How to remove the ‘Made with AI’ label on Instagram & Facebook
Short answer: Meta’s “Made with AI” label is triggered by C2PA content credentials baked into the file by the tool that generated or edited the image. Re‑encoding the image through a browser canvas drops those credentials, and the label stops appearing on new uploads. You can do it in five seconds at aiinforemover.com without uploading anything.
What actually triggers the label
When Instagram or Facebook decides to add “Made with AI,” they aren’t looking at the pixels. They’re looking at a C2PA manifest — a small, cryptographically signed block of metadata attached to the file.
The tools that embed these manifests today include:
- Adobe Firefly — everything it generates is signed.
- Photoshop Generative Fill — signs edits made with the Generative Fill brush.
- OpenAI’s ChatGPT image generator (DALL·E 3 backend) — signs since spring 2024.
- Microsoft Designer / Image Creator — signed via Microsoft’s C2PA implementation.
- Some Google models — selectively signing rolled out in 2024.
When Meta’s pipeline ingests an image, it parses any C2PA manifest it finds. If the manifest declares an AI tool as the author, it sets the label.
Why a browser canvas removes the credential
C2PA manifests travel outside the pixel data. In a JPEG, they live in an APP11 marker segment; in PNG, in a caBX chunk. The pixel grid itself doesn’t carry the signature.
When a browser draws an image onto a <canvas> and exports it with canvas.toBlob(), it re‑encodes a brand‑new file from the in‑memory bitmap. There’s no metadata to copy because the bitmap doesn’t have any. The C2PA manifest gets left behind.
This is what AI Info Remover does in one click. No server, no upload, no account.
Step‑by‑step
- Open aiinforemover.com on your phone or desktop.
- Drop or paste the AI‑generated image into the dropzone.
- Click Download on the cleaned file. The file is now C2PA‑free.
- Post that cleaned file to Instagram / Facebook. The “Made with AI” label should not appear.
Caveats — the parts honest articles skip
- Platforms change. Meta has hinted at additional detection signals (pixel‑level watermarks, model fingerprinting). If those ship, removing C2PA alone won’t be enough. Test on a throwaway post first.
- SynthID‑style watermarks are baked into the pixels themselves. Re‑encoding through JPEG quantization sometimes weakens them, but not reliably.
- Honesty is fine too. This guide is for cases where the label is wrong (you took the photo, then used Generative Fill to remove a power line) or genuinely misleading. We’re not trying to help anyone pass off a generated image as a documentary photo.
Why not use Photoshop’s “Export As”?
Photoshop’s own export still writes a C2PA manifest unless you go into preferences and disable it. Most users never find that toggle. A browser canvas re‑encode bypasses the question entirely.
TL;DR
The “Made with AI” label is metadata, not magic. A 30‑line HTML file in your browser strips it. Use AI Info Remover and the next post is invisible to the detector.