快速入口: 本页是 AI Image Inspector 的专注入口 ChatGPT (DALL-E 3). 前往 AI Image Inspector→
Drop a PNG, JPEG, or WebP image
AI generation fingerprints are detected in your browser. Nothing is uploaded.
Check if an Image is from ChatGPT — Free ChatGPT Image Detector
预期用途
- The inspector reports fingerprints it finds in the file. It does not prove the image is real when no fingerprints are present — a re-encoded screenshot of a ChatGPT image will also return 'No AI generation traces found.'
- Screenshots of ChatGPT images typically strip the C2PA manifest, the EXIF block, and the XMP block. If the original file is no longer available, no structural detector can recover the fingerprint.
使用方法
Drop the image you want to test
Drag a PNG, JPEG, or WebP file from your desktop, your browser, a chat, or a saved folder into the drop zone. The image is read into a Uint8Array in your browser. Nothing is uploaded — the detection runs as pure JavaScript in the same tab.
Read the ChatGPT fingerprint report
The inspector checks for three ChatGPT-specific signals: the C2PA Content Credentials manifest (OpenAI signs every ChatGPT image with a JUMBF payload), the EXIF Software tag (older DALL-E 2 builds set Software to 'dalle' or 'DALL·E'), and the XMP CreatorTool field. Each detection is shown as a card with a 10-bar probability meter and the byte-level evidence. The highest-confidence card is the answer.
Verify and clean up
If the inspector reports a ChatGPT fingerprint, open the C2PA Stripper to remove the manifest, or the Strip Image Metadata tool to wipe the full metadata block. Re-run the cleaned file through the inspector to confirm the score has dropped below the detection threshold.
常见问题
How can I check if an image is from ChatGPT?
The fastest way to check if an image is from ChatGPT is to drop it into the AI Image Inspector on this page. The inspector reads the C2PA Content Credentials manifest that ChatGPT embeds in every image, plus the EXIF Software tag and the XMP CreatorTool field, and reports each detection with byte-level evidence. The whole check runs in your browser — no upload, no server, no account.
Does ChatGPT put a watermark on its images?
Yes. Every image ChatGPT generates carries a C2PA Content Credentials manifest. The manifest is a cryptographically-signed JUMBF block embedded in the file (in PNG iTXt/zTXt chunks, in JPEG APP11 segments, or in WebP 'EXIF' chunks). The manifest names 'OpenAI' as the assertion signer, 'DALL-E' or 'ChatGPT' as the generator, and the generation timestamp. A reader that does not understand C2PA will see the bytes but ignore them; the inspector extracts and verifies them.
Will this work for DALL-E 2 and DALL-E 3 images?
Yes. DALL-E 2 images are signed with the older 'dalle' or 'DALL·E' software tag in EXIF and a C2PA manifest in newer exports. DALL-E 3 images (the default ChatGPT image generator since late 2023) carry a C2PA manifest that names 'OpenAI' and 'DALL-E 3'. GPT-4o native image generation uses the same C2PA manifest format with the generator set to 'GPT-Image-1'. All three are detected by the same inspector.
What if the image was screenshotted?
A screenshot of a ChatGPT image is re-encoded by your operating system's screenshot tool. The re-encoding usually strips the C2PA manifest and the EXIF block, so the inspector will report 'No AI generation traces found.' The screenshot also changes the pixel data, which can defeat pixel-level watermarks. Screenshots are a known limitation of every structural detector — the only way to catch a screenshot is to keep the original file.
Is this a ChatGPT image detector?
It is a structural ChatGPT image detector. It reads the metadata, the C2PA manifest, and the software tag that ChatGPT leaves on its files. It does not run a neural network on the pixels, so it will not catch a ChatGPT image whose metadata has been deliberately scrubbed. For any image that still carries the manifest, the answer is unambiguous.
What does the verdict mean?
The verdict is the aggregate answer across all six detectors in the inspector. 'No AI generation traces found' means none of the detectors produced a non-zero score — the image does not carry any known ChatGPT, DALL-E, or other AI fingerprint. 'Likely generated by a single AI vendor' means one detector produced a high score — the card at the top names the vendor. 'Some weak signals detected' means at least one detector produced a low score — inconclusive, often a false positive on heavily-edited images.
Are my images uploaded to a server?
No. The image bytes are read into a Uint8Array in your browser. The C2PA reader, the EXIF reader, the XMP reader, and every other detector all run as pure functions in JavaScript. No network call, no server roundtrip, no third-party API, no telemetry, no account. The file never leaves your device, and nothing is stored between sessions. You can verify this by watching the Network tab in DevTools while you drop a file in.
限制说明
- Structural detection, not pixel-levelThe inspector reads metadata, C2PA manifests, EXIF tags, and XMP fields. It does not run a neural classifier on the pixel content. A ChatGPT image that has been re-encoded with all metadata stripped will be reported as 'No AI generation traces found.' The inspector catches any ChatGPT image that still carries at least one of the three fingerprints — it cannot catch a fully-cleaned ChatGPT image that has been re-encoded.
- Screenshots defeat detectionWhen you take a screenshot of a ChatGPT image, your operating system re-encodes the pixels and typically drops the C2PA manifest, the EXIF block, and the XMP block. The screenshot will be reported as 'No AI generation traces found.' The only way to keep the fingerprint is to keep the original file that ChatGPT produced (or the file you downloaded from a chat that preserved it).
- PNG, JPEG, and WebP onlyThe C2PA chunk reader works on PNG iTXt/zTXt chunks, JPEG APP11 JUMBF segments, and WebP 'EXIF' chunks. EXIF and XMP are read from JPEG APP1, PNG iTXt, and WebP 'EXIF'/'XMP ' chunks. AVIF, HEIC, and TIFF use different containers and are not decoded by this inspector. If your file is in one of those formats, convert it to PNG or JPEG first.
平台说明
- macOS
- Preview.app and Quick Look do not surface C2PA Content Credentials, EXIF Software, or XMP CreatorTool in their info panels. The inspector is the right tool for diagnosing the provenance of a ChatGPT image dropped from Safari, a chat, or a downloaded stock photo. To inspect a specific tag, right-click the file → Get Info → More Info shows EXIF but not C2PA or XMP.
- Windows
- Windows Photos and File Explorer show EXIF and basic XMP, but not C2PA. Right-click → Properties → Details shows the Software tag if it is set, but the C2PA manifest and the XMP CreatorTool are not visible. For full ChatGPT provenance, drop the file into the inspector — the report covers every ChatGPT fingerprint in one pass.
- Linux
- CLI alternative for some ChatGPT fingerprints: `exiftool -api RequestAll=3 image.png` shows C2PA claim chains and XMP, and will reveal the OpenAI assertion signer. It does not run the SynthID or dwtDct detectors, but those are not ChatGPT fingerprints anyway. The inspector is the right pick when libexiftool is missing C2PA support or when you want a single report that covers every fingerprint.
- Web
- Runs entirely client-side. The image is read as a Uint8Array, the C2PA reader, EXIF reader, and XMP reader all run as pure functions, and the report is rendered as HTML. No upload, no server roundtrip, no telemetry, no third-party calls. The file never leaves your device.