OpenAI · ChatGPT Images 2.5

Remove Metadata from ChatGPT Images

Every image made with ChatGPT, Codex or the OpenAI API is saved with a signed C2PA manifest that names OpenAI's tools as the creator. Since 19 May 2026 OpenAI also adds Google's SynthID watermark to the pixels. BlankAI removes the manifest and any other metadata. It cannot remove SynthID, so OpenAI's and Google's checkers will still recognize the image.

Runs in your browser. Nothing is uploaded. Reviewed September 25, 2026

Check a ChatGPT image

Drop a ChatGPT download to see its C2PA manifest, then save a clean copy in the same format.

Drop images here or click to choose

JPG, PNG, WebP, AVIF, HEIC · Up to 20 images · Nothing is uploaded

EXIFGPSXMPIPTCC2PAPNG text

What OpenAI puts in ChatGPT images

Layer	Present	Details	BlankAI
C2PA Content Credentials	Yes	A signed manifest identifying the OpenAI product that generated the image, with an AI-generated action. OpenAI joined the C2PA steering committee in May 2026.	Removed
EXIF / XMP / IPTC label	Sometimes	Downloads carry little outside the manifest. The AI source type is recorded inside the C2PA data.	Removed
Invisible watermark	Yes	SynthID on ChatGPT, Codex and API images since 19 May 2026. It is designed to survive compression, resizing, cropping and screenshots.	Not removed
Visible watermark	No	ChatGPT images have no visible mark.	Nothing to remove

What OpenAI embeds, and why

OpenAI has attached C2PA credentials to its images since the DALL-E 3 era. In May 2026 it went further: every image from ChatGPT, Codex and the API now also carries SynthID, the invisible watermark Google developed, and OpenAI published a public verification tool. ChatGPT Images 2.5, released on 8 September 2026, keeps the same approach; OpenAI says ChatGPT and the API now produce more than 3 billion images a week.

These marks are not going away. The EU AI Act's marking duty has applied since August 2026 and California's AI Transparency Act requires covered providers to embed latent disclosures. OpenAI signed the EU's Code of Practice on AI-generated content.

Why clean a ChatGPT image at all

If you publish the image as AI-generated, you do not need to remove anything. Cleaning is useful when the file itself is the problem: some print services, marketplaces and content systems reject or mishandle files with embedded manifests, and manifests record tool details and timestamps you may not want to pass along. Designers who place a generated element into a larger composition also end up with a new file anyway, and BlankAI shows exactly what the original carried.

What stays after cleaning

SynthID stays. It is part of the pixels, and because OpenAI uses Google's system, both OpenAI's verification tool and Google's SynthID checks in Gemini can recognize a cleaned ChatGPT image. Removing metadata does not make a ChatGPT image undetectable, and no honest tool can promise that.

Platform rules also stay. TikTok, YouTube and Meta ask creators to label realistic AI content, whatever the file's metadata says.

How to remove metadata from a ChatGPT image

1Download the image from ChatGPT and drop it into the tool above.
2Review the report: it should list a C2PA manifest that mentions OpenAI.
3Keep “Same as original” to stay in PNG or WebP, then select Remove Metadata.
4Save the clean copy. The report re-scans it and confirms the manifest is gone.

What BlankAI removes, and what it does not

The same rules apply to every file, whatever made it. BlankAI cleans the metadata that travels with a file. It cannot reach anything that is part of the image itself.

Layer	Where it lives	What it holds	BlankAI
EXIF and GPS	File metadata	Camera or phone model, serial numbers, capture time and location coordinates.	Removed
XMP and IPTC	File metadata	Creator tool, edit history, captions, credits and the IPTC Digital Source Type AI label.	Removed
C2PA Content Credentials	Signed manifest inside the file	Who or what created the file and how it was edited, signed by the issuing app.	Removed
PNG text chunks	File metadata	Stable Diffusion prompts, seeds and model names, ComfyUI workflows.	Removed
Invisible watermarks	The pixels	SynthID (Google and OpenAI images), Meta's Content Seal, TikTok and ByteDance watermarks.	Not removed
Visible watermarks	The pixels	Logos, sparkle icons and corner badges drawn onto the image.	Not removed
AI classifiers	The platform	Models that judge an image by how it looks. File metadata plays no part.	Not affected

Questions

Does ChatGPT add a watermark to images?

Not a visible one. Since 19 May 2026 ChatGPT images carry an invisible SynthID watermark, alongside the C2PA metadata OpenAI already added.

Does removing metadata make a ChatGPT image undetectable?

No. It removes the C2PA manifest, but SynthID remains in the pixels and OpenAI's and Google's checkers can still identify the image.

What format are ChatGPT downloads?

ChatGPT usually saves PNG files. BlankAI keeps PNG by default so the clean copy stays lossless.

How can I check whether a ChatGPT image still has credentials?

Drop it into BlankAI or the EXIF Viewer, or upload it to contentcredentials.org/verify. A cleaned copy shows no manifest.