AI-generated images have exploded in popularity over the last couple of years. Tools like Midjourney and GPT are now creating visuals that blur the line between real and synthetic. As this kind of content becomes more common, it’s no surprise that people want a way to verify what’s made by a machine.
That’s where AI image detectors come in — and in this article, we’re comparing two of them: Undetectable AI’s image detector and AI or Not. Both tools claim they can reliably spot AI-generated images. We wanted to see how they handle images from one of the most advanced generators out there: Midjourney.
We ran 10 Midjourney-generated images through both detectors and recorded their responses. Some results were spot-on. Others were off. But overall? Let’s just say it made things interesting.
What is Undetectable AI
Undetectable AI originally launched as a tool to humanize AI-written content, helping it pass through detectors like GPTZero and Turnitin. But it’s recently expanded into new territory: image detection.
The image detector works similarly to their AI detection tool. You upload an image, and it scans for digital fingerprints that suggest it was generated by AI. These include things like texture uniformity, edge smoothness, and visual anomalies: factors most humans wouldn’t notice, but a trained model would.
Unlike some other tools, Undetectable AI also gives you a confidence score—telling you not just if an image was likely AI-generated, but how likely.
What is AI or Not
AI or Not is a standalone image detection platform built specifically for identifying AI-generated visuals. It’s widely used by educators, journalists, and fact-checkers, and is known for its simplicity: just upload an image, and it gives you a binary result—”AI” or “Human.”
The catch? Extra features are locked behind a paywall.
Undetectable AI vs. AI or Not: Midjourney Battle
We tested 10 images generated by Midjourney, all with realistic aesthetics — ranging from stylized portraits to detailed scenes. Each image was run through both Undetectable AI and AI or Not. Here’s how each tool performed:
Test #1
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #2
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #3
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #4
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #5
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #6
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #7
Undetectable AI: Incorrectly identified Midjourney image as human.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #8
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #9
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI Or Not: Correctly identified Midjourney image as machine-generated.
Test #10
Undetectable AI: Correctly identified Midjourney image as machine-generated.
AI or Not: Incorrectly identified Midjourney image as human.
Average Score
Both tools ended up with identical scores—but they didn’t miss the same images. Their strengths showed up in different places, which made this less about who “won” and more about how they differ.
The Bottom Line
If you’re trying to detect AI images—especially from Midjourney—both of these tools are viable. They’re not perfect, but they’re getting close.
Undetectable AI is better if you want more detailed feedback for free and confidence scores. That’s useful when you’re dealing with higher-stakes content or visuals that look borderline real. AI or Not is better when you want a quick yes-or-no answer without overthinking it.
But here’s what stood out the most: both tools scored a 90% accuracy rate. Which is great for AI detection as a whole.
Are both tools really that evenly matched? Or did we just happen to pick a batch of images that played to their strengths? Maybe. But either way, both tools delivered.
As AI visuals keep evolving, you can bet these detectors will need to keep up. We’ll be watching.
