Five checks that are often confused.
| Mechanism | Purpose | Candidate takeaway |
|---|---|---|
| Resume parsing | Extracts details such as employment history into fields. | Check names, dates, titles and text order after upload. Extraction is not an authorship verdict. |
| Candidate matching | Compares information with criteria for a role. | Show relevant experience accurately. A fit assessment is not proof that the wording was generated. |
| Application rules | Can route or reject answers under employer-configured rules. | Read eligibility questions carefully and answer truthfully. |
| Fraud checks | Look for signals that an applicant or submission needs scrutiny. | Keep identity and experience consistent. Fraud signals and prose authorship are different questions. |
| AI text detection | Estimates whether text resembles machine-generated writing. | Do not treat a detector score as proof of authorship or as a hiring outcome. |
“This platform uses AI” does not answer “does this employer test resume authorship?” Features, integrations and employer policies vary. A vendor’s public description also cannot establish every private configuration.
What the vendor documentation actually establishes.
- Greenhouse parsing: its support documentation describes extracting resume details and correcting failed or partial imports. Parsing documentation.
- Greenhouse Talent Matching: its audit statement describes matching against recruiter-defined criteria and says this feature does not automatically advance or reject candidates. This is specific to Talent Matching, not every Greenhouse workflow. Talent Matching statement.
- Greenhouse application rules: a separate feature can reject candidates based on configured answers. Auto-reject rules.
- Ashby fraud detection: the documented checks use device, IP, email and phone signals, with recruiter review of findings. These are not described in that release as a test of resume prose authorship. Ashby’s feature announcement.
These examples explain specific mechanisms. They are not a certification that every ATS can—or cannot—detect AI-written resumes. We have not tested employers’ private screening configurations.
AI-detector accuracy needs context.
The RAID benchmark, published at ACL 2024, found that text detectors’ performance could weaken across unfamiliar generation models and changes in how text was produced. That supports caution about universal accuracy claims.
It does not provide a current rejection probability for your resume. A general text benchmark, a vendor’s marketing score and an employer’s decision are separate measurements. We do not convert them into a single “ATS detection accuracy” percentage.
Review an AI-assisted resume for evidence, not a score.
- Read the posting’s rules about AI assistance, including any separate assessment rules.
- Verify each employer, title, date, qualification and numerical result.
- Replace generic claims with a task you performed and an outcome you can substantiate.
- Keep terminology relevant to the role without adding experience you do not have.
- Check text extraction and the fields populated after upload.
- Keep a copy of the exact version you submitted.
Vague: “Leveraged innovative solutions to drive exceptional business impact.”
Specific, only if true: “Built a support dashboard that grouped open tickets by account and helped the team find overdue requests.”
The resume text-extraction checker can help you inspect a file, while our PDF and DOCX test explains the limits of a local parsing experiment. Neither reproduces an employer’s screening decision.
Common AI-resume questions.
They may notice generic wording, inconsistencies or details you cannot explain. They may also use screening tools. You cannot infer every employer’s setup from the ATS brand, so follow the employer’s instructions and verify every claim in your resume.
There is no universal rule across employers and systems. An employer’s application policy and configured screening process matter. AI assistance does not make an inaccurate qualification or fabricated achievement acceptable.
Prioritize accurate, specific writing you can explain. Chasing a detector score can make wording less clear and does not establish that the employer will accept the resume.