DALL·E Detector
DALL·E 3 is one of the easier generators to confirm — OpenAI signs its output with C2PA Content Credentials and writes itself into the file. When that signature survives, the answer is immediate. When someone has stripped it, forensic analysis takes over.
The two ways to catch DALL·E output
OpenAI participates in the Content Authenticity Initiative, which gives you a shortcut that does not exist for Midjourney — but only if the file has not been laundered.
Signed C2PA manifest
OpenAI embeds a cryptographically signed manifest declaring the image as AI-generated, with a timestamp, using certificates on the C2PA Trust List. When present, this is definitive.
The Software field
Images produced through ChatGPT commonly carry a Software tag naming DALL·E 3. Our free EXIF viewer reads it in a second.
Signatures are trivially removed
Screenshot it, re-save it, pass it through a messenger — the manifest is gone. Absence of Content Credentials proves nothing at all.
Characteristic rendering
When metadata is gone, we fall back on pixels: DALL·E output tends towards smooth gradients and a clean, illustrative finish that our texture and frequency analyzers pick up.
Its own text problem
DALL·E 3 handles short text better than earlier models, but longer strings still degrade into plausible-looking nonsense.
No capture data
No camera, no lens, no exposure settings — the coherent block of shooting parameters a real photo carries is simply absent.
How the analysis works
Upload the original file
Full resolution, straight from the source. Screenshots and re-saved copies destroy most of the evidence we measure.
12 independent methods run in parallel
Three neural networks look at the image as a whole. Nine forensic methods examine it as a signal: frequency (FFT) analysis, noise pattern, pixel-level statistics, EXIF metadata, Error Level Analysis, texture coherence, colour distribution, Benford's Law and edge detection.
You see the breakdown, not just a number
Every method reports separately, so you can tell whether the verdict rests on one weak signal or on nine that agree.
When the metadata has been stripped, the answer has to come from the pixels — Start with our AI image detector — it covers any image type and runs the same twelve methods.
How much each method counts
Most detectors give you one number from one model. We run twelve independent analyzers and weight them — three neural networks carry the majority, but nine forensic methods examine the file as a signal and can overrule them. Here is the exact weighting we use in production:
| Method | Weight in final score |
|---|---|
| Neural network #2 | 25% |
| Neural network #1 | 15% |
| Neural network #3 | 15% |
| EXIF metadata | 10% |
| Frequency (FFT) | 6% |
| Benford's Law | 6% |
| Noise pattern | 5% |
| Error Level Analysis | 4% |
| Texture coherence | 4% |
| Edge detection | 4% |
| Pixel statistics | 3% |
| Colour distribution | 3% |
What people actually upload
These are our own numbers, not an industry estimate — every check run through FakeSec since launch. The split is closer than most people expect: roughly four in ten uploads really do turn out to be generated.
Where this comes up
- Newsrooms — An image circulating as a photograph.
- Education — Submitted work that was supposed to be original.
- Marketplaces — Product shots that were never photographed.
- Social platforms — Disclosure rules for synthetic media.
Accuracy — and where it breaks down
On our internal benchmark of 240,000 images (120,000 AI-generated, 120,000 real) the combined score reaches up to 98% on unedited, full-resolution files. That number drops, sometimes sharply, in four situations — and you should know them before trusting any result:
- Heavy compression. Social platforms re-encode everything. A photo pulled from Instagram has already lost much of its noise signature.
- Screenshots. A screenshot is a new image of an image. Original metadata and compression history are gone.
- Stylised content. Flat colour and clean lines remove the texture and noise evidence forensic methods depend on.
- Post-processing. Added grain, filters, upscaling and manual retouching all mask generation traces.
No detector on the market is 100% accurate — ours included. Treat a high score as strong evidence, not as proof. When a decision matters, look at which of the 12 methods agreed.
Order of operations matters here: check metadata first — it is free and instant — and if it is empty, run the full analysis. Empty metadata is the normal state of anything that travelled through social media, so it is a starting point, not a conclusion.
Frequently asked questions
Related tools
Check your first image now
Create a free account and get 3 checks. Full breakdown across all 12 methods, no card required.