Use AI as a Critic, Not a Ghostwriter
The best use of an AI writing tool on a resume is not drafting. It’s critique: hand it a resume you wrote and ask it what’s vague, what’s repetitive, what a reader would be confused by, and where a posting and your page don’t line up. Critique is the mode where the tool has an actual advantage over you — it hasn’t read your resume forty times — and it’s the mode where it can’t invent your experience, because you’re not asking it to supply anything.
Why critique is the mode that works
A generator has to fill space; a critic only has to notice. Asked to write your experience section, a tool knows your job title and nothing else, so it produces the average of every resume in that field. That average is smooth, confident, and interchangeable. Asked what’s weak about a section you wrote, it has real input to work with and its output is a list of observations you can accept or reject one at a time.
It has no familiarity blindness. You cannot see your own resume clearly after the tenth read — you fill in the context that isn’t on the page. A fresh reader catches that instantly, and a tool is a fresh reader every time.
Rejecting a bad critique costs nothing. If it flags a bullet that’s actually fine, you shrug. If it generates a bullet describing work you never did and you don’t catch it, that fabrication ends up in front of an employer. The asymmetry is the whole argument.
The critique passes worth running
Run these one at a time, on a resume you’ve already drafted. Paste the relevant section and ask for a specific kind of read.
1. The vagueness pass. “Here is my experience section. Point out every bullet that is vague or could apply to almost anyone, and say what specific information is missing from each.” The output is a punch list. You supply the missing specifics from memory — the tool can’t, and shouldn’t try.
2. The duplication pass. “Which of these bullets say substantially the same thing?” Repetition is very hard to see in your own writing and easy to spot from outside. Merging two near-identical bullets usually frees a line and sharpens both.
3. The stranger’s-summary pass. “Reading this once, what do you think this person does, at what level, and what are they good at?” Then compare the answer to what you intended. A mismatch means your resume is communicating something other than what you meant — the single most useful signal you can get, and the same question worth asking a human reader.
4. The confusion pass. “What here would confuse someone outside my company or industry?” This catches internal project names, unexplained acronyms, and org-chart references that are invisible to outsiders.
5. The gap-against-posting pass. “Here’s a job posting and here’s my resume. Where do they not line up?” Treat the answer as a checklist to consider, not instructions to follow. Some gaps are real and you should address them by reordering and re-emphasizing what’s true. Some are gaps because you genuinely don’t have the thing, and the correct response is to leave them alone — not to write the missing qualification onto your page. Read the posting yourself too; how to read a job posting covers what a tool won’t tell you.
6. The cliché pass. “Flag every phrase here that’s a stock resume phrase.” Tools are good at this because stock phrasing is exactly what they’ve seen the most of. Then replace each one with evidence, per resume clichés and what to write instead.
7. The consistency pass. “Check this for inconsistent date formats, tense, bullet punctuation, and capitalization.” Mechanical, tedious, and a genuinely good fit for a machine. Verify what it reports rather than accepting it — see tense, person, and punctuation.
Rephrasing is allowed — with a rule
Asking for a tighter version of a bullet you wrote is legitimate and useful. The rule that keeps it safe: the tool may improve how you say something; it may never decide what’s true.
In practice that means every rephrase gets checked for smuggled content. Feed it “rewrote the onboarding checklist, fewer questions in week one” and you may get back “Redesigned the new-hire onboarding process, reducing first-week support inquiries by 30% and improving ramp time across the team.” Three things were added that you never said: a percentage, a claim about ramp time, and a scope expansion from “checklist” to “process.” Delete all three. Keep the tighter verb if it’s better.
That check is not optional and it’s not paranoia — plausible-sounding additions are the characteristic failure of the tool, and they’re hardest to notice when they sound like something you might have written.
Where a tool has nothing to offer
- Supplying accomplishments. It doesn’t know your job. Anything it produces from a job title alone is invention.
- Supplying numbers. Any figure it generates is fictional. Use only figures you can source, and describe scale honestly when you can’t — how to quantify achievements without inventing numbers.
- Judging what matters at this employer. It has no read on this team, this hiring manager, or this market.
- Deciding what to cut. It will happily trim the specific detail that makes a bullet yours, because specificity is what an averaging process removes first. You make the cuts; what to cut from your resume first gives an order.
A workable session
- Draft the section yourself, ugly and honest. Facts first, phrasing later.
- Run passes 1, 2, and 4. Fix what you agree with, using your own words and your own memory.
- Ask for tighter phrasing on the three bullets you like least. Strip any content you didn’t supply.
- Run pass 3 and check whether the resume reads as the person you meant to present.
- Run pass 7 last, once the content has stopped moving.
- Read the whole thing aloud yourself. A tool doesn’t hear rhythm the way you do — that pass is in proofread your resume like an editor.
Your long source document is the natural place to do all of this, because that’s where the real material lives — see keep a master resume.
The line, stated plainly
Using a tool to make true statements clearer is good practice, and there’s nothing to feel uneasy about. Using one to decide what your statements are is a different activity, and it produces a resume you can’t defend in an interview — which is the room where the resume was only ever trying to get you.
Stay the author. Let the tool be the second reader you don’t have. That’s a real advantage, and it costs you nothing you’d want to keep.