Checks images from ChatGPT, Midjourney and every major AI tool
AI generators leave names, prompts and generation details in the files they export, and isGenAI reads them: common Stable Diffusion prompt fields, ComfyUI and InvokeAI records, declared AI source types and signed provenance records. The report tells you precisely which field each finding came from, so a generator name in a filename never gets mistaken for evidence.
Check by AI generator (Midjourney, DALL·E, Flux, Firefly, Gemini and more), by file format, or by the platform the image came from. New to the terms? Start with the AI provenance glossary. Comparing tools? See the best free AI image detectors and how isGenAI compares with other detectors.
Check images without leaving the page you are on: add the free isGenAI Chrome extension and right-click any image.
ChatGPT and OpenAI images
Upload the original download to read every saved AI label and creation record it carries. Reposted copies carry different details from the original, and the report shows you which copy you checked.
Midjourney and Stable Diffusion
Find generator names and saved prompts in seconds. The report names the exact field it read, so you always know a saved statement from a signed record.
Edited and reposted pictures
Compare the original with the version you received. Matching copies and recorded edits add hard context, without relying on how realistic the picture looks.
PNG, JPEG, WebP, AVIF and TIFF
Upload one supported file up to 20 MiB, private by default. Classic single-page TIFF files up to 40 million pixels get full file evidence and pixel diagnostics. Multipage TIFF, pyramid TIFF and BigTIFF are rejected up front, never silently converted.
Free AI picture checker: is this picture AI?
Picture, photo or image: the check is the same. Upload the picture you want to verify, and the free AI picture detector reads what the file itself records about how it was made, free and without an account. It answers “is this picture AI generated?” with the evidence it found, names the field each finding came from, and says plainly when a copy carries no record at all.
For a picture that came from a social feed or a chat, check the platform page first: Instagram, WhatsApp, Telegram, Discord and Snapchat each keep or strip different records, so the page tells you which copy to ask for.
Fake photo detector: what it catches and what it cannot
As a fake photo detector, isGenAI catches pictures whose file records say they were made or changed with AI: a declared AI source type, a saved generation prompt, a signed Content Credentials record naming a generator or an AI edit, and copies that match earlier records. A photo edited by hand, or an AI picture whose records were stripped by a screenshot or a repost, carries none of that, so a clean result means no record was found, not that the photo is genuine. For those, ask the sender for the original file, run a reverse image search to find earlier copies, use the KYC guide when the picture is an ID or a selfie, and the fake receipt checker for receipts.
Eight measurements. A clearer picture.
Every report measures dimensions, resolution, average brightness, brightness spread, near-black and near-white pixels, transparency and edge differences. Compare versions of the same picture to see exactly what an export changed. These numbers describe visual quality; the origin answer comes from the file’s records.
How to tell if a photo is AI: start with the file
The most reliable way to tell is to read the evidence saved in the original file, not to judge how the photo looks. A saved generation prompt, a declared AI source type and a signed creation record each answer a different question, and the best AI image detector names the source of every finding. This report separates an editable statement from a record whose signature was checked, and keeps pixel measurements in their own section, where brightness, sharpness and resolution describe the picture rather than its origin.
Can a file show that ChatGPT made an image?
Yes, when the copy you have still carries a supported label or signed record. Upload the original download and the report lists every supported OpenAI label or signed record inside it. A product name in a filename is not evidence, and screenshots or reposts strip the records the original carried, so a clean result on a repost means the records are gone, not that the generator was never used.
A repeatable way to review a result
Detect AI-generated images with a method you can defend: compare the report with the source you received and the original file. Open each finding, read which checks ran and keep your notes with the report. Matching copies add context from existing records. Try the controlled examples to see how identical pixels can carry completely different records, then use the metadata comparison tool to inspect supported records before and after export.
Find the prompt saved in your image
Use the free image prompt extractor to read AUTOMATIC1111, ComfyUI, InvokeAI and NovelAI files and copy their saved instructions word for word.
Try two pictures with identical pixels
Our downloadable example shows how removing saved details changes a check while the picture stays pixel-for-pixel the same. The files are clearly marked demonstrations, with everything you need to reproduce the result.
Is this photo AI? Read the result with its evidence.
A check finds supported creation records, surfaces an AI-related statement or reports a clean copy. Read what was found and which checks completed, then decide with the evidence in front of you. A saved field is editable; a signature check and a signer-trust check answer different questions. The original publisher and the context the photo appeared in complete the picture.
What does “No AI evidence found” mean?
The checks read the file’s records and found no AI labels or creation details in that copy: a verified clean copy, not a guess. It is not a camera certificate, because screenshots and reposts strip records. Open the report to see which checks completed, then check the original file or source. Only have a repost? Follow the original-source checklist to document earlier appearances and credits.
Is this picture AI if it has a watermark?
First identify the signal: a visible logo, an invisible watermark and signed Content Credentials are different kinds of evidence. isGenAI reads Content Credentials and metadata declarations; Google’s SynthID detector is a separate tool. The AI watermark verification guide explains which check to use and how to read a missing signal.
What should I do when two AI image detectors disagree?
Check whether both tools received the same file and what each result actually measured. Keep the full explanations, inspect the original and compare them with the publisher’s account. Use the detector-disagreement workflow to choose the next check instead of averaging unrelated scores.
Free image tools that run in your browser
Need to fix the picture rather than check it? These free tools work on your device with no upload: convert WebP to PNG, convert PNG to ICO, pixelate an image, invert image colors, make a photo black and white and remove metadata from an image. See all free image tools.
Checking words instead? Open the AI text and code detector. Have a video clip? The AI video checker shows how to read MP4 and MOV records, Content Credentials and keyframes. For a closer look at file records, read how an image was made. To upload any image file and list every AI record inside it in one pass, use the AI file detector; to identify AI images, text and files from one page, start with the AI identifier.