How to humanize an AI-written CV

Before and after

Six adjectives became two numbers, a budget, a timeframe and a decision. The second version is shorter in claim and longer in fact, and it survives the obvious follow-up question in an interview.

The failure mode is vagueness, not detection

Most advice about AI and CVs starts with whether a recruiter will run yours through a detector. That is the wrong worry. There is no reliable evidence that detection is routine in hiring, a CV is mostly fragments rather than prose, and detectors are least reliable on exactly that kind of text.

What is real is that a generated CV reads as generic, and a recruiter comparing two hundred of them notices it in seconds without any software. The pattern is the same every time: a summary paragraph made of qualities, bullets that describe responsibilities rather than results, and vocabulary that would fit any candidate in any industry.

So the fix is not to sound less like a machine. It is to say something only you could say, which happens to be the same edit.

The bullet formula that survives contact

Verb, object, number, outcome. What you did, to what, at what scale, and what changed as a result. Not every bullet will have all four, and a bullet with none of them is a bullet describing a job description rather than your work in it.

The most common miss is scale. "Managed a budget" is a responsibility; "managed a £40k monthly budget" is a job. Recruiters are calibrating what level you have operated at, and without a number they cannot, so they default downwards.

The second most common miss is the outcome. A great many CVs list what someone was responsible for and never say whether it went well. If the outcome was bad, the honest version is often still stronger — a campaign you shut down because it was not working demonstrates judgement that a vague success claim does not.

The summary paragraph is usually deletable

Generated CVs open with three lines about being a results-driven professional. Read yours and ask what a reader learns from it that the rest of the page does not tell them better. Usually nothing, and the space is worth more given to a bullet with a number in it.

If you keep a summary, make it specific enough that it could not sit on top of anyone else's CV. Two sentences naming what you actually do, at what scale, in what sector. If it would fit a stranger's CV unchanged, it is not doing any work.

The line you must not cross

Rewriting how you describe your work is normal and expected — everyone does it, with or without software. Changing what the work was is a different thing entirely, and it is the failure mode a rewriting tool makes frictionless.

Job titles, employment dates, degree classifications, employer names. None of these are phrasing. A tool asked to make a CV more impressive will happily promote you, and the resulting document is one background check away from a withdrawn offer and a much worse problem than a flat bullet point.

Read the output against the facts, line by line, before it goes anywhere.

Formats with the same problem

Cover letters — the prose half of the same application, where specifics matter just as much.

LinkedIn posts — the public version of the same voice, where the template is the whole problem.

Questions

Do recruiters run CVs through AI detectors?

There is no reliable evidence that it is routine, and a CV is a poor input for detection anyway — mostly fragments, well below the length most tools state as a minimum. The realistic risk is a human recognising generic writing, which needs no software.

Will an applicant tracking system reject an AI-written CV?

Applicant tracking systems parse and match keywords; they are not AI detectors. What actually gets a CV dropped by one is formatting — multi-column layouts, text inside images, unusual fonts — rather than how the sentences were produced.

Should I disclose that I used AI to write my CV?

There is no general obligation, and nobody expects a disclosure statement on a CV. Some employers now ask directly in an application form, and if asked, answer honestly — a false declaration is a much larger problem than the tool use it was hiding.

How many numbers is enough?

Aim for one in most bullets and do not invent any to hit that. A bullet with no number that describes a real decision beats a bullet with a number you cannot substantiate in an interview.

What to watch for

• •Every number on a CV is a claim you will be asked about. Check them before you send it.
• •Do not rewrite job titles, dates or employer names into something more impressive — that is not phrasing, it is misrepresentation.
• •Keep the file plain. Multi-column layouts and text inside images defeat the parsers before any human reads it.

Other use cases