Frequently asked questions

Not completely. Every detector produces false positives (human text flagged as AI) and false negatives (AI text marked as human). They're useful as one signal among many — never as sole evidence, especially in academic or legal contexts.

It's a rough heuristic that runs in your browser. Good for curiosity and quick checks. For anything important, use commercial tools like GPTZero or Sapling — and still apply human judgment. See how our detector works.

No. Analysis happens entirely in your browser via JavaScript. We never receive your text on a server.

For most people: GPTZero (10k words/month) or Sapling (generous API). For unlimited quick checks: our built-in tool or ZeroGPT (less accurate). Full comparison on the compare page.

Yes, sometimes. Tools like Undetectable AI rewrite text to evade detection. In our tests, they fool some detectors more than others — but no humanizer beats everything. We test both sides honestly.

Detectors often confuse "formal, predictable grammar" with "AI-generated." Non-native speakers who write carefully can trigger false positives. This is a known, serious problem — don't accuse anyone based on a score alone.

With extreme caution. Use detection as a conversation starter, not a verdict. Talk to the student. Look at their draft history. Consider their writing history. A detector score is not proof of cheating. More in our guide for teachers (coming soon).

Eventually: Google AdSense and affiliate links where tools offer referral programs. We'll label affiliates clearly. The site needs to stay honest to be worth anything.

Because it was designed to. The domain is HumanProduced.org — the aesthetic is the brand. Every competitor looks identical. We'd rather look like a notebook than a template.

When tools change significantly — new models, pricing shifts, or when our quarterly re-tests show different results. AI detection is an arms race; last month's verdict might not hold.