How to use AI to write your resume.

Recruiters read hundreds of applications, and most of them can spot generic AI writing in about four seconds. "Results-driven professional with a proven track record of leveraging cross-functional synergies" tells a hiring manager nothing at all, and whether a machine or a human typed it is almost beside the point — the sentence is empty either way. That is the real hazard of using AI on a job application: not that you get caught, but that you disappear into a pile of interchangeable paragraphs.

Used properly, though, AI is genuinely good at this job. Think of it as an editor and a translator, never an inventor. You supply the truth of what you actually did; it helps you say it in the employer's language, in the shape a hiring manager expects, far faster than you would manage alone. The facts stay yours. Only the phrasing is up for negotiation.

Start with the raw material

Garbage in, generic out. If you type "write me a resume for a marketing job," the model has nothing to work from except the average of every marketing resume ever written — and that average is exactly what you will get back. Before you ask for a single sentence, paste in three things:

• Your current CV, however rough. Even a messy list of employers, dates and duties gives the model real facts to anchor to.
• The actual job advert, in full. Not your summary of it — the whole thing, including the tedious requirements section, because that is where the employer's own vocabulary is hiding.
• A few sentences on what you really did day to day. Write it the way you would explain it to a friend: "I ran the weekly stock order, trained every new starter, and rebuilt the rota spreadsheet everyone hated." The unpolished version almost always contains the best material.

That last piece is the one people skip, and it is the one that makes the difference. Job titles hide the interesting parts; plain speech reveals them.

Tailor to the job advert

Once the model has both documents, ask it to do the comparison work you would otherwise do by hand:

Here is my CV and a job advert. Map my experience onto each requirement in the advert. Where I clearly meet a requirement, suggest wording that uses the employer's own terms. Where I only partly meet it, or don't meet it at all, say so plainly rather than papering over it.

The instruction to flag gaps honestly matters more than it looks. A model asked only to make you look good will happily blur "some exposure to" into "experienced in." A flagged gap is useful information: you can address it directly in your cover letter, decide the role is worth applying for anyway, or conclude that it isn't a fit.

Mirroring vocabulary is fair game as long as it stays true. If the advert says "stakeholder management" and you have always called it "keeping clients happy," use their phrase — you are describing the same real work in the reader's language. Adopting a term for something you have never done is a different thing entirely.

Write bullet points that land

Almost every strong resume bullet follows one pattern: a specific action verb, what you actually did, and a result you can measure. Most weak ones stop after the first half.

Before: "Responsible for the company's social media accounts." After: "Grew the company Instagram account from 2,000 to 9,000 followers in a year by replacing product posts with a weekly behind-the-scenes video series." Same job, same person — but only one of them gives the reader something to ask about.

AI is very good at reshaping bullets into that pattern, and very willing to invent the numbers if you let it. So don't let it:

Rewrite these bullet points as action + what I did + measurable result. Where you need a number I haven't given you, leave a bracketed question for me instead of estimating one.

You will get back a list peppered with [how many people?] and [over what period?], which is exactly right. Numbers don't have to be revenue, either — team size, volume, frequency, hours saved, error rates and turnaround times all count.

Applicant tracking systems (ATS)

An applicant tracking system is, at heart, a database. It receives your file, parses it into fields — name, employers, dates, skills — and lets a recruiter search and filter the pile. Most systems do not silently reject anyone; a human still decides. But if the parser mangles your document, that human may never see it properly.

What follows from this is unglamorous and effective: use standard section headings such as Experience and Education, keep to a single column, avoid tables and text boxes, don't bury contact details in a header or footer, and never put words inside an image. Include the genuine keywords from the advert in your actual descriptions, because those are the terms a recruiter will search for.

Two honest caveats. Free "ATS score" checkers vary enormously in quality, and plenty of them exist mainly to sell you a rewriting service, so treat a score as a hint rather than a verdict. And the old trick of stuffing keywords in white text at the bottom of the page backfires badly: parsers read invisible text perfectly well, it appears the moment anyone copies your CV into another document, and it reads as deception rather than cleverness.

Cover letters

This is where AI earns its keep, provided you treat the draft as scaffolding rather than a finished letter. Ask it for a structure — why this employer, what you bring, one concrete example, a short close — and let it produce a first pass so you are never staring at a blank page.

Then rewrite the opening yourself. The first two sentences are the only ones guaranteed to be read, and they are where AI phrasing is most obvious. Say something specific about this employer that no other applicant could have written. A generic cover letter is genuinely worse than none at all, because it proves you sent the same thing everywhere. Our cover letter templates are a useful starting structure if you would rather begin from a human-written skeleton.

What to never let AI do

• Invent anything factual. Employers, dates, job titles, degrees, certifications, metrics — none of these are the model's to generate.
• Inflate your seniority. "Supported the migration" is not "led the migration." That edit takes one word and can cost you the offer.
• Write anything you could not talk through confidently. If a bullet is on your resume, you should be able to spend five minutes on it: what the problem was, what you did, what happened next.

Fabrications rarely survive contact with reality. They surface under interview questioning, in reference checks and in background screening, and they do far more damage when discovered than the gap they were meant to cover. Practising the conversation is a much better use of AI — our interview prompts will happily grill you on your own bullet points until you can defend every one of them.

When you are ready to start, our resume prompts give you ready-made prompts for tailoring, rewriting and reviewing, and how to write better AI prompts covers the general technique behind all of them.