Claude Detector: was this written by Claude AI?
Check essays, articles and reports for text written by Anthropic's Claude. TextFlag also detects ChatGPT, Gemini and DeepSeek.
Detecting Claude-written text
Claude writes in a natural, often conversational style, which can make it harder to spot by eye than other AI writers. TextFlag's score is calibrated with real Claude text, so Claude-written passages are part of what the detector was tuned on, not an afterthought.
How well does it detect Claude?
We publish our weak spots. In calibration, fully Claude-written documents read about 92% on average rather than 100%, so expect Claude text to score high but not always at the top of the scale. Across all model families, nine documents in ten land within 9 percentage points of their true AI share.
| Mode | Flagged sentences that are AI | AI sentences caught | Human sentences flagged by mistake |
|---|---|---|---|
| Standard (default) | 95.3% | 41.2% | 1.2% |
| Thorough | 70.5% | 80.5% | 20.5% |
Measured on 132 held-out documents (46,862 sentences). Full methodology →
When to use Standard or Thorough
Use Standard when you are assessing someone else's work: it highlights only highly confident runs of sentences and rarely flags human writing. Use Thorough on your own draft to surface every passage that might read as Claude-written.
Keep exploring
Claude Detector FAQ
Can AI detectors detect Claude?
Yes, if they were trained or calibrated on Claude text. TextFlag was calibrated with Claude output and highlights Claude-style sentences, although pure Claude text averages about 92% rather than 100%.
Is Claude harder to detect than ChatGPT?
In our calibration, pure Claude documents scored slightly lower on average than other model families, which is why we report the result with a margin of error.
Is the Claude detector free?
Yes. Up to 10,000 words per check and 20 checks a day, with a downloadable PDF report.
Check your text for AI in seconds
Free for up to 10,000 words per check. No account needed.