Applications do not show the whole process
You may be reading a cover letter, written response or work sample without knowing how it was prepared. Our detector reads patterns in the passage, not the applicant's ability, honesty or suitability for a role. Use a result only as context for questions that let the applicant explain their work.
Applicants writing in a second language need the same opportunity to explain their choices. Our language calibration includes non-native writing, and we publish false flags on human writing. Errors remain possible. A formal style or marked passage should not become a shortcut for judging someone before you have spoken with them.
Turn passages into fair questions
Review what you need
Choose a relevant writing passage and consider the information you are submitting. Understand our text handling before pasting application material. Keep the review focused on the role and the work.
Read the limits with the result
Check confidence and range before sentence marks. Very short text gets no reading, and short text is rated Low. Do not treat a headline as a conclusion about an applicant.
Prepare an open question
Use a marked passage to ask how the applicant developed an idea or supported a claim. Give them room to explain. Paid batch checks do not replace that conversation.
Example: Discussing a written work sample
In this fictional example, you check a written work sample that shifts from a formal introduction to a practical explanation. The headline is "Mixed: part reads AI, part human" and "Our confidence" is Medium. You examine sentence marks and "Most machine-like sentences" to understand where the reading changes. You do not translate those marks into a judgment about the applicant's honesty, ability or suitability for the role.
You open "Why this reading" to review the length band and the independent model's reading. In the interview, you ask the applicant to explain a recommendation from the sample and discuss the reasoning behind it. You consider that explanation against the role's requirements and the wider application. You keep the detector reading separate from your assessment notes so it does not become an unsupported conclusion about how the applicant worked.
A conversation about the work
The useful outcome is a question you can ask, not a rejection label. Keep the reading separate from judgments about competence or honesty. Consider the applicant's explanation and the wider application. A detector result is not proof about a person and must never be the sole basis for a hiring decision.
The same figure can carry a different confidence rating when the range changes. A work sample that shifts register can have a wider range than an even passage. Do not compare applicants by their figures or rank them using detector results. The reading concerns the submitted text, not the person, and differences between readings do not establish differences in skill, effort or suitability for a role.
Questions
Should I reject an applicant because of a reading?
No, a reading is not a reason to reject an applicant. Human writing can be flagged and AI writing can be missed. The result does not measure ability or establish how the application was prepared. Use it, if relevant, to ask about the work rather than make a hiring decision from it.
What happens to pasted application text?
Pasted text is processed to provide the result and is not stored except in the history of a signed-in account. You can delete one entry or that whole history at any time, on any plan. We do not train models on your text, and we do not sell or share it. Review what application information you need to submit before pasting.
How can sentence marks help an interview?
Sentence marks can help you select a passage to discuss with the applicant. Ask how an idea developed, what supports a claim or why a particular explanation was chosen. Do not present the mark as proof. Read the applicant's answer alongside the work and account for writing in another language.
Can I check applications in a batch?
Yes, every paid plan can check several texts at once. Batch checks do not remove the need to consider each application's context or the reading's limits. For a separate length check, the free word counter has no cap and keeps typed or pasted text in your browser. Neither tool determines suitability.
Does the detector identify which AI tool was used?
No, the detector does not identify the tool or model that wrote a passage. It reads writing patterns rather than a model's signature, and it does not establish that AI was used at all. Avoid asking an applicant to explain a specific tool based on the result. Ask about the substance and development of the work without presenting an unsupported premise as fact.
Can a printable report establish how an application was written?
No, a printable report records a reading, not the application's authorship or drafting process. Pro reports include the range, confidence rating and marked text, along with language and calibration details, model version and date. They stay in your browser. If you retain a report, keep the applicant's explanation and relevant context beside it rather than treating the document as independent proof.
What should I review before relying on detector testing?
Review the accuracy page for the text's language, including false flags and the distinction between calibration results and held-out tests. We publish dated results and identify weaker performance on very short text and older model generations. Those tests describe performance on test material, not an individual applicant. They do not justify turning a reading into a hiring rule or a ranking of candidates.
Tools for this work
- AI detectorOne figure with our confidence rating and a mark on every sentence.
- Word counterWords, characters, sentences and reading time as you type.
Free checks with room for context
Free use needs no card. Each tool has a daily free check with a word cap; the grammar checker and both counters have no cap. Paid plans add batch checks. Pro adds printable reports that you can keep beside, not instead of, review notes.