Using AI to write your CV: what works and what does not

Large language models are strong at language tasks with clear inputs. On a CV, that makes them useful as an editor, a sparring partner and a checker, rather than as an author. The tasks where they tend to add real value include:

They are also patient. You can ask for five alternative versions of a profile, compare them and keep the best phrase from each, which is a quick way to get unstuck when you are staring at a blank page. In every case, though, the underlying facts have to come from you.

The core limitation is simple: an AI tool knows nothing about your career beyond what you tell it. If you give it vague input, it fills the gaps with plausible-sounding language, and plausible is the problem. A bullet like "spearheaded cross-functional initiatives that drove significant operational efficiencies" sounds impressive and says nothing a recruiter can use. It could describe almost anyone.

Worse, AI tools can invent specifics. Ask one to "make this bullet stronger" and it may add a percentage improvement, a team size or a budget figure that you never mentioned. These fabrications can be subtle, and they are easy to miss when you are reading quickly. A CV that contains an invented number is a liability: it can be exposed in an interview, in a reference check, or by a hiring manager who knows the real scale of the work in your sector.

AI also tends towards a recognisable house style. Certain words and structures recur: inflated verbs, triple lists of abstract nouns, and profiles that open with the same few adjectives. When many applicants use the same tools with similar prompts, their CVs start to converge. That makes it harder, not easier, to stand out.

Finally, AI tools do not know your target market as well as they sound. They may apply one country's conventions to another, recommend including personal details that a market does not expect, or confidently state rules about CV length or format that are not true everywhere. Treat any advice about conventions as a starting point to check, not as fact.

Recruiters are not usually running your CV through detection software and rejecting it for AI use, and detection tools are not reliable enough to justify that anyway. What they notice is the effect of unedited AI output: a CV full of confident claims with no concrete evidence behind them, language that sounds more like marketing copy than a person describing their work, and bullet points that could be swapped between candidates without anyone noticing.

They also notice mismatches. If the CV reads as highly polished and strategic but the candidate cannot explain their achievements in the same terms at interview, the gap is obvious. The same applies to cover letters that sound nothing like the person on the video call. Your CV should be something you can talk through comfortably, line by line, in your own words.

Some employers now state their own position on AI use in applications or assessments, particularly for written tasks and tests. If an employer gives guidance, read it and follow it. Using AI to polish a CV is broadly accepted; using it to complete an assessment designed to test your own ability may breach the rules of that process.

A CV contains a concentrated set of personal information: your name, contact details, employment history and sometimes your address or date of birth. Before pasting it into any AI tool, check how that tool handles your data. Many services let you control whether your conversations are used to improve their models, and business or enterprise accounts often have different terms from free consumer accounts. Read the settings and the privacy information for the specific tool you use.

The safest approach is to remove what the tool does not need. It does not need your phone number, email, home address, date of birth or any identity numbers to improve your bullets. Replace your name with a placeholder if you like. The content that matters for editing, your roles and achievements, works just as well without identifiers.

Take particular care with your employer's information. Client names, unreleased financial results, internal project names and confidential figures may be covered by your employment contract or a confidentiality agreement. Describe them in general terms before sharing them with any external tool, just as you would on the final CV itself.

Remember that you may be sharing more than your own data. Referees' names and contact details, colleagues mentioned in your achievements and client information all belong to other people. Leave them out of anything you paste into an external tool.

Whatever an AI tool produces, you are the one sending it and the one who will be asked about it. Before any AI-edited CV goes out, read it slowly with your original notes beside you and check each of the following:

The most reliable approach reverses the usual order. Instead of asking an AI to write your CV and then editing its output, start by writing down the facts yourself: for each role, what you were responsible for, what you changed, what the results were and any figures you can stand behind. This is the part only you can do, and it is where the value of a CV lives.

Then use the tool on that material with specific instructions: tighten these bullets without adding any new information; suggest a clearer structure for this section; list the requirements in this job description that my CV does not yet address. Constraining the task reduces the chance of invented detail and keeps the output grounded in your experience.

Finally, edit the result as the author, not as a proofreader. Remove any phrase you would not say out loud. Restore specifics the tool smoothed away. Check the whole document against the job you are applying for, and against the fact-checking list above. The finished CV should read as if a clear, capable person wrote it about their own work, because one did.

AI tools are a reasonable choice when your main need is editing and polish. They are a weaker choice when the real problem is strategic: when you are changing careers, entering a new market, stepping up to executive level, or unsure what your strongest story is. Those situations call for judgement about positioning, which comes from understanding your career and your target readers. That is where an experienced human reviewer or writer adds what a language model cannot.