Recruiters Can Spot an AI-Written CV. Here's How to Use AI Without Sounding Like One
The problem isn't that you used AI. It's that the output is fluent, confident and completely interchangeable with everyone else's. Here's the line between leverage and liability.
Ask a general-purpose chatbot to write your CV and you'll get something impressive in about eleven seconds. It will be grammatical, confident, well structured — and it will read exactly like the four hundred other CVs on the recruiter's screen that were produced the same way this month.
That's the actual risk, and it isn't detection software. It's sameness. A recruiter doesn't need a tool to tell them a document was AI-written; they can feel it, because the tenth one that day sounds like the first. Here's how to keep the leverage without the smell.
The five tells that give it away
None of these is about grammar. They're all about a specific kind of emptiness — text that describes a role rather than a person.
Fluent but unfalsifiable
"Leveraged cross-functional synergies to drive impactful outcomes." Nothing in that sentence could be checked, disproved, or asked about in an interview. Every line on a strong CV can be.
Suspiciously even rhythm
Every bullet the same length, the same shape, the same verb-object-result cadence. Real careers are lumpy: some roles have one big achievement, others have five small ones.
Adjectives doing the work of facts
"Highly motivated", "results-driven", "passionate", "dynamic". Adjectives are what text produces when it has no data to work with.
Zero specificity of place
No named systems, no named markets, no named constraints. A person who actually did the job mentions the ERP they hated and the region they covered.
The same three metaphors
"Spearheaded", "orchestrated", "championed". Recognisable house style. Once you've seen it twice in a day, you can't unsee it.
The document isn't rejected for being AI-assisted. It's rejected for being indistinguishable — and indistinguishable is a slower, quieter kind of no.
Where AI genuinely helps
This isn't an argument against using it. Used on the right tasks, it removes exactly the work that stops people applying — and it's very good at some things humans are reliably bad at about themselves.
| Task | AI is | Why |
|---|---|---|
| Structuring a messy history into sections | Excellent | Pure formatting work with a known right answer |
| Turning your notes into tight bullets | Excellent | You supply the facts; it supplies the compression |
| Matching your CV against a job description | Excellent | Comparison at speed, across vocabulary differences |
| Catching what you left out | Very good | You're blind to your own obvious achievements; it isn't |
| Translating a CV into another market's conventions | Very good | Conventions are learnable patterns |
| Inventing achievements you didn't mention | Dangerous | It will confabulate plausibly, and you'll be asked about it in the interview |
| Writing your CV from your job title alone | Useless | With no input facts, output is by definition generic |
The garbage-in rule, restated
If you give a model three words about yourself, it can only produce the statistical average of everyone with your job title. The quality of an AI-assisted CV is set almost entirely by how much true, specific material you feed it — not by how good the model is.
This is why our builder asks you questions before it writes anything, and shows every suggested change side by side so you decide what stays. The facts stay yours.
Try the guided builderThe method: draft with AI, own it back
Dump your raw material first
Before you generate anything, write out — badly, in fragments — what you actually did. Systems you used, numbers you remember, problems you fixed, people who depended on you. Ten minutes of ugly notes is worth more than any prompt.
Generate structure, not substance
Let the tool organise, compress and order. Its job is to turn your fragments into a readable page, not to decide what your career was.
Delete every unfalsifiable line
Read each bullet and ask: could a recruiter challenge this in an interview? If not, it's decoration. Cut it or replace it with something checkable.
Re-insert three specifics
A named system. A real number. A constraint you worked under (a merger, a regulator, a two-person team, a language). Three of these across the page changes the whole texture.
Break the rhythm on purpose
Let one bullet be long because the story needs it and another be six words. Uneven is human.
Read it aloud
The fastest detector anyone has. If a sentence is uncomfortable to say in your own voice, you won't be able to defend it in an interview either.
AI first draft
"Spearheaded transformative initiatives that drove operational excellence and delivered impactful results across the organisation."
After you own it back
"Ran the 2025 warehouse consolidation (3 sites → 1, 40 staff redeployed). Delivered 6 weeks late but €480k under budget; kept every SLA through the switchover."
The interview test
For every bullet, ask yourself: "Can I talk about this for two minutes if they ask?" A CV you can't discuss is worse than a modest one you can. The second draft's job is to make sure everything on the page is something you'd be glad to be asked about.
What about "AI detectors"?
Some employers run detection tools; most don't, and the tools themselves are unreliable in both directions — they flag careful human writing and clear obvious machine writing. Optimising against a detector is chasing a moving, inaccurate target.
Optimise for the thing detectors are a bad proxy for: specificity. A CV full of real numbers, named systems and awkward true details doesn't read as machine-written because machines don't have access to those facts. Specificity is the defence, and it's also just a better CV.
Disclosure: do you need to say you used AI?
Unless an employer explicitly asks — and a small number now do, usually for written assessments rather than CVs — no. Using a tool to format and sharpen a document you wrote is no more disclosable than using a spell-checker. Using one to fabricate experience is a different problem entirely, and disclosure doesn't fix it.
The one thing never to automate
The decision about what's true. Every fact on the page — every number, date, title, tool, and language level — has to survive contact with a reference check and a follow-up question. Models generate confident plausible detail when they're short of input, and "the AI added that" is not a recoverable position in an interview.
- Numbers you can't source back to something real
- Tools you've read about but never used
- Inflated titles that don't match what payroll would confirm
- Language levels one notch above the truth (the interview may switch language)
- Dates smoothed over to hide a gap
The short version
- The risk isn't detection, it's sameness — fluent, unfalsifiable, interchangeable text.
- AI is excellent at structure, compression and matching; useless at inventing your career.
- Dump raw material first; the quality of the output tracks the quality of your input.
- Delete anything a recruiter couldn't challenge, then re-insert three hard specifics.
- Read it aloud, and never automate the decision about what's true.
Use AI the way it actually helps
Import your CV or answer a handful of questions, and get a structured, tailored document where every AI suggestion is shown as a tracked change you approve or reject. Your facts, sharpened — not a generated stranger's career.
Build my CV freeFree to build and preview. Pay only when you download.
Frequently asked questions
Can recruiters actually tell if a CV was written by AI?
Often, yes — not through software but through pattern recognition. Generic phrasing, uniform bullet rhythm, adjective-heavy claims and a total absence of named systems or real numbers are the giveaways. The fix isn't hiding the tool; it's adding the specificity a tool can't invent for you.
Will using AI to write my CV get me rejected?
Not by itself. Most employers have no policy against it and no reliable way to detect it. What gets you rejected is the *output* being interchangeable with everyone else's, or containing claims you can't defend in an interview. Use it for structure and sharpening, keep the facts yours.
Should I tell an employer I used AI on my application?
Only if they ask, and some do for written assessments. Using AI to format and tighten a document you wrote is comparable to using a grammar checker. Using it to fabricate experience is a different issue, and disclosing it doesn't make that acceptable.
Are AI detectors used on CVs?
A minority of employers use them, and the results are unreliable in both directions — well-edited human writing gets flagged, and edited machine writing passes. Writing specifically to defeat a detector is wasted effort; writing with real detail happens to defeat it anyway, and produces a better CV.
What's the safest way to use AI on a job application?
Give it real material — your notes, your numbers, the actual job description — and use it to structure, compress, translate and match. Review every change rather than accepting a wholesale rewrite, and check that you could talk for two minutes about any line on the page.