Check text
Check text
Paste a passage on the AI text detector and Sens says whether it reads like ChatGPT, Claude, Gemini or another model wrote it. It is tuned to avoid false accusations and honest about what it cannot tell.
What to paste
At least 80 words and at most 10,000 characters. Longer passages give steadier results; under about 150 words every detector is shaky. For a long document, check a section at a time.
The Chrome extension adds Check this text: is it AI? to the right-click menu on selected text and opens the detector with the selection filled in.
How it decides
Unlike images, text carries no provenance to lean on, so the check is a statistical classifier:
- The free check runs an open-source classifier (desklib v1.01, MIT licensed) on Sens servers. It reads the text in chunks of about 500 words and scores how much each reads like a language model wrote it. It only says likely AI-written when the score is high enough that, in testing, no human writing reached it.
- The second opinion, for signed-in users, sends the passage to Pangram, a commercial detector that independent studies rate well ahead of open models on current-model text. It also marks which passages look AI-written or AI-assisted. Pangram is the stronger of the two, so its read leads; when the two disagree the verdict says so, and if only the free classifier sees AI the result is inconclusive.
The headlines
| Headline | Meaning |
|---|---|
| Likely AI-written | The classifier (or Pangram) scored the passage above the threshold no human text reached in testing. |
| Partly AI-written | Pangram found a mix of human and AI or AI-assisted passages. |
| Inconclusive | The signals are in the middle, or the two detectors disagree. |
| No clear signs of AI / No signs of AI writing | Nothing scored as AI. This is weak evidence: the check deliberately misses a lot of AI text. |
| Couldn't check | The classifier did not run (for example its server was cold or unavailable). Nothing was concluded; try again in a minute. |
Every result shows the AI-like score, the evidence behind the headline and the caveats, in the same shape as the image verdict.
Where it struggles
- Text from the newest models, which the free classifier often scores as human.
- AI text that has been paraphrased, heavily edited or run through a "humanizer".
- Human text polished with AI tools such as grammar rewriters.
- Short passages.
No tool can read the hidden watermarks Google, OpenAI or Anthropic add to some models' text, so every result is a statistical estimate. A high score is evidence worth a conversation, not proof. Do not use it on its own to accuse a student, a colleague or anyone else.
Limits
- Rate limit. Anonymous checks are capped at 20 per 10 minutes per address and signed-in users at 60. Past that you get "That's a lot of checks in a few minutes"; wait a moment or sign in.
- Second opinions. Free accounts get 4 Pangram second opinions a month, Pro accounts 100. A second opinion is counted when you ask for it, even if Pangram then fails. See Plans and limits.
What happens to your text
The text is sent to the Sens server to be checked and then discarded. A fingerprint (a hash) of it and the scores are kept for 90 days so checking the same text again is instant. A second opinion also sends the text to Pangram. Unlike image checks, nothing about text checking runs in your browser.