Labeling AI Edits When You Post — What Platforms Actually Require
Every prompt page on this site ends with the same reminder: label the result as AI-edited when you share it. That is our house advice. It is not the same thing as what each platform requires, and the rules are narrower — and more specific — than most people assume.
Here is what four big platforms actually say, as of October 2026, and how it applies to the kind of edits on this site. Policies change, so each section links to the platform’s own page. Check it before relying on this.
The One Idea All Four Share: Realism
None of these platforms bans AI edits, and none requires a label just because an AI tool touched the image. What they care about is whether a viewer could mistake the result for something that really happened.
That gives a simple test you can apply to any edit before posting: could someone who does not know you believe this scene is real? A clay figurine on a desk, a Polaroid frame, a painted storybook style — no one thinks those are photographs of a real moment. You standing in an English country garden you have never visited is a different matter.
Instagram and Facebook (Meta)
Meta’s own announcement says it will “require people to use this disclosure and label tool when they post organic content with a photorealistic video or realistic-sounding audio that was digitally created or altered”, and that it “may apply penalties if they fail to do so.” (Meta, February 2024)
Note what that covers: video and audio. For still images, Meta’s approach is to label them itself when it detects industry-standard AI signals in the file, or when you choose to disclose. The label shown is “AI info”, which replaced “Made with AI” in July 2024. (Meta, updated 2024)
What this means for you: an animated version of your photo is the case where Instagram expects you to switch the AI label on yourself. A still edit may get labelled automatically anyway.
TikTok
TikTok’s help page says it requires creators to disclose realistic AI-generated content, and encourages — but does not require — disclosure for content that is not realistic. It may also apply an “AI-generated” label automatically when it detects that content was made or significantly edited with AI. (TikTok support)
The same realism test applies. A photo-to-video clip of a real person doing something they did not do — turning, laughing, walking down a street — is realistic. Switch the AI-generated setting on when you post it.
YouTube
YouTube’s Help Center is the most specific of the four. Creators must disclose when they “use AI to meaningfully alter or generate photorealistic content”, including content that:
- makes a real person appear to say or do something they didn’t do
- alters footage of a real event or place
- generates a realistic scene that didn’t actually occur
It also lists what does not need disclosure, and several of them describe edits from this site: “applying beauty filters”, “color adjustment or lighting filters”, and “special effects filters, like adding background blur or vintage effects”. Repeatedly skipping a required disclosure can lead to a label being added for you, or penalties up to removal of content or suspension from the Partner Program. (YouTube Help)
So a vintage film grade on your own photo is exempt. A Short where your still photo turns and smiles — a real person appearing to do something they did not do — is not.
Pinterest shows an “AI modified” label in the close-up view of an image Pin when its systems detect that the image was generated or modified with AI. It checks the file’s metadata against the IPTC standard and also runs classifiers that can flag images without obvious markers. A Pin can also get the label when its owner says it was made or modified with AI. If you think a label is wrong, you can appeal through Pinterest support. (Pinterest Help)
Pinterest is the one place where the label is mostly applied to you rather than by you.
Your Edits May Already Carry the Label
The automatic labels above work because AI tools increasingly mark their output. Meta names the C2PA and IPTC standards as the signals it reads, and Pinterest names IPTC. So whether or not you add a label, the platform may add one from the file itself.
That is a good reason to label first. A post you labelled reads as honest; a post the platform labelled after you did not can read as though you were hiding something.
What We Suggest
Our rule of thumb, stricter than most platforms require:
- Video of a real person: always label. Meta, TikTok and YouTube all put realistic video of a real person inside their disclosure rules.
- Photorealistic stills that change the place, clothes or situation: label. The Old Money edit above is the example.
- Obvious styles — figurines, illustrations, frames: label anyway if you like. No one is misled, and no platform above requires it.
- Restoration and colourisation of old family photos: say so in the caption. No label is required for a colour adjustment, but a restored photo can be mistaken for the original, and family members deserve to know which is which.
A short caption line does the job everywhere:
Edited with AI from my own photo.
And the permission rule matters more than the label. Only edit and post photos of people who have agreed to it. YouTube’s rule about making a real person appear to do something they didn’t is written with exactly that in mind.
Every prompt in the prompt library and every prompt the prompt generator builds is meant for your own photos. Labelling the result costs one line and keeps it that way.
FAQ
Will labelling my post reduce its reach?
The penalties YouTube’s page describes are for failing to disclose when it is required, not for disclosing. We have not measured reach on labelled posts ourselves, so we will not claim a number either way.
The platform labelled my edit and I don’t think it should have. What can I do?
Pinterest has an appeal route through its support team. Elsewhere, check whether the label is the platform’s automatic one or one you set at upload, and look at the platform’s help page linked above for the current options.
Does cropping or a filter count as an AI edit?
Not by itself. YouTube explicitly lists colour adjustment, lighting filters and vintage effects as not needing disclosure. The label question starts when an AI tool generates new content — a new background, new clothing, motion that never happened.
