Home / Blog / ChatGPT for job applications: what it does well and badly

ChatGPT for job applications: what it does well and badly

· Updated 2026-09-18 · 8 min read
A chat window beside a CV document, with invented sentences marked

ChatGPT works well for one half of an application and badly for the other. The split is predictable, and knowing where it falls is most of the skill.

Using ChatGPT for a job application works well for one half of the task and badly for the other. The split is predictable, and knowing where it falls is most of the skill.

What it is good at

It is good at arrangement: ordering, shortening, changing register, and turning something you already wrote into something readable. It is bad at knowing what you did, and it will not say so.

TaskWorks?Why
Shortening a bullet you wroteYesPure rearrangement of your own material
Matching your text to a postingYesIt has both documents in front of it
Writing a bullet from a job titleNoIt has no facts, so it supplies plausible ones
Estimating a number you forgotNoIt will produce a confident figure that is fiction

The prompt that works

Give it your material and the posting, and forbid it from adding anything not present in what you supplied. The instruction that changes the output most is the one telling it what it may not do.

What it cannot fix

It writes text, it does not produce a file, and the file is where most applications fail. Paste the finished document into a plain text editor and read what a machine sees: the format that parses.

Checking the output before you send it

Read the result against your source line by line, because the failures are additions rather than errors. A wrong sentence looks exactly like a right one until you check where it came from.

Table of seven application tasks marked good, neutral or bad for a chat model, with the reason for each verdict
The pattern across the whole table: it is reliable when it edits facts you supplied and unreliable the moment it has to supply one. Interview practice is the most underused item on the good side.

It is reliable while it edits facts you supplied and unreliable the moment it has to supply one. Everything it adds is something you now have to check.

The line that runs through every use

Sources

Every source here was fetched and checked before it was linked. Our own parser numbers are reproducible: render the same CV in both layouts and extract it.

Stop sending the same CV

Keywordise writes a fresh, truthful CV for every job and applies for you. 7-day free trial, 27 free applications, no card.

Start free →

Frequently asked questions

Can I use ChatGPT to write my job application?

For arranging, shortening and matching your own material to a posting, yes. For producing content about work you have not described to it, no: it has no facts, so it supplies plausible ones and states them confidently.

What is a good ChatGPT prompt for a CV?

Paste your real bullets with the numbers first, then the posting, then instruct it to use only facts present above, invent nothing, and leave gaps as gaps. Ask it to flag any sentence it could not source from your material.

Will employers know I used ChatGPT?

Detection tools are unreliable, but a human notices a document that does not sound like you, and the interview is where that costs you. Keep your phrasing and your numbers.

Is it safe to paste my CV into ChatGPT?

Treat it as sending the document to a third party, because that is what it is. Remove your address and phone number before pasting if that matters to you; the tailoring works the same without them.

Can ChatGPT make my CV ATS friendly?

It can fix wording and headings. It cannot fix the file, and the file is where most applications fail. Export, paste into a plain text editor and check your name is the first line.

Should I let ChatGPT estimate numbers I forgot?

No. It will produce a confident figure that is fiction, and a borrowed number collapses in the interview and takes your true ones with it. Leave the gap or go and find the real number.

Does it work in German and Dutch?

Yes, and the same split applies: good at arrangement, bad at facts. Check idiom carefully in both, because a model will produce grammatical text that a native speaker can tell was translated.

Is a paid tool better than ChatGPT for this?

For one document, no. A paid tool earns its place on repetition and on producing a file that parses, which is the part a chat window does not do.