TL;DR (June 2026): Build one plain-text master resume, then use AI to spin a JD-tailored variant in about 5 minutes. For the rewrite itself, Claude Opus 4.7 is the safer default — it invents fewer metrics and uses less marketing filler than GPT-5.5, and its 1M-token context swallows your full resume plus the JD plus a LinkedIn export without the copy-paste shuffle. The hard rule across every prompt below: AI does the wording, you own every number and every seniority claim. No major ATS (Workday, Greenhouse, iCIMS) actually flags “AI-written” resumes, but a human recruiter will spot suspiciously round metrics and generic filler in seconds, so specificity beats polish.
The task
You have applied to 32 roles in the last 6 weeks. Three callbacks. The resume is the same one you have used for two years, with one new bullet bolted on after each role. The format is fine, the content is mostly true, and yet recruiters either skim past it or the ATS quietly drops it before any human reads it. You want one master English resume that holds together, plus a fast way to spin a JD-tailored variant in 5 minutes, not 90.
Which model to use
All three frontier assistants can rewrite a resume. They are not equally good at it. Prices below are per month, as of June 2026.
| Model | Best for | Watch out for | Plan that unlocks it |
|---|---|---|---|
| Claude Opus 4.7 | The full rewrite and strategic reframing; hallucinates fewer metrics, less filler, 1M-token context for resume + JD + LinkedIn at once | Slower than GPT-5.5 | Claude Pro $20 (bundles Claude Code + Cowork); free tier runs the weaker Sonnet 4.6 |
| GPT-5.5 (ChatGPT) | Fast first drafts, keyword density, knows industry jargon (OKRs, sprint velocity) out of the box | Most likely to invent numbers and team sizes; US free tier now shows ads since Feb 2026 | ChatGPT Free $0 (tight limits) / Plus $20 |
| Gemini 3.1 Pro | If you live in Google Docs / Workspace; 1M context, edits inline | Less surgical on seniority calibration | Google AI Pro $19.99 (formerly “Gemini Advanced”) |
Practical rule: draft fast in GPT-5.5 if you want momentum, then run the final pass through Claude Opus 4.7 to strip invented metrics and overstated seniority. If you only pay for one, pay for Claude.
Where AI helps, and where it does not
AI is genuinely good at three resume tasks: tightening loose bullets into STAR structure, mirroring a JD’s vocabulary back into your experience without sounding like a parrot, and ranking your weakest bullets so you know what to fix first. It will also reformat for ATS without losing readability, preserving section order, removing graphics, and fixing the date format hiring systems prefer. Where AI fails: it does not know which of your past roles is the right one to emphasize for a Series B fintech versus a FAANG infra team. It also cannot verify your metrics; if you tell it “I improved retention by 30%,” it will write that number and a hostile interviewer will ask you to defend it. The strategic decisions (what to emphasize, what to drop, which metrics to claim) stay yours.
A common failure mode: the model defaults to making every role sound like senior-level work. Your IC2 bullet reads like a director’s bullet, the resume passes the keyword filter, and then the screening call falls apart in 8 minutes because nothing matches your actual level. Tell the model your real level for each role explicitly.
One more 2026 caveat: AI-detection tools are not your enemy at the ATS layer. As of June 2026 no major ATS (Workday, Greenhouse, iCIMS) screens for AI-generated text, and standalone detectors run 30-50% false-positive rates, so they are unreliable. The real risk is human: recruiters now flag resumes that read as AI-default. The tells are suspiciously round numbers (“increased sales 40%”) instead of asymmetric real ones (“$847K to $1.2M”), and bullets of near-identical length and rhythm. Specificity is the defense.
What to feed the AI
- Your current resume in plain text — not a PDF, not a screenshot
- The target job description in plain text (the full posting, not the marketing tagline)
- Your real level for each past role — IC3, manager of 5, senior staff, founding eng
- The 3 strongest concrete wins you want to surface (quantified if you have the number, with the source)
- The career arc you are telling — “infra to platform,” “IC to manager,” “second-time founder,” “regulated industries to consumer”
- Hard caveats — gaps you need explained, a role you want minimized, a title that does not match your real work
- Geographic context — US resumes are different from UK / EU / SG / HK; pick one target market
- Banned phrases — your AI tics (“delve,” “tapestry”), and your industry’s spam-filter phrases (“synergize,” “thought leader,” “passionate”)
Copy-ready prompt
You are a senior tech recruiter with 8 years of experience in {US / UK / EU / SG / HK}.
Rewrite my resume to pass ATS filters and match the target JD below.
Requirements:
1) Reverse-chronological layout. Section order: Summary, Experience, Projects, Skills, Education.
2) Each bullet uses STAR structure and emphasizes outcomes (quantified where I gave you a number).
3) No fluffy adjectives. Replace each adjective with the artifact or delete it.
4) Naturally include keywords from the JD; do not keyword-stuff. The keyword must appear in a sentence that uses it in context.
5) Match my real level at each role (I tell you below). Do not promote my IC2 work into staff-level bullets.
6) Length: one page if I have <10 years of experience, two pages max otherwise.
7) Do NOT invent metrics. If a bullet would need a number to be strong and I did not give you one, leave a "[NUMBER]" placeholder and ask me at the end.
My career arc in one sentence: {one sentence}
My real level at each past role: {role 1: IC2, role 2: senior IC, role 3: tech lead, etc.}
Strongest 3 wins to surface: {paste with metrics where you have them}
Hard caveats: {gaps to explain, role to minimize}
Banned phrases: {your AI tics, your industry's spam-filter words}
Current resume: {paste plain text}
Target JD: {paste plain text}
After the resume, list the 3 weakest areas of my background relative to this JD and how I could strengthen them.Shorter variant — single-role rewrite
Rewrite only the bullets for the {role name} section of my resume.
Current bullets: {paste}
My real level at this role: {level}
The JD context I am rewriting these for: {paste 3-5 lines of JD}.
Return 3 alternative versions of each bullet, ranked by strongest specificity. Do not invent metrics.Shorter variant — recruiter 20-second scan
Act as a recruiter scanning the resume below in 20 seconds.
Identify the 5 weakest bullets and rewrite each one more compelling — without inventing facts.
Then name the single bullet that would best move on top of page 1.
Resume: {paste}
Target JD: {paste}Sample output
A good STAR bullet rewrite: original “Improved checkout conversion.” becomes “Owned the checkout funnel redesign at Mercado Pago — 7-step to 3-step flow with conditional KYC; six-month read showed drop-off dropped from 38% to 14% and monthly new-merchant net adds grew from 4.2k to 11.6k.”
A good summary line: original “Results-driven product manager with passion for building great products.” becomes “Product manager focused on regulated payments — 4 years at Mercado Pago shipping KYC, merchant onboarding, and dispute resolution; built and currently run the team that owns Latam checkout.”
A useful weak-area callout: “Your resume shows 4 years of payments PM work but no exposure to underwriting or credit risk, which this JD names as a ‘plus.’ Two ways to strengthen: (1) add one bullet from the dispute resolution work that touches risk-adjacent decisions, (2) be prepared in the screen for the question ‘how would you approach a credit-risk PM problem if you have not owned credit before.’”
How to refine
- If every role reads like senior work: “Recalibrate to the levels I gave you. The IC2 bullet should not use ‘led’ or ‘owned’; it should say ‘contributed,’ ‘built,’ ‘shipped’ with a smaller scope. Match the language to the level.”
- If keywords feel parroted: “Every JD keyword you used must appear in a sentence that uses it in context, not as a list. If a keyword does not fit naturally, drop it.”
- If bullets have invented numbers: “Replace any metric I did not give you with a
[NUMBER]placeholder and ask me for the real one. Do not invent.” - If the summary feels generic: “Rewrite the summary in 2 sentences. Sentence 1: who I am right now (current role + scope + domain). Sentence 2: the one differentiator. Cut every adjective.”
- If the resume is too dense: “Cut 15% of the words. Anything that survives is signal. Prioritize bullets with numbers over bullets without.”
Common mistakes
- Letting AI invent metrics: the screen call ends the moment you cannot defend a number; recruiters can smell an inflated bullet from the second sentence.
- Keyword-stuffing for ATS: modern parsers weight where and how a keyword appears, not raw count; a term buried in white text or a skills dump reads as suspect to both the filter and the human reviewer. Put the keyword inside a real sentence.
- Pasting the same resume to every company: the JD-tailored version takes 5 minutes; the cost of not doing it is ~3x more applications for the same callbacks.
- Promoting your real level: “led a team of 8” when you mentored 2 people during a sprint, and the resume gets you past the filter, but you fail the screen.
- Adjective stacking in the summary: “results-driven, passionate, detail-oriented product manager” is the spam-filter sentence; replace with the role + scope + domain.
- One-page rule applied wrongly: 12 years of experience compressed onto one page becomes unreadable; the rule is one page if you have under 10 years, otherwise two.
- Bullets without artifacts: “Improved processes” tells the reader nothing; lead with what shipped, what changed, who used it.
- Ignoring geography: US resumes do not include a photo; some European markets expect one; UK leans more formal; SG/HK tolerate longer; pick the convention for the target market.
FAQ
- Should I keep my master resume version-controlled?: Yes. Keep one master in Google Docs or a private repo, and spin per-JD variants as duplicates. The master accumulates wins, the variants get pruned for the JD. Re-merge wins into the master every 3 months.
- What about creative resumes — designer portfolios, infographics?: Different track. The two-column “creative” resume often fails ATS. Submit a plain-text master through the ATS, and link to your portfolio site for the visual version.
- How many tailored variants per job search?: Realistically, 3-5 templates per target archetype (FAANG infra, Series B fintech PM, etc.), then 5 minutes of per-company tweaks. Anything more is over-engineering.
- Should I use AI to write the cover letter from the resume?: Yes, but in a separate session and a separate prompt — see the linked cover letter article. The resume voice and the cover letter voice are different jobs.
- My ATS keeps rejecting my resume — is the AI rewrite enough?: Probably not, if the underlying file format is wrong. As of June 2026, modern parsers (Workday, Greenhouse, Lever, Ashby) extract text from a text-selectable PDF as cleanly as from a
.docx, so either works for those. DOCX stays the safest single choice because some legacy systems (Oracle Taleo, older iCIMS) still parse it more reliably. The one fatal mistake is an image-only PDF or a Canva export with embedded graphics — the parser sees zero text. Test before you submit: paste your “submitted” file into a plain-text editor; if anything scrambles or sections drop out, the ATS sees the same scramble. - What format does the parser actually want?: Single column, reverse-chronological, a system font (Arial, Calibri, Helvetica) at 10-12pt, standard section headers (“Work Experience,” “Education,” “Skills”). No tables, text boxes, multi-column layouts, or text in the header/footer region — those are the layout features that confuse parsers most. The AI rewrite handles the wording; you handle the file structure.
- Will the round numbers AI gives me get me flagged?: They can, by a human. As of June 2026 no mainstream ATS detects AI text, but recruiters are trained to distrust clean round metrics. Replace every AI-supplied “improved X by 30%” with your real, asymmetric figure (“from 38% to 14%”). If you do not have the number, cut the claim rather than round it.
Related
- ATS Resume Optimization Prompts
- Resume Bullet Rewrite AI
- JD Matching Analysis AI
- LinkedIn Bio AI
- STAR Interview Answers
- ChatGPT Resume Workflow
- Cover Letter Prompts
- AI Cover Letter
Sources
- Jobscan, “Can ATS Detect AI Resumes?” — confirms no major ATS flags AI text and explains how parsers read files.
- Anthropic, Claude pricing and OpenAI, ChatGPT pricing — current plan tiers cited above.
Tags: #Resume #Job search #AI writing #Prompt