How to Use AI to Write a Resume (Without Sounding Like AI)

Recruiters now read dozens of AI-written resumes a week. They can tell.

The tools got good fast, and almost everyone is using them. That’s the problem. When thousands of applicants feed the same prompts into the same model, the output converges: the same rhythm, the same buzzwords, the same flawless blandness. The Dutch work psychologist Djurre Holtrop put it plainly in a 2026 study: when everyone uses the same tools with similar prompts, you get applications that all look alike, and looking alike is the opposite of standing out.

So the goal isn’t to avoid AI. It’s to use it without producing the generic draft that gets your application skimmed past.

What AI is genuinely good at

Used as an assistant, a model like ChatGPT saves real time. It’s strong at:

• Restructuring messy notes into a clean, logical order.
• Pulling the keywords and required skills out of a job description.
• Turning a vague, passive bullet into a tighter, active one.
• Generating three versions of a sentence so you can pick the best.

None of that is writing your resume for you. It’s editing, sorting, and rephrasing material you already supplied. That distinction is the whole game.

What it can’t do, and where it bites back

AI doesn’t know your facts. It has no idea you cut churn 18% or onboarded 11 hires, so if you let it write freely it invents plausible-sounding filler, or worse, fabricates numbers you can’t defend in an interview. Recruiters have learned to read for exactly that. In a 2025 Jobseeker survey of over 1,000 recruiters, 81.6% had already received AI-written applications, and they flagged them by a lack of personalisation (61%), generic phrasing (57%), and robotic tone (41%). More than half said an unedited AI application made them less likely to move a candidate forward.

There’s also a quieter risk: a resume that reads better than you interview. If the page sounds like a polished stranger, the gap shows the moment you start talking.

The workflow that works

Write the raw material yourself first. Note each role, what you actually did, and the results, with real numbers. Then hand that to the model and ask it to tighten, not invent. A prompt like “rewrite these bullets with active verbs, keep every number, and add nothing I didn’t state” keeps you at the wheel.

Then do the part most people skip: rewrite the output in your own words. Cut the tells. “Results-driven professional with a proven track record of delivering cross-functional impact” is the sound of a machine. Replace it with something only you could say: a specific project, a specific number, a specific problem you fixed. Rule of thumb: if half the draft still isn’t in your voice, you’re not done.

One more thing, and it matters more in some countries than others. Don’t paste private data into a public AI tool. Your home address, date of birth, and especially anything like a national ID number can be stored and resurface elsewhere. Give the model your experience, not your identity.

The read-aloud test

Before you send anything, read the resume out loud. The AI tells surface fast that way: sentences that are all the same length, transitions like “furthermore” and “additionally,” adjectives with no example behind them. Anywhere you’d never actually say it in an interview, rewrite it until you would. A recruiter is running roughly the same test in their head, and they’ve had a lot of practice.

Let the proof do the talking

The best defence against sounding AI-generated is concrete evidence. A model can fake fluent prose, but it can’t fake “cut fulfilment costs 22% across two warehouses.” Lead every bullet with a result a human can verify, and the question of who wrote it stops mattering, because no generic draft reads like that.

AI has made polished resumes cheap and everywhere. That’s exactly why a specific, human one stands out more now than it has in years. Use the tool to save time on the wording, then spend what you saved making sure the content could only have come from you. Build your resume in that order: your facts first, the phrasing second.