Illuminarty is an AI content detector at illuminarty.ai. You upload an image (or paste text) and it estimates the probability that it was AI-generated. Paid plans add a map of which regions look generated and a guess at which model made the image. A high score is a probability, not proof, and it can land on real photos that were edited heavily. If Illuminarty flags an image you made and you want it to read as a photograph, fix the image: Imagera's AI Image Humanizer rebuilds skin, light and lens detail in one pass. For a second opinion, run the same file through Imagera's AI Image Detector.
What Illuminarty checks
From Illuminarty's own site, checked on 30 September 2026:
| Feature | What Illuminarty says | Plan |
|---|---|---|
| AI image classification | "Find out the probability of AI generation for a given image" | Free and paid |
| AI text classification | Probability for text, with the passages "most likely the result of AI generation" marked on paid plans | Free and paid |
| Localized image detection | Shows "which regions of the image have been generated" | Paid |
| Model identification | Estimates which AI model an image came from | Paid |
| API | Automated image checks from your own software | Paid |
| Browser extension | "We are working on a web browser extension" | Not yet available |
The site says it "combines various computer vision algorithms" for images and that basic AI detection "will always be free." Its own disclaimer says it makes "no representations or warranties" about the completeness or accuracy of the information on the site.
How to read an Illuminarty result
- A probability, not a verdict. A high number means the image shares features with generated images the tool has seen. It doesn't prove how the image was made.
- Heatmaps show where, not why. A hot region around a face or hair usually means that area has the smooth, even texture generators produce.
- Model guesses are guesses. Identifying the generator is harder than spotting generation, so weigh that label lightly.
- Real photos can score high. Beauty filters, AI retouching and heavy compression all push scores up.
Imagera's detectors carry the same caveat: Imagera does not publish an accuracy percentage, because it has not run a benchmark that could support one. See our AI image detector comparison for why tools disagree on the same file.
Why AI images get flagged
Region maps tend to light up in the same places people notice first:
- Skin: poreless, waxy, evenly lit
- Hair: strands that melt into each other or the background
- Eyes and teeth: too symmetrical, too clean
- Backgrounds: smooth gradients with no grain or lens falloff
Before and after from the AI Image Humanizer demo: same photo and pose, only the skin rendering and grain change.
Fix the image, not the score
- Repair broken anatomy first. Faces, hands and melted text go to the AI Anatomy Fixer.
- Humanize the surface. Open the AI Image Humanizer in Sandbox, add the image and run it. It reads the photo first, then removes the plastic look from skin, light and lens. The credit cost shows before you run.
- Use grain only when composition is locked. Real Film Grain adds camera sensor grain without moving anything.
- Check at 100% zoom for faces, hands, text and edges before you publish.
- Disclose AI use where a platform, contest or client requires it.
No tool can guarantee a particular Illuminarty score. Detectors update, and results change with every export and crop. The goal is an image that looks photographic to the people who see it.
Checking someone else's image
If you're verifying an image you received, don't rely on one tool. Run it through Illuminarty and Imagera's AI Image Detector (15 credits a scan, with the visual signals behind the score), and check the source: where it was first posted and whether the original file is available. For a face that may be swapped, use the deepfake detector.
Credits come with the Starter plan at $19.99/mo for 300 credits, or in packs from $39.99 for 350 credits. See pricing.
