15 GPT-6 Prompts to Tailor Your CV to Any Job (and How to Check the Result)
Quick answer: The best GPT-6 CV prompts do three things. They pull real achievements out of you, map them to a specific job description, and tighten every bullet without inventing anything. Use the 15 prompts below in that order. Then do the step most people skip. ChatGPT can't reliably grade its own writing, so score the new version in a dedicated CV checker like HRLens and confirm it's better than the old one before you send it.
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.
What GPT-6 changes for CV writing, and what it still gets wrong
OpenAI's GPT-6 generation is the strongest writing assistant most job seekers have ever had. GPT-6 Astra is the flagship. GPT-6.1 Sol launched at DevDay on September 29, 2026, and OpenAI says it nearly matches Astra on professional work at a fifth of the API price. In ChatGPT, Sol is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, though not yet in regular chat. Either model can turn a messy work history into clean, confident bullets in seconds.
Three failure modes haven't gone away, though. The model fills gaps with plausible fiction: a rounded-up metric, a promoted title, a tool you never used. It optimizes for sounding impressive rather than for getting parsed, so it will happily suggest formatting that an applicant tracking system scrambles. And it grades its own work generously. Ask "is this better?" and the answer is almost always yes.
The prompts below are built around those weaknesses. Each one tells the model to ask instead of invent, and the workflow ends with an independent score.
Before you prompt: give GPT-6 the right inputs
Garbage in, polished garbage out. Before the first prompt, paste four things into a fresh ChatGPT conversation.
1. Your current CV as plain text. 2. The full job description: the whole text, not a link, because many job sites block automated reading. 3. A raw achievements dump: ten minutes of bullet points about projects, numbers, praise you received and problems you solved, however messy. 4. An objective diagnosis of your current CV.
That fourth input is the one most people don't have. Run your CV through a free HRLens CV analysis first. You'll get a score out of 100 and five category scores (experience, tech stack, impact and ownership, clarity and structure, and ATS compatibility). If you paste in the job description, you'll also see the skills you're missing for that role. The full analysis adds a prioritized fix list and likely rejection reasons. Paste those findings into ChatGPT and you've given GPT-6 a precise brief instead of "make it better".
| Input | Why it matters | Where it comes from |
|---|---|---|
| Current CV (text) | The baseline to improve | Your file |
| Full job description | Defines what 'better' means for this role | Copy the job ad text |
| Raw achievements dump | Real facts, so the model doesn't invent | 10 minutes of brain-dump |
| Objective diagnosis | Tells the model what to fix first | HRLens CV analysis |
Prompts 1–5: mine your real achievements
1. The interview-me prompt: "Act as a senior recruiter for [target role]. Interview me one question at a time to uncover achievements for my CV. Focus on outcomes, numbers, scope and ownership. Ask at least 10 questions, and don't write anything until I say 'done'."
2. Duties to outcomes: "Here's a bullet that describes a duty: [bullet]. Ask me three questions that would turn it into an outcome. What changed, by how much, and for whom?"
3. Find the hidden numbers: "From my achievements dump, list every place a number could exist but doesn't: team size, users, revenue, time saved, volume, error rate. For each one, ask me for the real figure or a defensible range."
4. Ownership check: "For each bullet, tell me whether it reads as 'I did it', 'I led it' or 'I was involved'. Flag the ones where my real role was bigger than the wording suggests."
5. Fix-list implementation: "Here are the fixes from my CV analysis: [paste HRLens fix list and blind spots]. Work through them one at a time. For each fix, show the original text, your proposed change and why, and wait for my approval before moving on."
Prompts 6–10: tailor your CV to the job description
6. Decode the job ad: "Analyze this job description. List the top 8 requirements in order of importance, separate must-haves from nice-to-haves, and quote the exact phrases the employer uses for each."
7. Gap map: "Compare my CV to those 8 requirements. For each one, mark it as clearly shown, implied or missing, and quote the CV line that supports it. Don't rewrite anything yet."
8. Keyword mirroring, honestly: "For requirements that are 'implied', suggest rewrites that use the job ad's exact terminology, but only where my experience genuinely matches. Mark any suggestion that would stretch the truth with ⚠️."
9. Reorder for relevance: "Reorder the bullets under each role so the ones most relevant to this job come first. Don't change their wording."
10. The tailored summary: "Write a 3-line professional summary for this specific role. Include my years of experience, my two strongest matching achievements with numbers, and the job title as the employer writes it. No clichés like 'results-driven' or 'passionate'."
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.
Prompts 11–15: tighten, cut and sanity-check
11. The 2-page cut: "My CV is too long. Cut it to [1 or 2] pages by removing or merging the bullets least relevant to this job description. List what you removed so I can veto it."
12. Bullet tightening: "Rewrite each bullet to 1–2 lines, starting with a strong verb and ending with the result. Keep every number. Don't add new claims."
13. Consistency pass: "Check my CV for inconsistent tenses, date formats, job-title formatting and punctuation. List the fixes, don't apply them."
14. Recruiter skim test: "You're a recruiter spending 7 seconds on this CV. What do you notice first? What job do you think I'm applying for? What would make you stop reading?"
15. Fact audit: "List every claim in this CV that includes a number, a tool or a title. Next to each one, mark 'from my input' or 'added by you'. Remove anything added by you."
- +Pulls achievements out of you with good questions
- +Decodes job descriptions and mirrors their language
- +Tightens bullets and cuts length fast
- +Great at drafts and alternatives
- −Can't tell you how an ATS will parse your file
- −Grades its own rewrites generously
- −Tends to inflate numbers and titles unless constrained
- −Outputs chat text, not an ATS-safe document
Check the result: why ChatGPT shouldn't grade its own CV
You now have a sharper, tailored CV. But is it *actually* better, and will it survive screening? Don't ask the model that wrote it. Language models reliably rate their own rewrites as improvements, and none of them can see how an ATS parser will read your final file.
Upload the new version to HRLens with the same job description and compare the numbers. Did the overall score go up? Did impact and ownership improve after prompts 2 and 3? Did clarity and structure hold up after the cuts in prompts 11 and 12? Is ATS compatibility still high? With the Premium Package or All-Access, HRLens also shows a before-and-after comparison: score change, category changes, and which issues you fixed.
If a category dropped, you know exactly which prompt to revisit. That feedback loop is the difference between "ChatGPT rewrote my CV" and "my CV now gets interviews".
| Prompts | What they change | HRLens category to check |
|---|---|---|
| 1–5 | Real outcomes, numbers, ownership | Impact and ownership; Experience |
| 6–10 | Match to the job description | Missing skills list for that job |
| 11–13 | Length, tightness, consistency | Clarity and structure |
| All | Final file and layout | ATS compatibility + visual analysis |
Turn the draft into an ATS-safe file
The last trap is formatting. People paste great ChatGPT text into a flashy two-column template they downloaded somewhere, and all that work gets scrambled by the first applicant tracking system it hits.
Use the HRLens CV builder instead. Paste your improved text (or your old CV) into the chat, and the builder produces a clean CV you can keep refining in plain language. Choose from six ATS-friendly templates and export to PDF or Word. If you ran an analysis with a job description, you can also open the tailored CV HRLens generated and customize it directly.
Keep one master CV and one tailored version per job family. Run prompts 6–10 and a quick HRLens re-score for every role you really want. It takes about 15 minutes per application, and it's the highest-return 15 minutes in your whole job search.
Frequently asked questions
What's the best ChatGPT prompt to tailor a CV to a job description?
Start with a gap map: ask GPT-6 to list the job's top 8 requirements, then mark each as clearly shown, implied or missing in your CV, quoting the supporting line, without rewriting anything yet. Then rewrite only the 'implied' items using the job ad's exact terms, and only where it's true.
Can GPT-6 check if my CV is ATS-friendly?
No, not reliably. It sees your text, not how a parser reads your file's layout. Use a dedicated checker like HRLens, which scores ATS compatibility as its own category and analyzes your CV's visual layout.
Is GPT-6.1 Sol or GPT-6 Astra better for CV writing?
OpenAI says Sol nearly matches Astra on professional work at a fifth of the API price. For CV writing, your inputs (real achievements, the full job description, a diagnosis of your current CV) matter far more than which of the two models you use.
Will recruiters know my CV was written with ChatGPT?
They notice generic AI patterns, like cliché summaries, vague claims and inflated numbers. Using prompts that force real facts, plus a fact audit (prompt 15), keeps your CV specific and credible.
How do I know my rewritten CV is actually better?
Score both versions with the same job description in HRLens and compare the overall and category scores. Don't ask the model that wrote the new version, because it will almost always say yes.
Is your CV good enough?
Upload your CV and get an instant AI score out of 100, an ATS-compatibility rating and a breakdown across five categories — free.