How do I decide when to label a social post as AI-generated?

Label a post when a reasonable viewer could mistake something made or materially changed by AI for something real: a realistic person, place, event or voice that did not exist or did not happen that way. Help with wording, spelling or layout does not usually need a label.

Platform rules and local law change, and they differ by platform and country. Read the current rules for each platform you post on, and treat this guide as a way to organise the decision, not as legal advice.

• Usually label: a realistic image or video of a person, product result or place generated by AI; a real person's face or voice altered or synthesised; a real event shown as happening differently from how it did.
• Usually no label needed: a caption drafted or tidied with AI; colour correction, cropping, background blur; clearly stylised illustration or animation nobody would take for a photograph.
• Decide case by case: a product shot with an AI-generated background; a voice-over read by a synthetic voice; a composite of real photographs.
• When in doubt, label. A label on a post that did not strictly need one costs little; a missing one on a post that did can cost the audience's trust.

1. 1Set the brand's default. If most of the brand's visuals are generated, default to labelling and make exceptions for real photography. If they are mostly real, default to no label and require one for generated or altered media.
2. 2Name who decides. One person per brand signs off exceptions to the default. Without an owner, the decision is made by whoever happens to post.
3. 3Decide per campaign, then per post. A campaign built on generated imagery can take the label throughout. Inside it, a post that uses only real photography can be the exception.
4. 4Use the platform's own label first. Where a platform offers a disclosure setting when you publish, use it; it is the label viewers and the platform recognise. Where it does not, a short line in the caption, such as "Image made with AI", does the job.
5. 5Record the decision with the post. Keep whether a post was labelled, and why, next to its approval. If anyone later asks, the answer should not depend on someone's memory.

Add two columns to the campaign plan: AI media (none, edited, generated) and label (yes or no). Fill in the first when the artwork is made and the second when the post is checked, and colour any row where generated media has no label.

Add the label decision to whatever the reviewer sees. A client approving a post should know whether it will carry a label, because it changes how the post reads. See how to run client approval.

"Mark my posts as AI-generated" is a brand default you can override for a campaign or a single post. When a post goes out, CampaignFlow sends the label wherever the platform's publishing interface has one; today that is YouTube's altered or synthetic content disclosure, and Pinterest's once Pinterest is out of platform review. Where a platform has no such field, nothing is sent, so add a caption line if your policy needs one there.

See the features page for the rest of what happens when a post is published.