Author
Arshad Hossain
Facebook and Instagram ads specialist, Dhaka
Arshad Hossain works in paid social advertising on Facebook and Instagram, which are the platforms that apply the AI labels this site exists to deal with. Images published to Meta are read for provenance on upload, and that is a daily problem for anyone running creative at volume.
He built Remove AI Label after auditing the existing tools in this category and finding that the market leader reported a file as "clean for upload" while it carried GPS coordinates and an EXIF Software field reading "Adobe Photoshop 25.0 (Generative Fill)".
He wrote the parsers this site runs on: hand-written, dependency-free readers for JPEG marker segments, PNG chunks and WebP RIFF containers, built so that metadata removal never decodes or re-encodes the image. Every operation on this site is verified against that guarantee before it reports success.
What that work established
- Measured a leading competitor altering 52.3% of pixels on a single metadata strip, at a PSNR of 39.78 dB, while its page claimed the pixels were unchanged
- Documented that the same tool detected only C2PA, reporting plain EXIF, XMP, IPTC and PNG text chunks as "None" across all five of its reporting surfaces
- Wrote the JPEG, PNG and WebP container parsers used on this site, including the RIFF size and VP8X feature-flag corrections most tools omit
Those are measurements rather than opinions, and the method is published so anyone can repeat them. See how this site came about and what the tool cannot do.