, under Attributes, the "AI use" setting takes a yes or no. You must answer yes when your video makes a real person appear to say or do something they didn't, alters footage of a real event or place, or generates a realistic scene that didn't happen. AI-generated music and synthetic depictions of real places are on YouTube's list of examples.

YouTube is explicit about what doesn't need disclosure, and the list is useful for businesses:

• clearly unrealistic content (animation, fantasy, obvious effects);
• beauty filters, colour adjustment and lighting filters;
• using generative AI for a script, thumbnail, title or infographic;
• automatically generated captions;
• cloning your own voice.

Photorealistic AI content is labelled on the video player itself; other AI content gets a label in the expanded description. YouTube may add labels on its own where content was made with its tools, carries C2PA metadata or is detected by its systems. Creators who consistently fail to disclose risk labels being applied for them, removal of content, or suspension from the YouTube Partner Program.

TikTok: the AI-generated content setting

TikTok requires creators to label all AI-generated content containing realistic images, audio and video, and encourages labelling anything completely generated or significantly edited with AI. To label in the app, tap More options before posting and turn on "AI-generated content". On a computer, upload the video, click See more, and turn on the same setting.

• What viewers see: Creator labeled as AI-generated. TikTok may also apply its own "AI-generated" label when content was made with its AI effects or uploaded with Content Credentials attached.
• It's permanent: you can't remove the label after posting, so decide before you tap Post.
• It doesn't hurt reach: TikTok says the label won't affect distribution.
• It doesn't make everything allowed: even labelled AI content can break TikTok's rules, for example by showing fake authoritative sources, fake crisis events, or public figures endorsing something. And labelling unaltered content as AI is itself against TikTok's terms.

LinkedIn, product listings and the metadata you shouldn't strip

LinkedIn doesn't ask you to toggle anything. Instead, images and videos cryptographically signed with C2PA Content Credentials show an icon; clicking it shows whether AI was used to generate or edit the content, the app or device, the issuer and a timestamp. Many AI image tools add these credentials automatically.

That metadata matters beyond LinkedIn. Meta, YouTube and TikTok all read industry-standard signals to label automatically, and Google's Merchant Center requires AI-generated product images to keep their IPTC "TrainedAlgorithmicMedia" metadata. The quickest way to lose it is to screenshot an image, run it through a compression tool, or re-export it from a design app that drops metadata. If your workflow does any of those, test one image: upload it somewhere that reads credentials and see whether the information survived.

Don't assume every tool marks its output yet. The EU AI Act requires providers of generative AI systems to mark outputs in a machine-readable format, but systems already on the market before 2 August 2026 have a grace period until 2 December 2026, so some tools you use may not embed marks until then.

What the EU AI Act adds for EU audiences

If your posts reach customers in the EU, Article 50(4) of the AI Act applies to you as the business using the tool (a "deployer"): if an AI system generates or manipulates image, audio or video content constituting a deepfake, you must disclose that the content has been artificially generated or manipulated. The Act defines a deepfake as AI-generated or manipulated content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.

Two practical readings for a small business:

• A realistic AI image of your own premises, a real street, or a real person (including you or your staff) sits squarely in scope. Label it.
• For evidently artistic, creative, satirical or fictional work, the duty is lighter: disclose that generated content exists in a way that doesn't spoil the work, such as a line at the end of a video.

Platform labels and a caption line together are a sensible way to meet the duty. Whether a particular image counts as a deepfake can be a fine judgement; if a campaign depends on realistic AI people or places, ask your adviser. The companion duty for chatbots is in AI chatbot disclosure at the start of a chat, and the wider question of whether AI images are legal to use at all is in can you legally use AI-generated images in your marketing.

A mortgage adviser's month of posts, labelled

An illustrative three-adviser mortgage brokerage plans twelve posts for the month across Instagram, Facebook, LinkedIn and YouTube, five of them using AI. Here is how each gets handled:

The first row shows a sensible habit rather than a rule: labelling realistic AI images even where a platform doesn't insist. A photorealistic couple with keys implies real clients, and for a firm whose business rests on trust, a small label costs nothing next to a comment asking whether these are real customers. The last row shows the strongest case for a label: a real, identifiable person apparently speaking. Getting the adviser's written agreement to the avatar and to each script is part of the job too.

In practice this added about a minute per post for the person scheduling: one question per post ("is anything in this realistic and not real?"), one toggle, one caption line.

Caption wording that works

Platform labels vary in wording and prominence, so a caption line gives you one consistent disclosure everywhere. Keep it short, specific about which part is AI, and honest about who stands behind it:

• "Image created with AI." (Fully generated image.)
• "Background generated with AI; the product is real." (Real product, AI scene.)
• "Voiceover generated with AI." (Synthetic narration.)
• "This video uses an AI avatar of [first name], who wrote and approved the script." (Digital likeness.)
• "Illustration made with AI tools." (Where you choose to disclose non-realistic work.)

Prefer the caption line to a hashtag. Instagram caps posts at five hashtags, so an #AIgenerated tag uses one of your five and says less than a sentence does. Avoid vague wording like "enhanced with technology", which reads as though you are hiding something.

Mistakes that turn a label into a problem

A realistic example first. An illustrative recruitment agency posted a "day in the life at our client's office" video made entirely with AI: synthetic employees, a synthetic office, an upbeat narrator. It went out unlabelled on Instagram and TikTok. Within a day the comments asked whether these were real staff and whether the office existed, and a candidate who had interviewed there pointed out it looked nothing like the real one. The agency deleted and reposted with labels and a caption, but the screenshots and comments had already done the damage. The lesson isn't only "label it": a realistic AI video of a real client's workplace misrepresents the job, and no label fixes that.

Other mistakes that come up:

• Stripping metadata by screenshotting or re-exporting, so the platform can't read provenance and you lose your own record of what was generated.
• Labelling on one platform but not another when the same video is cross-posted. Decide the label once per asset, not per upload.
• Labelling real content as AI to look innovative, which TikTok treats as a breach of its terms.
• Using a real client's or employee's likeness in AI content without written consent, labelled or not.
• AI "customer" testimonials. A label doesn't make a fabricated testimonial acceptable; it's a different and bigger problem.

Keeping a simple log of AI assets

Once a business posts AI content regularly, a one-line record per asset solves two problems at once. It keeps labels consistent when the same file is cross-posted weeks apart by different people, and it gives you an answer if a customer, platform or regulator ever asks how a particular image was made. Any spreadsheet everyone who posts can open will do. Rows from the mortgage brokerage's log (illustrative):

Keep the original files, with their metadata, in the same folder as the log. If a platform applies a label you didn't expect, or removes a post for missing one, the log and the original file are what you use to check what actually happened.

A pre-post labelling check for whoever schedules posts

Put these five questions next to your scheduling tool, and ask them for every asset, not every upload:

1. Was any part of this image, video or audio generated or altered with AI?
2. If so, does it look or sound realistic, or show a real person, place or event?
3. If realistic: is the platform label switched on for every platform it goes to, and is the caption line added?
4. Is the original file's metadata intact (not a screenshot or a re-export that strips it)?
5. If it shows a real person's likeness, is their written consent on file?

This check fits naturally into a broader pre-publish routine; the social media pre-publish checklist for AI-drafted posts covers the words, and the image error check before posting covers the visual mistakes that are a separate reason to hold a post back.

Labelling AI content: other questions

Does labelling a post as AI reduce its reach?

TikTok states that adding its AI-generated label won't affect your video's distribution. Meta's guidance treats labelling as a transparency tool rather than a penalty. What does carry penalties on these platforms is failing to label content that needed it. If engagement drops after you start labelling, look first at whether the AI content itself is weaker than your real photos.

Do I need to label images I only edited with AI?

Usually not for small fixes. YouTube lists colour and lighting adjustments and beauty filters as edits that need no disclosure, and Meta doesn't show its visible label on content its systems identify as AI-modified rather than generated, putting the details in the post menu instead. Label when the edit changes what the picture shows: a person added, a place altered, an event that didn't happen.

Should staff label AI content on their personal accounts when posting about work?

Yes, the platform rules apply to whoever posts, and a staff member sharing an AI avatar video of your firm without a label puts the firm's reputation at risk as well as their account. Include personal accounts in your social media policy: same labelling rules, same caption wording, and a named person to ask when unsure.

Further reads

• How to Write an AI Disclosure Statement for Your Website — The site-wide AI statement that sits behind your labels.
• Should You Tell Customers When You Use AI in Your Marketing? — Whether and how to disclose AI use in marketing more broadly.
• How to Keep AI-Made Videos On-Brand: Captions, Voice and Intros — Keep AI-made videos on-brand once they're labelled.
• How to Make TikToks and Reels for a Small Business With AI — Making TikToks and Reels with AI in the first place.
• Brand Guidelines for AI Images: Colours, Style and Limits — Set the colours, style and limits your AI images follow.
• How to Spot Deepfake Voice and Video Scams Aimed at Your Business — The misuse side: deepfakes aimed at your business.
• Should You Tell Customers You Use AI? A Disclosure Guide — When to tell customers you use AI and when you needn't, with a decision table for 11 everyday uses and wording you can copy for chats, quotes and emails.
• How Photographers Cut Editing Time With AI and Keep Their Style — Train an AI editing profile on your own finished work, test it on a job you've delivered, and review by scene so the look stays yours.
• How Photographers Can Fill Mini-Session Slots With AI Marketing — A three-week, waitlist-first campaign for mini sessions: slot maths, past-client segments, copy prompts and honest ways to fill the last slots.
• How to Plan a Month of Barbershop Instagram Content With AI — A one-hour planning session, a sample four-week grid, a shot list for filming the lot in one afternoon, and captions that sound like your shop.
• How to Create Outfit and Styling Posts for a Boutique With AI — Turn a boutique's stock list into weekly outfit carousels: AI plans combinations and writes the copy, while the photos stay real and the sizes stay honest.
• How Tattoo Studios Handle Enquiries and Deposits With AI — How a tattoo studio can use AI to sort and answer enquiries, collect what artists need to quote, and take deposits safely through booking software.
• Can AI Mock Up a Flower Arrangement Before You Make It? — What AI mock-ups of arrangements get right and wrong, three ways to make one, a prompt that respects real flowers, and wording that sets expectations.
• AI Model Shots for Boutiques: Try-On Images Without a Photoshoot — How AI on-model images work, what tools like FASHN and Photoroom cost, which garments come out wrong, and how to check and label every image honestly.
• Does Writing Etsy Listings With AI Hurt Your Search Ranking? — How Etsy search actually ranks listings, the specific ways AI-drafted copy loses ground, and a simple test to run on your own shop.
• AI Mistakes That Put Couples Off Your Wedding Business — Eight ways AI makes a wedding supplier look careless or fake to couples, each with an example from real-looking enquiries and the fix.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Facebook Help Centre (Label AI content on Facebook); Meta Transparency Centre and Meta help on AI in ads; YouTube Help (disclosing altered or synthetic content); TikTok Support (AI-generated content); LinkedIn Help (Content Credentials); EU AI Act Articles 3 and 50; Gibson Dunn briefing on the Digital Omnibus. Checked September 2026. Practical guidance, not legal advice.