AI Image Detector: Check if an Image Is AI-Generated

Check whether an image was made by an AI image generator such as Midjourney, DALL-E, ChatGPT, Gemini, Flux or Stable Diffusion. Upload, paste or link an image to get an AI likelihood score in seconds. Free, with no sign-up.

Drop an image here or paste it (Ctrl+V) JPG, PNG, WebP, GIF or BMP, up to 8 MB.

How to check if an image is AI-generated

  1. Drop an image into the box, click Choose image, or paste a screenshot with Ctrl+V. You can also paste a link to an image.
  2. Click Check image.
  3. Read the score and what it means. The result also says which checker was used.
  4. Use the original file when you can. Screenshots, crops and heavy compression remove some of the signals detectors look for.

How to read your result

ResultWhat it means
Likely not AI-generated (under 40%)Few signs of AI generation were found.
Unclear (40% to 74%)The signals are mixed. The image may be AI-generated, edited with AI, or a heavily processed photo.
Likely AI-generated (75% and above)The image shows strong signs of being made by an AI image generator.

How it works

Your image is checked by Sightengine, an image analysis company whose AI detection model is trained on images from many AI image generators. When that checker is busy or has reached its daily limit, the check runs instead on a free open-source model inside your own browser. That model is less accurate, and the result tells you which checker was used.

How accurate is it?

We tested both checkers on recent images from the OpenFake research dataset, made with 2026 generators such as Flux 2, GPT Image, Seedream 5, Sora 2, Ideogram and Z-Image, alongside real photos. The main checker judged all 16 images in its test correctly, scoring the AI images 99% and the real photos under 1%. The free in-browser model judged 73% of 97 images correctly and struggled most with Flux 2 images, so read its results with extra care.

No AI image detector can prove how an image was made. Treat the score as one signal, together with where the image came from, who posted it and whether other copies exist. Never accuse someone of using AI based on a detector score alone.

Why a real photo can score high

  • Strong filters, beauty retouching or AI-powered phone processing.
  • Photos edited with AI tools such as generative fill or background replacement.
  • Screenshots, heavy compression, or images resized many times.
  • Digital art, 3D renders and illustrations, which can share traits with AI images.

Other ways to spot an AI-generated image

  • Look closely at hands, teeth, ears, jewellery and text in the background, which AI tools often get wrong.
  • Check for lighting and shadows that do not match, or textures that look too smooth.
  • Run a reverse image search to find the original source and the earliest copy.
  • Look for Content Credentials or an AI label added by the platform or the generator.

Privacy

Images checked by the main checker are sent securely to Sightengine for analysis and are not stored by spellmistake.io. When the in-browser model is used, the image never leaves your device. Do not upload private or sensitive photos of other people.

Frequently asked questions

Is the AI image detector free?

Yes. There is no sign-up and no charge. To keep it free, each visitor gets a few checks a day with the main checker, and further checks run on the in-browser model.

Can it detect images made with ChatGPT or Gemini?

The main checker is trained on images from many generators, including ChatGPT (GPT Image), Google Imagen and Gemini, Midjourney, Flux, DALL-E, Stable Diffusion, Firefly, Ideogram and Seedream. New generators appear often, so results on the newest tools can be less reliable.

Can it check AI-edited photos?

It is built to spot images generated entirely by AI. A real photo with a small AI edit, such as a removed object, may score low or land in the Unclear range.

Can I check text for AI too?

Yes. Use our free AI detector to check whether a text reads like it was written by ChatGPT or another AI writing tool.