You made a teaser graphic in Canva, wrote the caption yourself, and posted it. Then Instagram stamped an "AI info" label on it. You never opened a text generator. The words are yours, the story is yours, and now there is a tag sitting under your post telling readers a machine had a hand in it. For an author whose entire pitch is that a real person wrote this book, that stings in a very specific way.
None of this is a judgment on your work, and it is not a glitch. Meta labeling content as AI when there is no AI in sight is a metadata problem, and once you see how it works, it stops feeling like an accusation.
Why your human-made post gets flagged
When Meta rolled its "Made with AI" labels out in 2024, the system was not reading your caption or evaluating your creativity. It was reading the technical metadata baked into your image file, using standards like C2PA and IPTC that record which tools touched the file along the way. The problem is that a lot of ordinary editing tools write AI-related tags into that metadata for edits that have nothing to do with generating an image from scratch.
The examples that broke this open were photographers. Meta labeled a real photo by former White House photographer Pete Souza, and a genuine press photo of the Kolkata Knight Riders winning a cricket final, both because the photographers had used everyday tools like Photoshop’s Generative Fill to do something minor, like clean up a background. The photo was real. The metadata said an AI-capable tool had been in the room, and the label followed.
The same thing happens to author graphics. Use an AI-powered "remove object" or a generative feature inside Canva or Photoshop on your teaser, and the export can carry a tag that trips the label, even though you wrote every word and chose every design element yourself. Meta heard the complaints and, by mid-2024, renamed the label from "Made with AI" to the softer "AI info," and later tucked it further into the post menu for content that was only edited rather than generated. The wording got gentler. The underlying trigger did not go away.
It is still inconsistent, too. The metadata standards do not fully talk to each other, so the same image can pick up a label on one Meta surface and stay clean on another. You are not imagining the randomness.
What it costs you, and what to actually do
The label is not primarily an algorithm penalty that buries your reach. The real cost is trust. A reader who is already primed to side-eye anything that smells like AI sees that tag, and for a beat they wonder whether your book is the real thing. That is the exact opposite of what you are on that platform to build.
Start by checking your own pipeline. If a graphic is genuinely your own work and only got flagged because you used an AI-assisted edit somewhere in the file’s history, re-exporting it through a tool that does not write those tags, or skipping the AI-specific feature on that particular asset, will usually clear it. This is not about hiding anything. It is about a file’s metadata telling a story that misrepresents what you actually made.
Do not delete good content over a label, and do not let it scare you off posting. The reach hit is smaller than the panic suggests. And if you do use AI tools in your graphics work, the strongest move is to own that plainly in your own voice rather than get quietly outed by a tag. Readers forgive a tool used honestly far more readily than one they feel they caught.
The bigger lesson sits underneath all of it. A label you cannot control, applied by rules you did not write, on a platform that can change them next week, is one more reminder that your visibility on Instagram is rented. Your email list is the one audience a metadata standard cannot touch. Meta labeling content as AI is annoying, and it is also a small, useful nudge to keep building the parts of your platform that no algorithm gets a vote on.
If you are trying to work out which of your marketing actually belongs on rented land and which needs to move somewhere you own, that is the untangling I do. Tell me what feels messy, and we will map it out.