- Why the first prompt everyone tries doesn't work
- Start with a data dump, not a writing request
- The seven prompts I use in order
- What Denise's CV looked like before and after
- The bit that's uncomfortable to say out loud
- Tailoring the same CV for different jobs, fast
- Fixing the summary line specifically
- Making the career gap or career change sound normal
- Don't forget the LinkedIn match
- The cover letter and follow-up, briefly
- A quick word on format
- Frequently asked questions
The short version: ChatGPT can build you a solid CV from nothing, but only if you feed it your actual numbers, titles, and dates before you ask it to write a single sentence, because if you skip that step it will invent confident-sounding fluff instead. Below are the seven prompts I use in order, a real example from a workshop I ran two months ago, and the uncomfortable bit nobody selling "AI CV templates" wants to say out loud: most AI-written CVs now sound identical, and recruiters have started noticing.
Why the first prompt everyone tries doesn't work
Most people open ChatGPT and type something like "write me a professional CV for a marketing manager role." It obliges. What comes back reads well, uses the right buzzwords, and is completely useless, because ChatGPT doesn't know a single fact about you. It fills the gaps with the safest, most generic phrasing it has seen a million times: "proven track record of driving results," "dynamic team player," "results-oriented professional." Every other candidate applying to the same job typed the same lazy prompt and got the same three phrases back.
I have read enough CVs in the past year, my own included, to say this plainly: recruiters can now spot an AI-written CV at a glance, not because AI writing is bad but because unedited AI writing is samey. The fix isn't avoiding ChatGPT. It's refusing to let it guess.
Start with a data dump, not a writing request
Before you ask ChatGPT to write anything, get your raw material into the chat. This is the step almost every "best CV prompts" list skips.
Prompt 1, for gathering your facts:
- "I'm going to give you messy notes about my work history. Don't write anything yet. Just read it and ask me follow-up questions to fill in gaps, especially anywhere I've used a vague word like 'helped' or 'managed' without a number attached."
Then paste in everything: job titles, dates, a rough sense of what you did, any numbers you remember (team size, budget, percentage growth, number of clients, hours saved). Even half-remembered figures are worth including, ChatGPT can help you phrase them honestly ("around 200 customers" rather than a fake precise figure).
The seven prompts I use in order
This is the sequence I gave a group of women at a job search workshop in Reading in late 2025, most of them returning to work after redundancy. It works whether you're starting completely from scratch or rebuilding an old CV that hasn't been touched since 2019.
- Facts first. "Here are my notes on my last three jobs. Ask me questions until you have specific numbers, tools, and outcomes for each role, not just duties."
- The professional summary. "Using only the facts I've given you, write a three-line CV summary. No adjectives like 'passionate' or 'dynamic.' Every line must contain either a number, a job title, or a named skill."
- Bullet points per role. "For my role as [job title] at [company], turn these notes into four bullet points using the format: action verb, what I did, the measurable result. If I haven't given you a number for a bullet, ask me for one instead of inventing one."
- Skills section. "Read the job advert I'm about to paste in and compare it against my CV draft. List the skills the advert mentions that I already have evidence for in my notes, and list the ones I'm missing."
- ATS keyword pass. "Go through my CV draft and the job advert side by side. Which exact phrases from the advert are missing from my CV that I could honestly include?"
- Tightening. "Cut this CV down so no bullet point is longer than two lines and no section header takes up more than four lines total."
- The gut check. "Read this CV back as if you were a recruiter who has 200 CVs to get through today and 7 seconds each. What would make you skip past this one?"
That last prompt is the one people skip and it's the most useful. A widely cited applicant tracking system reality is that most CVs get filtered by software before a human ever reads them, so writing for keyword match and writing for a fast human skim are two different jobs, and you need ChatGPT to check both.
What Denise's CV looked like before and after
At that Reading workshop, one attendee, Denise, had already asked ChatGPT for a CV three separate times before she came to me. Each version was fluent, well-formatted, and interchangeable with a stock template. Her summary line read: "Dynamic and results-driven administrator with a passion for operational excellence." It said nothing. It could have been written for anyone, because it had been, functionally, by an AI that had nothing specific to work with.
We ran the facts-first prompt. It turned out Denise had run a team of six through a systems migration that cut invoice processing time from four days to one, and had personally trained eleven new starters over eighteen months. None of that was in her CV because she'd never thought to mention it, it felt like "just doing my job." Once ChatGPT had those numbers, the rewritten summary read: "Operations administrator who led a six-person team through a systems migration, cutting invoice processing time from four days to one, and trained eleven new starters over 18 months." Same person, completely different CV, because the input changed, not the tool.
The bit that's uncomfortable to say out loud
Here's the part most posts on this topic won't tell you: a CV isn't primarily written for the hiring manager. It's written to survive a keyword scan first and get 7 seconds of human attention second. That's not cynical, it's just how hiring at scale works now, and pretending your CV is a personal essay about your career journey wastes time you don't have. Use ChatGPT to match the exact language in the job advert, not because it's dishonest but because if the advert says "stakeholder management" and your CV says "worked with people across departments," the software may never let a human see the difference.
The other uncomfortable truth: you cannot outsource the substance. ChatGPT can structure, tighten, and phrase, but if you don't know your own numbers, no prompt fixes that. Spend twenty minutes before you open ChatGPT writing down every metric you can half-remember from the last five years. That twenty minutes matters more than any prompt on this page.
Tailoring the same CV for different jobs, fast
Once you have one strong CV, don't rewrite it from zero for every application. Use this prompt instead:
- "Here is my master CV and here is a new job advert. Adjust the summary and reorder the bullet points so the most relevant achievements to this specific advert appear first, without changing any facts."
This takes about ten minutes per application once your master CV is solid, versus the hour or two people spend starting from scratch each time. If you're job hunting while also trying to keep a small business or side project running, the same discipline applies to your week generally, and I've written before about the ChatGPT prompts I use to run my actual week that work well alongside a job search routine.
Fixing the summary line specifically
The top three lines of a CV get read even when nothing else does. If you only fix one section, fix this one. Try:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
- "Write three versions of my CV summary at different lengths, one line, two lines, and three lines. Each must lead with my most impressive number, not my job title."
Leading with a number over a title matters more than people think. "Led a team of six through a systems migration that cut processing time by 75%" gets read differently to a recruiter than "Experienced Operations Administrator," even though the second one is what most CVs still open with.
Making the career gap or career change sound normal
If you're changing industries, returning after time out, or moving from employment into running your own thing, ChatGPT is useful here, but only if you tell it the truth rather than asking it to disguise anything. Try:
- "I had an 18-month gap between these two roles where I was [caring for a parent / freelancing / made redundant and retraining]. Suggest one honest sentence I could add to my CV that explains this without apologising for it."
If part of that gap involved starting something of your own, even informally, that's worth including rather than hiding. I've covered the practical side of that shift in ChatGPT prompts for small business owners, several of which are useful if your CV needs to explain freelance or self-employed work in a way that sounds like a deliberate choice rather than a gap you're covering up.
Don't forget the LinkedIn match
Whoever reads your CV will check your LinkedIn within the same five minutes. If the two don't match, it reads as sloppy or, worse, exaggerated in one direction or the other. Ask ChatGPT:
- "Compare my new CV summary with this LinkedIn headline and about section. Tell me anywhere they contradict each other or use very different numbers for the same achievement."
This is the same instinct behind keeping your social presence consistent generally, which is something I bang on about constantly with clients, and the same logic that sits behind the ChatGPT prompts I use with social media manager clients for keeping messaging consistent across a whole profile rather than one post at a time.
The cover letter and follow-up, briefly
A CV rarely goes out alone. If you're also drafting a follow-up email after an interview or a speculative note to a hiring manager, the tone matters as much as the content, and the same "don't let it invent, feed it facts" rule applies. If you want more on getting that kind of writing to sound like a person rather than a template, the same thinking runs through ChatGPT prompts for email writing that don't sound like every other message in someone's inbox, and it transfers directly to job search emails.
A quick word on format
ChatGPT will happily generate a two-column, graphics-heavy CV if you ask, and most applicant tracking systems will mangle it. Keep it single column, standard headings (Experience, Skills, Education), no tables, no text boxes, and save it as both a.docx and a PDF unless the job advert says otherwise. Ask ChatGPT: "Format this CV using only plain headings and bullet points, nothing that would break in an older CV scanning system." It knows this and will comply, it just won't offer unless you ask.
Try my free ChatGPT prompt generator before you post.
This guide is part of my ChatGPT Prompts and Limits: 30 Guides to Fun Prompts, Photo Edits and Daily Caps.
Frequently asked questions
Can ChatGPT write a whole CV for me from just a job title?
It will, but it will invent generic-sounding achievements to fill the gaps, which is exactly the problem. Give it your real notes, numbers, and job history first, and only then ask it to write, or the CV will read like it belongs to nobody in particular.
Will recruiters know my CV was written with ChatGPT?
Not from the writing quality itself, but they will notice if it sounds identical to every other CV they've read that week, full of phrases like "results-driven" with no specific evidence attached. The fix is feeding ChatGPT your real numbers, not avoiding the tool altogether.
How many jobs should I include on a CV built from scratch?
Two pages maximum for most people, which usually means your last 10 to 15 years of relevant work in full detail and older roles summarised in one line each. Ask ChatGPT to condense anything older than a decade unless it's directly relevant to the job you want.
Should I use the same CV for every job application?
Build one strong master CV, then use a short prompt to reorder and reweight it for each specific advert rather than rewriting from nothing each time. It takes about ten minutes per application once the master version is solid, and keeps your facts consistent across every version you send out.