On this page
- Why most resume prompts give you generic output
- Set up once: a reusable context block
- Six resume prompts for ChatGPT
- The order to run them in
- Worked example: tailoring in practice
- Worked example: one bullet, before and after
- Fixing the problems you'll actually hit
- Tips that change the results
- Keep a version for each kind of job
- Next step
The best resume prompts for ChatGPT do three things: they give the model your real experience, they paste in the actual job description, and they forbid it from adding anything you didn't say. Below are six templates that follow that pattern, the order to run them in, and a worked before-and-after so you can see what changes in the output.
Every template uses [placeholders] in square brackets. Replace them, paste, and send.
Why most resume prompts give you generic output
The typical prompt is "Rewrite my resume for a marketing job." ChatGPT has no idea which marketing job, which parts of your history matter, or what it's allowed to change. So it fills the gap with confident filler: "results-driven professional with a proven track record." It will also happily add a metric like "increased conversions by 35%" that you never told it.
Two fixes cover most of the problem:
Give it source material and a boundary. Your current resume or raw notes go in, and an explicit rule says nothing outside them may appear in the output.
Give it a target. The full job description, not a job title.
The rest is asking for the right kind of output: a rewrite, a gap list, a question list. They are different jobs, and one prompt shouldn't try to do all of them. Most of the weak results people complain about come from a single giant prompt that asks for a rewrite, a summary, a keyword check, and a cover letter at once.
Set up once: a reusable context block
You'll paste the same background into most of these prompts, so write it once and keep it. Here is the block we'd use.
BACKGROUND ABOUT ME Name: [your name] Current or most recent role: [title, company, dates] Target role: [job title] Seniority: [e.g. 6 years in the field, first management role] Facts you may use: everything in the resume text I paste below, plus anything I state in this chat. Facts you may NOT use: anything else. If you need a number, tool name, or outcome that isn't provided, ask me instead of guessing.
That last line is the most valuable one. It turns a model that invents into a model that asks.
Save this block somewhere you can reach in two clicks. If you keep your prompts in PromptPal, it sits happily in your library, and you paste it at the top of any template below. A notes app works too. What matters is that you stop retyping it.
Six resume prompts for ChatGPT
1. Tailor your resume to a job description
This is the one most people want. It doesn't rewrite everything. It reorders and rewords what's already true so it matches what the employer asked for.
You are helping me tailor my resume to one specific job. Use only the facts in my resume below. Do not add skills, tools, numbers, or outcomes that aren't there. JOB DESCRIPTION [paste full job description] MY CURRENT RESUME [paste resume text] Do this in order: 1. List the 6-8 requirements in the job description that matter most, ranked by how prominent they are in the posting. 2. For each, tell me whether my resume shows it clearly, shows it weakly, or doesn't show it. Quote the line you're basing that on. 3. Rewrite my experience bullets so the strongest matches come first and use the job description's wording where it's accurate. Keep each bullet to one or two lines. 4. List any requirement I don't cover, so I can decide whether I have relevant experience I left out. Do not rewrite my summary or skills section yet.
Step 4 is what makes this useful. Instead of papering over a gap, you get told about it, and you often remember something relevant you'd cut.
2. Rewrite one weak bullet
Work bullet by bullet when a specific line is bothering you. Asking for three versions gives you options without letting the model drift.
Rewrite this resume bullet three ways. Each version must stay true to the facts I give you. Original bullet: [paste bullet] What actually happened, in plain words: [2-3 sentences, include any real numbers] Target role: [job title] Version A: lead with the action. Version B: lead with the result. Version C: lead with the scope (team size, budget, volume), only if I gave you one. Under 25 words each. If a version needs a number I didn't provide, write [NUMBER?] there instead of making one up.
The [NUMBER?] instruction is a small trick that works well. You get a visible marker where evidence is missing instead of a believable fake.
3. Turn rough notes into bullets
Useful when you're writing from scratch or updating after a new role.
Turn my notes into resume bullets for the role below. Use only what's in the notes. Role: [title, company, dates] Target job: [job title] Notes (messy is fine): [paste notes] Write 4-6 bullets. Start each with a past-tense verb (present tense for my current role). Each bullet should say what I did and what changed because of it. If the notes don't say what changed, list those bullets separately under "Needs an outcome" with one question for me to answer.
The "Needs an outcome" list is the useful part. It turns vague duties into a short interview with yourself, and the answers are usually better than anything the model would have guessed.
4. Write the summary last
Write the summary after the bullets are settled, so it reflects what's actually on the page.
Write a 3-sentence resume summary for me. Base it only on the resume text below. Target role: [job title] Resume text: [paste final bullets and skills] Sentence 1: who I am professionally, with years of experience if stated. Sentence 2: the two strongest things I'd bring to this role, taken from the bullets. Sentence 3: the kind of problem I'm best at solving. Ban these words: passionate, dynamic, results-driven, proven track record, team player.
The banned-words line is blunt, but it works. Without it, you'll get at least two of those in the first draft.
5. Audit for gaps and weak spots
This one doesn't write anything. It reads like a skeptical recruiter.
Act as a recruiter screening resumes for [job title] at a company like [company type]. Read my resume and the job description. Don't rewrite anything. JOB DESCRIPTION [paste] RESUME [paste] Tell me: 1. The three things that would make you hesitate to interview me. 2. Any bullet that sounds like a duty rather than an achievement. 3. Any unexplained gap or sudden change in direction. 4. Five questions you'd ask me in a screening call.
The screening questions are a bonus: they double as interview prep. If you can't answer one in two sentences, the matching bullet probably needs more detail.
6. Check the final version for invented content
Run this last, after you've accepted edits from any of the other prompts. Models drift over a long chat, and a line you approved at step three may have changed by step six.
Compare the two texts below. TEXT A is my original resume. TEXT B is the revised version. List every claim in TEXT B that is not supported by TEXT A: new numbers, new tools, new job titles, new responsibilities, changed dates. Quote each one. If everything is supported, say so. TEXT A: [paste original] TEXT B: [paste revised]
This takes a minute and catches the thing that actually costs people interviews: a claim they can't back up when someone asks about it.
The order to run them in
The templates work alone, but they work better as a sequence for each application. Here is the one we'd follow.
Open a fresh chat and paste the context block.
Run template 1 (tailor) with the job description and your master resume. Read the gap list before you look at the rewritten bullets.
For any gap where you do have experience, jot a few messy notes and run template 3 to turn them into bullets.
Run template 2 on the two or three bullets that still read like job duties.
Run template 4 for the summary, now that the bullets are final.
Run template 5 as a recruiter pass. Fix what it flags by hand or with another round of template 2.
Run template 6 against your original resume before you export anything.
Keep one master resume that holds everything you've done, and tailor copies from it. Tailoring a tailored resume stacks small changes on top of each other, and by the third round you're hard-pressed to say which version is true.
Worked example: tailoring in practice
Say the posting is for a lifecycle email marketer, and it asks for HubSpot experience, A/B testing, and segmentation. Your resume has one relevant line: "Ran email campaigns for the product team."
Template 1 gives you something like this for steps 2 and 4, trimmed:
HubSpot: Doesn't show it. No platform is named anywhere.
Segmentation: Shows it weakly. Based on "Ran email campaigns for the product team", no audience split is mentioned.
A/B testing: Not covered.
That's a useful answer, because it points at exactly what's missing. In this case, you remember that you used HubSpot for two years and tested subject lines on every send. You add those facts in a sentence of notes, run template 3, and the bullet becomes specific and true. Without the gap list, ChatGPT would more likely have written "Optimised campaigns with A/B testing in HubSpot" on its own and left you to find out in the interview whether you could back it up.
Worked example: one bullet, before and after
Here's template 2 on a real-looking input.
Original bullet: Responsible for customer onboarding emails.
What actually happened: I rewrote the five-email onboarding sequence for our project management tool. I also set up a trigger in the email platform so people who hadn't logged in after three days got a different message. I don't have exact numbers, but support tickets about setup dropped.
Output, trimmed:
A: Rewrote a five-email onboarding sequence and added a three-day inactivity trigger for new users.
B: Reduced setup-related support tickets by rewriting the onboarding sequence and adding an inactivity trigger. [NUMBER?]
C: Rebuilt a five-email onboarding flow for a project management product, with a separate path for inactive users.
The original said nothing about what was done. The rewrites name the work, and version B marks exactly where a real figure belongs. If you can pull the ticket count from your support tool, drop it in. If you can't, use A or C and skip the claim.
Fixing the problems you'll actually hit
The model keeps adding numbers. Repeat the rule at the end of the prompt, not only the start, and add the [NUMBER?] marker from template 2. If a number still slips in, paste the output into template 6.
Every bullet sounds the same. ChatGPT tends to open every line with "Spearheaded" or "Leveraged". Tell it which verbs to avoid, or ask for plain verbs like built, wrote, ran, fixed, trained. Plain verbs read better to a human reviewer anyway.
The tailored resume stuffs in keywords. If the job description's phrases show up three times in one section, add this line: "Use each phrase from the job description at most once, and only where it accurately describes what I did."
Text pasted from a PDF comes out broken. Columns and bullets from a designed resume often paste as a jumble. Copy the text into a plain document first, fix the line breaks, and paste that. The model will read clean text far more reliably than a mangled layout.
The answer is too long. Set limits in the prompt: number of bullets, words per bullet, lines per role. Models follow a concrete limit better than "be concise".
Tips that change the results
Paste the whole job description. Not a summary. The wording in the posting is what recruiters scan for, and the model can only match it if it sees it.
One job per prompt. Tailoring, rewriting, and summarising in a single message gives you a mushy answer on all three.
Start a new chat per application. Earlier jobs bleed into later ones, and you'll find last week's company name in this week's cover letter.
Read the output against your own resume. Always. ChatGPT is fast, not accountable. You're the one in the interview.
Expect the output to differ by model and version. These prompts are plain enough to work in ChatGPT, Claude, and Gemini, but the tone of the bullets will vary. We covered the Claude version in Resume Prompts for Claude.
If you want to understand why some prompts hold up, and others fall apart, How to Write Better ChatGPT Prompts covers the general principles behind everything above.
Keep a version for each kind of job
After a few applications, you'll tweak these prompts: a tighter bullet length, a banned-words list that fits your field, a different tone for creative roles. Save the version that worked rather than rewriting from memory each time. PromptPal's prompt version control lets you keep earlier versions so you can go back when an edit makes the output worse.
If a prompt gives you a bland answer and you can't see why, the PromptPal Chrome extension can refine it inline before you send it to ChatGPT.
Next step
Browse the public prompt library for more role-based templates, then save your tailored resume prompts to your own library so the next application starts from a prompt that already works.
Stay Updated
Get the latest blog posts and AI prompt engineering tips delivered to your inbox.




