Can ATS Detect AI Resumes? What 6 Systems Do
ResumeAdapter EditorialShare this post
Send this to a friend who’s also job searching.
No major applicant tracking system documents a feature that detects AI-written resumes. We read the AI documentation of six of the most used systems on October 17, 2026: Workday, Greenhouse, iCIMS, Lever, Oracle Recruiting and SAP SuccessFactors. Their AI does something else. It reads your skills and experience and compares them with the job. So the question that decides your application is not who wrote your resume, but whether it proves what the posting asks for.
In short: ATS software grades and ranks, it does not detect. Workday's HiredScore grades you A to D on the job's qualifications. Greenhouse says it never uses AI to rate candidates or auto-reject. Oracle Recruiting even offers candidates AI to write a summary of their resume. AI detectors are a poor fit for resumes: OpenAI withdrew its own for low accuracy. The real risk of AI is a line you cannot defend in the interview.
Used AI on your resume already? Check your resume against the job you're applying to: paste the posting and see which of its required skills and keywords your resume still does not state.
Table of contents
- What an ATS does with your resume
- What 6 major systems do with AI
- Detectors and graders are not the same thing
- Why AI detectors are a poor fit for resumes
- What recruiters notice instead
- The real risk: a line you cannot defend
- Worked example: one bullet, three versions
- How to use AI on your resume honestly
- FAQ
What an ATS does with your resume
An applicant tracking system is the employer's database of applications. When you apply, it does four things with your resume:
- Parses it. It reads the file and pulls out your titles, employers, dates, skills and education into fields.
- Stores it next to your answers to the application questions.
- Lets recruiters search it. A recruiter looking for "Salesforce" or "variance analysis" finds the resumes that state those words.
- Sometimes grades or ranks it against the job, if the employer turned on the system's AI matching.
None of these steps asks who wrote the text. Parsing cares about layout, search cares about words, and grading cares about qualifications. Automatic rejections, where they exist, come from rules the employer sets, such as a "No" to a work authorization question. Our guide to ATS knockout questions lists the common ones.
What 6 major systems do with AI
Here is what each vendor's own documentation says its AI does, read on October 17, 2026. The second column is the question you came for.
| System | Documents an AI-writing detector? | What its AI does, in its own words | Who decides |
|---|---|---|---|
| Workday (HiredScore) | No | Assigns a grade "based on how well their resumes match the skills and qualifications specified in the job description", A to D | Workday describes human review as part of the workflow |
| Greenhouse | No | Talent Matching surfaces "discrete categories (e.g., Strong, Good, Partial, Limited) with explanations rather than composite fit scores" | "AI and automation can inform, summarize and surface insight, but it is never the final decision-maker"; in January 2025 it told candidates it "never uses AI to rate candidates or auto-reject applications" |
| iCIMS | No | "Rank candidates based on skills and experience" | "Recruiters stay firmly in control" |
| Lever | No | Talent Fit "uses AI-powered matching and customizable criteria to surface high-potential candidates" | "Recruiters always have the final say" |
| Oracle Recruiting | No | AI Assist generates a work summary for the candidate "based on the candidate's resume to match the job that they're applying for" | The candidate reviews and edits it before applying |
| SAP SuccessFactors | No | Applicant screening "identifies direct skill matches, related skills, and skills the applicant may easily acquire" and "extracts skills from resumes" | Helps "recruiters and hiring managers make better decisions" |
Sources for every row are listed under Primary sources. "No" means none of the pages we read describes such a feature. A vendor could add one, and an employer could run a separate tool outside its ATS, so read this table as what the systems document today, not a promise about every employer.
The Oracle row deserves a second look. Oracle's own recruiting software offers to write part of your application with AI, from your resume, and attaches it to your profile "as a cover letter for the application". A system that writes AI text for candidates is not built to penalize candidates for AI text.
Detectors and graders are not the same thing
Most of the fear comes from mixing up two kinds of AI.
- A detector tries to answer "did a machine write this?". None of the six systems documents one.
- A grader tries to answer "does this person match this job?". Several of them have one, and it is the one that matters to you.
Workday's HiredScore is the clearest example of a grader. Its documentation defines the grades this way:
| Grade | Workday's definition |
|---|---|
| A | "Candidate meets all basic (required) qualifications, most preferred qualifications, and has an average qualification match above a defined threshold." |
| B | "Candidate meets all basic qualifications." |
| C | "Candidate meets most but not all basic qualifications." |
| D | "Candidate does not meet most basic qualifications." |
Nothing in those definitions looks at style. A grader reads what your resume claims and checks it against the job's required and preferred qualifications. That cuts both ways. A human-written resume that never names a required skill grades low, and an AI-polished one that names it grades higher. Workday also published a summary of an independent analysis of HiredScore Spotlight, run by the firm Secretariat on U.S. applicant data for jobs in the greater New York City area, which found "no evidence of disparate impact" in the ratios it reported.
What graders reward is the match. Across the first submissions we analyzed for our ATS statistics, the average resume does not state 54% of the keywords in its target job description. That gap, not authorship, is what a grader sees.
Why AI detectors are a poor fit for resumes
Suppose an employer did run a resume through a separate AI detector. The best evidence on how well that works comes from OpenAI, the company behind ChatGPT, about its own classifier.
- On OpenAI's test set, it correctly flagged 26% of AI-written text and wrongly labeled human-written text as AI 9% of the time.
- OpenAI wrote that it was "very unreliable on short texts (below 1,000 characters)" and that AI text "can be edited to evade the classifier".
- On July 20, 2023, OpenAI withdrew it: "the AI classifier is no longer available due to its low rate of accuracy."
A resume is the hardest case for a detector. A bullet point is one or two lines, and a whole job entry is often under 1,000 characters. Resumes are also written in a formulaic style on purpose (verb, task, result), which is exactly the predictable text a detector struggles to tell apart. So a "98% AI" score from a free detector tells you very little, and paying a "humanizer" to rewrite your lines can change your facts without making anything truer.
What recruiters notice instead
Software is not the problem. People are, in a useful way. Recruiters read many AI-polished applications, and three things stand out:
- Generic lines. "Results-driven professional who leverages cross-functional synergies" could describe anyone, so it proves nothing.
- Claims without evidence. A 35% improvement with no scope, no method and no context invites the question "how?".
- A resume that disagrees with you. Titles that do not match your LinkedIn profile, tools you cannot discuss, numbers you cannot explain.
Some employers have also started asking directly. Application forms and policies can ask you to certify how you used AI in your application, and federal essay questions ask you to certify that answers are your own. If you are asked, answer the question as it is written. Our guide to tailoring your resume with ChatGPT covers what those certifications look like.
The real risk: a line you cannot defend
The danger in using AI is not that a machine catches you. It is that the AI writes something you did not do, and the interviewer asks about it.
Asked to "make this resume match the job", a chatbot tends to:
- move your title toward the target title,
- round up or invent results,
- add tools and certifications the posting names,
- upgrade verbs ("supported" becomes "led", "built" becomes "managed").
Every one of those changes can lift a grade and then fail in a 30-minute interview. Our rule for every line is simple: adapt the resume, never invent the candidate. Put your real experience in the posting's words, and leave out what you have not done.
Worked example: one bullet, three versions
Everything below is fictional: the candidate, the employer and the numbers. It shows the mechanics, not a real ResumeAdapter result.
The candidate. Marcus has four years in customer support, the last two as escalation lead for 40 business accounts. He is applying for a Customer Success Manager role whose posting asks for "account management", "quarterly business reviews (QBRs)", "churn reduction" and "Salesforce".
His original bullet: "Handled escalated tickets for 40 business accounts and ran monthly check-in calls."
What a chatbot gave him when he asked it to match the job: "Spearheaded enterprise customer success strategy in Salesforce, reducing churn by 35% across a 40-account portfolio."
It reads well and hits four keywords. It also invents a strategy role, a 35% result Marcus never measured, and a tool he has never used. No ATS flags it. The interviewer will: "Walk me through how you cut churn by 35%."
The honest adapted version: "Managed 40 business accounts as escalation lead; turned the monthly check-in call into a quarterly business review (QBR) format the support team adopted."
This states "account management" and "QBRs" in the posting's words, and every word is defensible. "Salesforce" stays out because Marcus has not used it, and "churn reduction" stays out because he never measured it. If he can find a real retention number in his account records, he can add it then.
How to use AI on your resume honestly
- Start from your real resume and the full posting. Paste both, not a summary of either.
- Ask for a match before a rewrite. Have the AI list each requirement in the posting next to the evidence your resume already has, and mark the ones with no evidence.
- Rewrite only the matched lines. Ask it to keep every title, employer, date and number exactly as they are.
- Check every changed line with four questions: Did I do this? Can I explain it with an example? Is the number mine? Does it use the posting's term?
- Check the result against the posting. Check your resume against the job you're applying to to see which required skills and keywords it still does not state, then add only what you can support. ResumeAdapter cannot see the employer's setup or rules, and it cannot guarantee an interview.
- Make your LinkedIn profile agree. Same titles, same dates, same skills.
Our step-by-step prompt for the first three steps is in How to Tailor Your Resume With ChatGPT.
Related reading
- How to Tailor Your Resume With ChatGPT, Then Check It gives the prompt that keeps the AI honest and the check before you send.
- Do ATS Automatically Reject Resumes? explains what Workday, Greenhouse, iCIMS and Taleo reject on their own.
- ATS Knockout Questions lists the screening questions that decide most automatic rejections.
- What Is a Good ATS Score? shows where your match score sits among real first submissions.
- ATS Keywords List shows how many keywords a resume needs, counted from real postings.
- ATS Optimization Hub collects every ATS guide in one place.
Primary sources
All pages read on October 17, 2026.
- Workday HiredScore documentation: Candidate grades
- Workday: Responsible AI and bias mitigation for HiredScore Spotlight
- Greenhouse: Our AI principles
- Greenhouse: How does Greenhouse use AI? Here's everything candidates need to know (January 8, 2025)
- iCIMS: AI recruiting software
- Lever: AI recruiting features
- Oracle Recruiting: Add work summary to job applications using AI Assist
- SAP: Applicant screening in SAP SuccessFactors Recruiting
- OpenAI: New AI classifier for indicating AI-written text (January 31, 2023, with the July 20, 2023 update)
Before you send it
No ATS is checking who wrote your resume. It is checking what your resume proves against one job, and a person will check it again in the interview. Use AI for the wording, keep every fact your own.
Check your resume against the job you're applying to and see which requirements your resume leaves out before a recruiter does.
Frequently Asked Questions
Common questions readers ask about this topic.
Can an ATS detect an AI-written resume?
Not as far as their makers document. We read the AI documentation of Workday, Greenhouse, iCIMS, Lever, Oracle Recruiting and SAP SuccessFactors on October 17, 2026, and none of them describes a feature that checks who wrote a resume. Their AI matches, grades or ranks your stated skills and experience against the job. Oracle Recruiting even offers candidates an AI Assist feature that writes a work summary from their resume.
Can recruiters detect ChatGPT?
Not reliably with software. OpenAI withdrew its own AI-text classifier in July 2023 because of its low rate of accuracy: it caught 26% of AI-written text and wrongly flagged 9% of human text, and it was very unreliable below 1,000 characters, which is longer than most resume sections. What recruiters do notice is generic wording, claims with no evidence behind them, and a candidate who cannot explain their own resume in the interview.
Do job recruiters care if you use AI?
Most care about what the resume claims, not how it was typed. The ATS vendors we checked all say a person makes the decision, and their AI compares your qualifications with the job. Where AI use becomes a problem is generic text that reads like every other application, and invented details you cannot defend. Some employers also ask you to certify how you used AI; if an application asks, answer truthfully.
Does ATS reject AI-generated resumes?
No documented ATS rejects a resume for being AI-written. Automatic rejections come from rules the employer sets up, such as knockout questions on work authorization, location or years of experience. Greenhouse states it never uses AI to auto-reject applications, and Workday's HiredScore grades candidates A to D on how well their resume matches the job's qualifications, not on who wrote it.
Should I run my resume through an AI detector or a humanizer?
There is little point. The systems that screen you do not document a detector, and detectors are unreliable on short text like resume bullets, so a score from one tells you little about how your application will be read. A humanizer that rewrites lines to fool a detector can also change your facts. Spend the time checking that every line is true and states the posting's requirements in its own words.
Should I say that I used AI on my resume?
There is no general rule that you must. If the application form or the employer's policy asks you to certify how you used AI, answer the question as it is asked. Either way, the safe test is the same: you should be able to explain every line of your resume in your own words, with an example, if an interviewer asks.
Can ResumeAdapter make my resume undetectable?
No, and it does not try to. ResumeAdapter compares your resume with the job posting you paste, shows which required skills and keywords your resume does not state, and helps you adapt your real experience to the posting's words without inventing any. It cannot see an employer's setup or screening rules, and no tool can guarantee an interview.