Resume bullets fail when they describe duties (“responsible for…”) instead of outcomes. These 15 prompts force the action-impact-evidence structure recruiters scan for. Quantified bullets matter: eye-tracking studies show recruiters give a resume roughly a 6-second first pass, and bullets that open with a number draw notably more attention than bullets that open with text.
TL;DR
- Paste a real bullet plus your target role and level; the prompt rewrites it as verb + what + how + measurable result, capped at 30 words.
- Tell the model never to invent numbers — make it leave a
[CONFIRM: ...]placeholder you verify with your manager. - Use Claude Opus 4.7 for honest, narrative rewrites; GPT-5.5 for fast keyword-dense bullet surgery; Gemini 3.1 Pro when you want full keyword names (“Apache Kafka”, not “Kafka”) and live market context.
- Run the ATS check (template 9) separately: tables, multi-column layouts, and graphics scramble parsers that read top-to-bottom.
Which AI to use (as of June 2026)
All three flagship chat models handle resume work on free tiers, but they behave differently:
| Model | Strength for resumes | Best for |
|---|---|---|
| Claude Opus 4.7 (Pro $20/mo, free Sonnet 4.6) | Treats your experience as claims to examine, not facts to polish; preserves narrative logic | Career-change stories, executive summaries, honesty checks |
| GPT-5.5 (ChatGPT Free $0, Plus $20/mo) | Fastest, most keyword-dense output; strong at format control and prompt chaining | Bullet surgery, ATS keyword passes, batch rewrites |
| Gemini 3.1 Pro (Google AI Pro $19.99/mo, free tier) | Web-aware, uses full keyword names and live salary/market context | Tailoring to a specific named company or role |
A practical workflow: draft with Claude for voice, run a GPT-5.5 keyword pass against the JD, then sanity-check with Gemini if you are targeting one company. Free tiers are enough for a single resume; the paid tiers mainly buy higher message limits and longer context for pasting a full resume plus JD.
Who this is for
Engineers, PMs, and designers refreshing a resume for a target role; career switchers; new grads turning internships into full-time pitches; anyone whose resume is stuck on “responsible for”.
When not to use these prompts
Skip them for academic CVs (publications-first), federal or government resumes (which follow their own format), and one-page resumes where you only need to tighten two or three lines. Never use AI to fabricate numbers — every metric has to survive an interview probe.
Prompt anatomy / structure formula
A resume-rewrite prompt should carry six elements:
- Target role + level: SWE L4, senior PM, junior designer — drives verb choice and metric weight.
- Original bullet (or raw notes): the model rewrites better than it invents from nothing.
- Strong-verb + outcome formula: “verb + what + how + measurable result” — the structure recruiters scan.
- ATS keywords from the JD: weave 2-3 in naturally; never paste a keyword list.
- Tone: confident, specific, never bragging, never lying. Tell the model to flag any claim it had to invent.
- Length cap: 30 words per bullet, 6 bullets per role, 2 pages total.
Best for
- SWE / PM / Designer resumes
- Career switchers
- Internship-to-FT transitions
- Senior / staff promotions
- A recruiter-replyable LinkedIn intro line
15 copy-ready prompt templates
Placeholders in brackets like [bullet] are yours to replace before sending.
1. Duty → outcome rewrite
Rewrite this resume bullet from "responsible for" voice to "action + outcome + evidence" voice. Quantify if possible. Preserve facts.
Original: "[bullet]"
Role: [role]
Level: [level]2. JD-keyword alignment
Below is my current bullet and the JD I am applying to. Rewrite the bullet to naturally weave in 2-3 JD keywords without stuffing. Mark which keywords you used.
Bullet: "[bullet]"
JD excerpt: "[jd]"3. Project-to-bullet generator
I worked on this project: [3-sentence description]. Generate 4 resume bullet variants. Format: strong verb + what + how + measurable result. Each under 30 words.4. Internship resume bullet upgrade
I had a [company] internship. Tasks: [list]. Outcome I can claim honestly: [outcome]. Write 3 resume bullets. Don't over-claim; intern-level wording is OK.5. Switch-fields bullet
I'm switching from [prev field] to [target field]. My experience: [bullet]. Rewrite to highlight transferable skills relevant to [target field], without lying.6. Side-project bullet
My side project: [description]. Stack: [stack]. Users: [audience / N]. Write 3 resume-worthy bullets — quantified, not fluffy.7. Open-source contribution bullet
My OSS contribution: [repo, what I did, merge status]. Write a 1-line resume bullet that names the project, the contribution type, and the (real) impact.8. Bullet-pruning pass
Below are 12 of my bullets across one role. Cut to the 6 strongest. For each you cut, name why. For each you kept, mark what makes it strong.
[paste bullets]9. ATS pass check
Below is my resume bullet. Identify any phrasing that may confuse an ATS (special characters, awkward verb tense, abbreviations, anything that breaks top-to-bottom parsing). Suggest fixes.
[paste bullet]10. Leadership-without-title bullet
I didn't have the title, but I led: [what / when / who]. Write a bullet that honestly captures the leadership without claiming a title I didn't hold.11. Metric-hunter pass
Use when bullets read fine but contain zero numbers.
Below are 8 of my bullets. For each, propose 2 metric angles I could legitimately add (latency, $ saved, % adopted, headcount unblocked, time-to-X reduced, NPS, MAU, conversion). For each metric: a sample value range and the question I should ask my manager to confirm a real number. Never invent specifics — only suggest the SHAPE of the metric.
[paste bullets]Variables to swap: paste 8 of your current bullets.
12. Senior / staff resume rewrite
I am applying for a [senior / staff / principal] role. Below are 6 of my bullets written at IC4 voice. Rewrite each in [target level] voice: emphasize scope (team, org, $), ambiguity navigated, and second-order impact (downstream metrics, cultural change, hiring leverage). Do not invent scope I do not have.
[paste bullets]13. Resume gap explainer line
I have a [N-month / N-year] gap on my resume ([caregiving / sabbatical / health / layoff search]). Write 2 versions of a one-line resume entry that names the gap honestly plus the most relevant thing I did during it (course, certification, project, freelance). Keep it confident, not apologetic.14. Quant resume → narrative resume
Useful for design, writing, or customer-facing roles where numbers don’t tell the story.
My current resume is metric-heavy but reads cold for a [design / writing / customer-success / strategy] role. Rewrite 5 bullets in a more narrative voice: 1 specific user problem, 1 concrete action I took, 1 short outcome (qualitative OK). Each bullet still under 30 words.
[paste bullets]15. Resume sanity-check vs JD
Below is my resume and the JD I want to apply to. For each JD requirement, mark: covered (where in resume), partially covered (what to strengthen), missing (whether to add or accept the gap). Then suggest the 3 highest-impact resume edits in order.
Resume: [paste]
JD: [paste]Common mistakes
- Listing duties, not outcomes.
- No quantification anywhere — research links quantified bullets to a markedly higher interview rate.
- The same verb starting five bullets in a row.
- Inventing metrics to please the model; recruiters will probe them.
- Sending one resume to 10 roles with no JD-alignment pass.
How to push results further
- Always tell the model the target level (junior / senior / staff). The same bullet reads differently at each tier.
- Demand the action-outcome-evidence triple in every bullet. If one part is missing, the bullet is weak.
- Cap bullets at 6 per role. Models love to expand to 12; recruiters scan, they don’t read.
- Keep one defensible metric per bullet. Numbers without referents (“improved performance by 40%”) read as invented.
- Run an ATS pass separately (template 9). Tables, columns, text boxes, and graphics break parsers that read top-to-bottom, so keep skills in plain bullets or pipe-separated lines.
- For senior+ roles, ask the model to emphasize scope and ambiguity navigated, not just deliverables.
- Save the JD-aligned version as a copy; keep your master resume neutral so you can re-tailor for each role.
FAQ
- How many bullets per role? Current role: 4-6. Past roles: 3-4. Old or unrelated roles: 1-2, or move them to an “additional experience” line.
- Should I tailor the resume for every application? Yes for roles you actually want. The 3 strongest edits per JD (template 15) take about 10 minutes and meaningfully lift callback rates.
- How quantified is “quantified enough”? Every bullet should carry at least one number — users, $, %, time, headcount, or rank. If you genuinely don’t have one, replace it with a specific name (a system, a customer, a launch).
- Are AI-rewritten bullets detectable? Bullets aren’t the problem. Resumes get flagged when the language reads generic. Keep one specific, company-internal detail per bullet (project codename, internal tool, real scale).
- Which model should I start with? Claude Opus 4.7 for honest, narrative rewrites; GPT-5.5 for fast keyword-dense passes; Gemini 3.1 Pro when targeting one named company. All three work on free tiers for a single resume.
- What if the model keeps inventing metrics I never had? Add: “Do not invent specific numbers. If a metric is uncertain, leave a
[CONFIRM: ...]placeholder I can verify with my manager.”
Related
- Cover letter prompts
- JD matching prompts
- LinkedIn bio prompts
- Behavioral question prompts
- Resume Achievement Quantification Prompts for Bullets
- How to Use AI to Write a Resume That Actually Gets Interviews
- How to Use AI to Rewrite Resume Bullets: From Duty-Lists to Outcome-Led, JD-Aligned
- ATS Resume Optimization Prompts: 12 Templates Without Keyword-Stuffing
External references: Jobscan — why ATS tables and columns break parsing and College Recruiter — the 6-second resume scan.
Tags: #Prompt #Job search #Resume