Free resource · Job seekers
Where AI helps a job search, where it quietly ends one, and why the line sits exactly where it does — from thirteen years of reading applications from the other side of the table.
So most people guess — and guess badly in both directions.
Half the candidates I talk to won't touch AI for anything job-related, because they've absorbed a vague sense that it's cheating. They're writing worse applications than they could, out of a caution nobody asked them to have.
The other half use it in ways that end their candidacy in the first four minutes of a conversation, and never find out why.
The people who actually know where the line is are hiring managers and recruiters, and they have no particular reason to publish it. So here it is. This is not a policy and it isn't law — it's what the person on the other side of the table is thinking.
Four uses that are simply fine. Use them.
The test, if you only remember one thing
Does the finished thing still represent you — and could you defend every line of it out loud, unprompted, with nothing in front of you?
That question resolves nearly every case. Help saying what's true passes it. Anything that produces a claim you couldn't stand behind in a room does not. The line isn't about how much AI touched the work. It's about whether the work is still yours.
Not the gatekeeper people imagine. Something quieter, and harder to notice.
Most employers past a certain size run applications through an applicant tracking system. It takes your document and tries to turn it into structured fields — employers, titles, dates, skills — and files the result. A recruiter then searches that database.
The myth is that the system scores you and rejects you. Mostly it doesn't. There is no robot grading your resume out of ten. What actually happens is duller and worse: if the parse goes badly, your record lands in the database incomplete, and when a recruiter searches for the exact thing you spent four years doing, you are not in the results. Nobody rejected you. You simply weren't there to reject.
Parsing breaks on layout, not on language. Two-column designs. Anything living in a header or footer. Text inside tables or graphics. Section headings the system doesn't recognize, so it can't tell where your experience starts. The choices that make a resume look considered to a human eye are very often the ones that make it illegible to the software doing the filing.
One column. Standard headings. Real text, not pictures of text. It is a plainer document than most people want to send, and it is the one that survives the trip.
Worth sitting with, on a page about AI: you have been thinking hard about which tools you're allowed to use. The employer has one too, and it read your resume before any person did.
Less than you fear about the writing. Much more than you'd like about the gap.
Start with what's not true: nobody can reliably identify AI-assisted writing by reading it, and the detection tools that claim to are unreliable enough that thoughtful employers have stopped leaning on them. If your worry is that a well-edited paragraph will be spotted and held against you, put it down.
Here is the real exposure, and it has nothing to do with detection.
Your paper and your conversation have to be the same person. A resume describes a scope; twenty minutes later you're asked what was hardest about it. If the document was built by something that never lived the work, the answer arrives thin — no texture, no constraint, no name of the person who pushed back. Nobody concludes "AI wrote this." They conclude something worse and vaguer: that you oversold, and that they don't quite believe you. Then they move on, and you never learn that's what happened.
That gap is the entire risk. It closes the moment the paper only claims things you actually did.
Increasingly you will be. It is not a trap.
Norms are still forming, and they vary. Some employers now state what they expect; most say nothing and form a private opinion. If a policy is published, follow it — that is the whole answer.
If you're simply asked, the honest version lands well, because it is the answer of someone who thinks about their tools: you used it to prepare and to write more clearly, the experience and judgment are yours, and you can walk through any of it. The candidates who look worst answering this question are the ones performing an abstinence nobody believes.
Two ways forward, depending on how much you want to do yourself.
Most resumes that don't get calls aren't badly written. They describe duties instead of outcomes, and no amount of editing fixes a framing problem. If that sounds like yours, start here.
$24 · instant download
Nine pages on writing bullet points that lead with the result — the five shapes a strong bullet takes — plus an editable template laid out the way the section above describes: one column, standard headings, real text, nothing hiding in a header. Built to parse cleanly. Filled-in example on page two.
Get the guide →$175 flat
I rewrite the resume, rebuild the LinkedIn headline and About section, and give you a one-page breakdown of why you weren't getting calls. From the side of the table that screens them.
Book the Resume Read →This page is educational and reflects hiring practice as of August 2026, not legal advice or any employer's policy. Where an employer publishes its own rules on AI use, those govern.