Resume prompts with the guardrails already written in
Six copy-paste prompts for ChatGPT, Claude, or any chatbot — for rewriting bullets, recovering your missing metrics, and tailoring to a posting. Each one bakes in the rule that matters most: the model works with your facts and never invents new ones.
Free · No sign-up · No credit card · Works with any chatbot
The prompts
1. Rephrase your bullet points
Use it when: Your bullets describe duties instead of outcomes, or all start the same way.
Rephrase my resume bullet points. Rules, and the first one is absolute:
1. Never invent anything — no numbers, percentages, tools, skills, titles, or outcomes that are not in my original bullet. If a bullet would be stronger with a metric it doesn't have, ask me for the real number instead of estimating one.
2. Keep each rewrite under 30 words. Lead with what changed because of my work, not with the activity.
3. Do not open a bullet with any of these: was responsible for, duties included, tasked with, participated in, involved in, helped with.
4. Do not use filler self-descriptions like: results-driven, detail-oriented, team player, hard worker, hard-working, thought leader.
5. Keep my voice. If my original is plain, make it clear — not fancy.
My bullets:
[paste your bullets here]
Rule 1 is the whole game. Left to itself, a chatbot will improve your bullet by inventing a percentage — and a number you can't back up in an interview is worse than no number.
2. Get interviewed for your missing numbers
Use it when: You know your bullets need metrics but can't remember the numbers offhand.
Read my resume bullets below, then interview me to recover the numbers they're missing. Ask me one question at a time — specific and answerable from memory: how many, how often, how big, over what period, compared to what. Up to ten questions total.
When I've answered, rewrite each bullet using only the numbers I actually gave you. If I couldn't give you a number for a bullet, leave it unquantified — an honest plain bullet beats an estimated impressive one.
My bullets:
[paste your bullets here]
Asking a model to "add metrics" produces fiction. Asking it to ask *you* produces your own numbers — the interview format is the only honest route to quantification.
3. Tailor your resume to one job posting
Use it when: You're applying to a specific role and want your existing experience to answer it.
I'm tailoring my resume to the job posting below. Help me, under these rules:
1. Work only with experience I already have. Never suggest claiming a skill the posting wants but my resume doesn't demonstrate. If there's a real gap, name it plainly so I know it exists — don't paper over it.
2. Find every place my resume already demonstrates a requirement but in different words than the posting uses, and suggest using the posting's wording — recruiters search for the posting's words.
3. Tell me which bullets to move up, trim, or cut for this role, and why.
4. Don't rewrite my whole resume. Give me the shortlist of edits with the highest payoff first.
My resume:
[paste your resume text here]
The job posting:
[paste the posting here]
The strongest pull in tailoring is to quietly add the missing keyword. Rule 1 blocks it; rule 2 does the legitimate version of the same thing — surfacing matches you already earned.
4. Make an AI-drafted resume sound like you
Use it when: You drafted with a chatbot and it reads like everyone else's chatbot draft.
The resume text below was partly written with AI, and it reads that way. Rewrite it to sound like a person:
- Cut buzzword self-descriptions (results-driven, detail-oriented, team player, hard worker, hard-working, thought leader) and any sentence that could appear unchanged on a stranger's resume.
- Vary how bullets and sentences open.
- Prefer concrete nouns and plain verbs over abstractions ("built the rota system", not "spearheaded scheduling optimization initiatives").
- Do not add a single new claim, skill, or number. This is a cut-and-rephrase pass only.
My text:
[paste your resume text here]
Hiring managers' complaint about AI resumes isn't that AI touched them — it's unedited, interchangeable slop. This pass removes the tells while leaving your facts alone.
5. Write a summary from facts, not adjectives
Use it when: Your current summary says what kind of worker you are instead of what you've done.
Write me a resume summary of two sentences, three at most, using ONLY facts that appear in the resume text below: my actual role, years, domains, and one or two concrete results.
Banned: any adjective about my character or work ethic (results-driven, detail-oriented, team player, hard worker, hard-working, thought leader, "passionate", "proven track record"). If it can't be verified from the text below, it doesn't go in the summary.
My resume:
[paste your resume text here]
A summary is the one section recruiters skim first, and the one most tools fill with unverifiable praise. Restricting it to facts from the page below it makes it the rare summary that survives a skeptical read.
6. Explain an employment gap without spin
Use it when: There's a gap in your dates and you keep rewriting the same awkward line.
I have an employment gap of [length] between [previous role] and [what came after, or "now"], because [the real reason — caregiving, layoff, health, education, travel, job search].
Give me one plain sentence I could put on the resume or say in a screening call, plus two variants in a different register. Rules: never suggest stretching dates or reformatting the resume to hide the gap — recruiters notice, and getting caught costs more than the gap does. If anything I did during the gap (courses, freelance, volunteering) is worth listing as its own entry, say so.
Most gap advice is concealment advice. A stated reason in one calm sentence outperforms a resume engineered to hide the dates — and can't blow up in a background check.
Is it cheating to use these?
It is the question that stops people from using tools they already have open, and the evidence on it is unusually one-sided — just not in the direction the worry expects.
The people reading it already said it is fine
90% of hiring managers say generative AI in application materials is acceptable (Canva, 2024, n=10,200). Not grudgingly tolerated — acceptable. A separate 2025 survey of 600 hiring managers found only 14.5% opposed to AI at any stage of writing (TopResume).
Their side of the table uses more of it than yours
87% of employers now use AI somewhere in recruiting, and their use of AI resume screening went from 35% to 58% in a single year (Resume Genius, 2026, n=1,500). It is an odd rule that would let the software read ten thousand resumes and not let you edit one.
Nothing in the pipeline is scanning for it
Independent audits through 2025 and 2026 found no major applicant tracking system running AI detection at all. General-purpose detectors would not help them if they did: all 14 tools in one 2023 evaluation scored under 80% accuracy (Weber-Wulff), and Stanford researchers found a 61.3% false-positive rate on writing by non-native English speakers that same year. Resume bullets are fragments — the worst possible input for any of them.
The strongest study says assistance helps
A field experiment with roughly 480,000 job seekers, published in Management Science in 2025, found algorithmic writing assistance raised hires by about 8% and wages by 8.4%, with no drop in employer satisfaction afterwards. The assistance was structured — repair the writing, never invent the content — which is the same line every prompt on this page draws.
The line worth respecting is fabrication, not authorship: a number you did not earn or a skill you cannot demonstrate is a problem whoever typed it. Everything a recruiter dislikes about AI resumes is a symptom of unedited default output — interchangeable phrasing, no specifics, nobody in particular on the page. That is a description of generic writing, and it is exactly what the rules inside these prompts are built to prevent. The detection question, with the evidence.
What a chatbot gets wrong without these rules
The prompts read strict because the default behavior is the problem. Three failure modes account for nearly all of it.
It invents your metrics
Ask for punchier bullets and you'll get percentages nobody measured. Every prompt here forbids new numbers and routes the model to ask you instead — the same rule our own paid rewrites enforce in code.
It adds skills you'd fail an interview on
Tailoring prompts drift toward claiming whatever the posting asks for. The tailoring prompt above only lets the model surface experience you already have, worded the way the posting words it.
It converges on everyone's resume
Chatbots reach for the same openers and self-praise for every user who asks. That shared vocabulary is exactly what readers have learned to discount — the de-AI prompt strips it and keeps your phrasing.
The banned-word lists inside the prompts aren’t our opinion of the week — they’re the published lists our own scoring matches against, in full on how scoring works.
Curious whether these failure modes are real or just plausible-sounding? We actually ran the naive version — no rules, no guardrails — on five resumes and published the results. See what a chatbot does left to its own defaults.
Now score what the chatbot gave you
A prompt can’t see the finished resume. Paste the result below and get the 0-100 score, category breakdown, and every remaining issue — each quoting the line it came from.
Prompt library — common questions
Do these prompts work with ChatGPT, Claude, and Gemini?⌄
Yes — they're plain instructions with no model-specific syntax, so they work pasted into any current chatbot. The rules inside the prompt matter far more than which model reads them.
Why are you giving these away free?⌄
Because you'll use a chatbot anyway, and we'd rather the output be honest. There's no catch: no sign-up, no credit card, nothing to unlock. When you want to know whether the result actually reads well, you can score it here in about 20 seconds — that part is free too.
Is it cheating to use AI on my resume?⌄
Using AI to rephrase your own real experience is editing, and nobody faults it. It crosses into misrepresentation the moment invented numbers or unearned skills appear — which is exactly what these prompts are engineered to prevent.
Will recruiters reject an AI-written resume?⌄
What gets discounted is generic, unedited output — the same buzzwords and rhythm on thousands of resumes — not AI involvement itself. Keep your facts, your specifics, and your phrasing, and the draft's origin stops being detectable or relevant.
How do I check the result?⌄
Paste the rewritten text into the checker on this page. You'll get a 0-100 score and every issue found, each quoting the line that caused it — the anti-filler word lists the prompts reference are the same ones the scoring runs.