AI Text Detector — Free & Explainable

An AI text detector takes any text — essay, email, blog draft, cover letter — and estimates the probability it was machine-written. "AI text detector" is the umbrella term searchers use (182,000 searches a month by itself); this page is the practical version: what detectors measure, which text types get falsely flagged, and a free one you can use right now.

The two numbers every detector produces

Under the branding, detectors converge on two measurements. Perplexity: how predictable each word is given the previous ones — models pick high-probability words, so AI text reads "low-perplexity." Burstiness: how much sentence lengths vary — humans fluctuate wildly, models stay steady. Our detector reports the burstiness-style rhythm variance openly, alongside AI-vocabulary density and per-line flags, so you can see the why behind your number.

Which texts get falsely flagged

• Formal business and academic writing — trained on the same patterns as model output.
• Non-native English — cleaner, simpler sentence structures score as machine-made.
• Technical documentation — passive voice plus standard phrasing.
• Short texts — under ~100 words, statistical signals get noisy.

If your own writing keeps getting flagged, that's the detector's bias, not your guilt. Check it, fix the specific lines, keep your drafts — the full playbook is in why detectors flag human writing.

Free detector, no login

Paste up to any length (detection is unlimited): get a 0-100 score, the flagged terms in your text, your burstiness value, and the five most AI-sounding lines. Then, if you want, humanize in place and watch the score move. Also see the essay-specific guide or the tool comparison.