The short answer is no, with one exception that matters more than all the rest: if the AI writes something about you that is not true, you have not used a tool badly. You have lied on an application.
That single test does more work than any rule about which software is allowed. It is also the test most of the advice online skips, because arguing about whether AI is "allowed" is easier than asking what you actually put on the page.
Why the question feels harder than it is
Job hunting is the one kind of writing where most people have no idea what good looks like. You write a resume a handful of times in your life, you never see anyone else's drafts, and you almost never find out why you were rejected. So when a tool appears that produces confident, professional-sounding paragraphs in four seconds, it feels like it must be a shortcut past something you were supposed to earn.
But nobody thinks it is cheating to have a friend in the industry read your resume. Nobody thinks it is cheating to use a template, a spellchecker, or the phrasing you picked up from a colleague's LinkedIn profile. Employers have never asked candidates to write unassisted. They ask candidates to be accurate.
Where the line actually sits
The line is not between "wrote it myself" and "had help". It is between describing your experience well and describing experience you do not have.
Everything on the safe side of that line has the same shape: you supply the facts, the tool supplies the phrasing. Everything on the wrong side has the opposite shape: the tool supplies the facts and you hope nobody checks.
Four uses nobody sensible objects to
- Turning a flat duty into a result. You know you cut the monthly close from ten days to six. You may not know that the strongest way to say it leads with the number. That is editing, not invention.
- Finding the words for what you did. Plenty of people undersell themselves because they do not know their work has a name the industry uses. Asking what a role like yours is normally called is research.
- Rehearsing. Having a model interview you about a project, then push back on your vague answers, is the closest most candidates get to practice. The answers you give are yours.
- Catching problems. Formatting a machine cannot parse, three bullets that all start with the same verb, a gap you have not explained. Faster than a human reviewer and available at midnight.
Three that will hurt you
- Letting it invent metrics. Models will happily produce "increased efficiency by 40%" because that is what resumes sound like. If you cannot say where the number came from, take it off.
- Accepting skills you do not have. Ask a model to tailor your resume to a posting and it may quietly absorb the posting's requirements into your history. Read every line against your own memory.
- Sending the voice, not the content. A candidate whose resume reads like a press release and who then speaks like a person in the interview creates a gap the interviewer notices, even if they cannot name it.
What recruiters can and cannot detect
People worry about AI-detection tools. In practice, detectors are unreliable on short professional text, and a resume is short professional text that was already full of standard phrasing before any of this existed. Treat any confident claim about detection rates with suspicion, in both directions.
What recruiters do notice, reliably, is inconsistency. A cover letter written at a different reading level from the resume. A bullet that claims ownership of something the interview reveals you observed. Five candidates who all describe themselves as a "results-driven professional with a proven track record". None of that requires a detector.
The risk is not that you get caught using AI. It is that you get caught being generic, or being wrong about your own history.
The disclosure question
Should you tell them? Almost always, no, and not because you are hiding anything. Employers do not ask how your resume was produced, any more than they ask which word processor you used. Volunteering it invites a conversation about tools instead of about you.
Two exceptions. If an application explicitly asks you not to use AI, respect it: a take-home exercise is often a test of exactly the skill they are asking you to demonstrate unaided. And if you are applying for a role where AI use is part of the job, being specific about how you use it is a point in your favour rather than a confession.
A test you can apply in ten seconds
Read any line the model wrote and ask: could I be interviewed on this for five minutes without getting uncomfortable?
If yes, it is your experience in better words, and you should use it. If you feel a flicker of hope that they will not ask, that line is a liability. Delete it or make it true.
That test scales to everything. It covers the resume, the cover letter, the LinkedIn summary, and the answer you rehearsed the night before. It does not care which model you used or whether the phrasing came from you, and it is the same standard a careful candidate applied long before any of this was possible.
What to do next
Start with the parts of the search where the model has the least opportunity to invent anything: formatting, structure, and practice. Have it check your resume against the posting for missing vocabulary rather than asking it to rewrite your history. Have it interview you rather than describe you. Then go back over the finished document line by line, with the five-minute test in mind, before anything gets sent.