What the hiring system can see
Your application goes into an applicant tracking system, such as Greenhouse, Lever, Ashby or Workday. It stores your file, your answers to the job's questions and the date you applied. Greenhouse's own page for job seekers says applications there “are reviewed by real people, one by one, in the order received,” and that “AI doesn't score or rank applications” (Greenhouse).
There is automatic filtering, but it runs on answers, not on writing style. Greenhouse's application rules let an employer auto-reject a candidate based on the response to a custom question, and the rule can set a rejection reason and send the email. That is a recruiter's rule about work authorization or a licence, not a judgement about who wrote your sentences.
One thing the system does show is your other applications. In Greenhouse, the Application History section “provides an overview of other jobs the candidate has applied to,” for reviewers with the right permissions. So the system can't see your AI use, but it can see that the same person applied to five different roles there this week.
What Baela does here: it applies only to jobs that fit your titles, level, location and salary, and never applies to the same job twice, so your Application History doesn't fill up with roles you don't fit. It can't change how an employer reviews what it receives.
Do AI detectors work? Not well enough to trust
Tools that claim to spot AI writing exist, and some employers have tried them. The public record on their accuracy is poor.
- OpenAI pulled its own. OpenAI launched an AI Text Classifier in January 2023 and withdrew it that July “due to its low rate of accuracy.” Even at launch, it labelled human text as AI about 9% of the time (The Decoder, July 25, 2023).
- The strongest vendor claim still means mistakes. Turnitin, whose detector is used across schools, says its document-level false positive rate “is less than 1% for documents with 20% or more AI writing,” and that its “sentence-level false positive rate is around 4%” (Turnitin). On a one-page resume, 4% of sentences is not a rounding error.
- A large user switched it off. Vanderbilt University disabled Turnitin's AI detector and wrote that “AI detection is already a very difficult task for technology to solve (if it is even possible),” and that “we do not believe that AI detection software is an effective tool that should be used” (Vanderbilt, August 16, 2023).
- They are biased against non-native English writers. A 2023 study by Stanford researchers found detectors “consistently misclassify non-native English writing samples as AI-generated,” while native samples were classified correctly. The authors advise against using these tools “in evaluative or educational settings” (Liang et al.).
Read that last one twice if English is your second language. A detector can flag careful, plain writing as machine-made when no machine was involved.
Honest limit: no tool, Baela included, can promise a detector will never flag your writing. Baela changes the thing a detector is a bad guess at: whether the words are true about you. Every line it writes comes from your own experience, and you can read it before and after it goes out.
What recruiters actually notice
Not the tool. The result. Four things stand out from a pile of applications.
1. Wording that could describe anyone
“Results-driven professional with a proven track record of driving cross-functional impact” says nothing. It fits a nurse, a controller and a sales lead equally well. A recruiter reading hundreds of applications passes over it quickly, not because it looks AI-written, but because it gives nothing to judge. The fix is specifics only you have: the systems you ran, the size of the team, the market you covered.
What Baela does here: it rewrites your real lines in the job's own words rather than generating a generic profile, so the specifics stay yours.
2. A resume that doesn't match you in the interview
This is the real test, and it's not automated. If your resume says you led a migration, you will be asked what broke and how you decided. In Greenhouse's 2025 AI in Hiring Report, 91% of U.S. recruiters said they had spotted candidate deception, and 65% of hiring managers said they had caught applicants using AI deceptively. A follow-up question costs a recruiter nothing and beats any detector.
What Baela does here: it never adds jobs, skills, numbers, titles or results you don't have, and a second check throws out any rewrite that claims more than your original line. So what you walk into the interview with is what you can talk about.
3. The same text, everywhere
Identical applications are visible in a way AI use isn't. A cover letter addressed to the wrong company, a summary that names a skill the job never mentions, five applications to one employer in one hour: all of that is on the record. The same Greenhouse survey found 34% of U.S. recruiters spend up to half their week filtering “spam and junk applications.”
What Baela does here: each application is tailored to that job, and Baela paces itself so an employer sees a real candidate rather than a burst. It never applies to your current employer, and you can block companies you don't want it to approach.
4. Tricks aimed at the software
Hidden white text, keyword stuffing and instructions aimed at an AI screener are a different thing from writing help, and employers are watching for them. Greenhouse's 2025 survey reports 41% of U.S. job seekers admitting to “prompt injections, hidden text designed to bypass AI filters.” If an employer finds that in your file, it looks like an attempt to cheat the process, because it is.
What Baela does here: it never uses hidden text, white-on-white keywords or instructions aimed at an employer's software. Your resume is a resume a person can read.
The real risk is lying, not AI help
An employer who finds out your resume claimed something untrue can drop you at any stage, including after an offer. The checks that find it are ordinary: an interview question, a reference call, an employment verification.
For a federal job in the United States, the stakes are written into law. Under 18 U.S. Code § 1001, knowingly making a materially false statement in a matter within the jurisdiction of the federal government can be fined or punished by up to five years in prison. That applies to what you sign, no matter which tool helped you write it.
This is why “did a human write it” is the wrong question. Nobody will ever prove which editor you used. Everyone can check whether what you wrote is true.
A safe way to use AI on an application
- Start from your own facts. Write the real line first, then ask for a clearer version of it. Don't ask for a resume to be written from the job post.
- Keep every number yours. If you can't say where a figure came from, take it out.
- Read it out loud. If it doesn't sound like you, it won't match you in the interview.
- Answer screening questions yourself. Work authorization, notice period, salary, on-site days: a wrong default answer rules you out or puts something untrue on the record.
- Keep a copy of what you sent. You'll be asked about it weeks later.
Our wider checklist is on how to auto apply to jobs safely, and you can check your own wording against a job post with our free resume keyword checker.
How Baela does it differently
Baela is built by recruiters who hired at OpenAI and Google. It uses AI on your application, and it is built so that use stays honest:
- Your real experience only. It tailors your resume and answers from what you have actually done. It never adds jobs, skills, numbers, titles or results you don't have.
- A second check on every rewrite. Any rewrite that claims more than your original line is thrown out, so tailoring can't drift into exaggeration.
- You see exactly what was sent. Every application shows the exact resume, cover letter and answers, plus the employer's confirmation.
- No tricks on the employer's software. No hidden text, no stuffed keywords, no instructions written for an AI screener.
- Fit first, not volume. It applies only to jobs that fit your titles, location, level and salary, never applies to the same job twice, and paces itself.
- It applies on employers' own career sites, including ones built on Greenhouse, Lever, Ashby and Workday, and never signs in to your LinkedIn account.
- Seen by recruiters, too. On top of applying, Baela shares your profile directly with recruiters at 500+ companies.
It never applies to your current employer, and you can pause it or skip any job anytime.
Sources
All checked in November 2026. Pages and figures change, so check the live pages too.
- OpenAI withdrew its own AI Text Classifier “due to its low rate of accuracy” (The Decoder, July 25, 2023; OpenAI’s own post blocks automated checks)
- Liang, Yuksekgonul, Mao, Wu and Zou, “GPT detectors are biased against non-native English writers” (arXiv, April 2023)
- Turnitin: Understanding the false positive rate for sentences of our AI writing detection capability (June 14, 2023)
- Vanderbilt University: Guidance on AI detection, and why we’re disabling Turnitin’s AI detector (August 16, 2023)
- Greenhouse: 2025 AI in Hiring Report press release (November 19, 2025)
- Greenhouse for job seekers: What really happens after you apply for a job?
- Greenhouse Support: Application rules overview (Auto-Reject)
- Greenhouse Support: Application review stage (Application History)
- 18 U.S. Code § 1001, false statements (Cornell Legal Information Institute)