Resume vs. the Screener

What this is for

You're applying to a specific job and want to know, before you hit submit, whether your resume clears the keyword filter and whether the recruiter skimming it will see what they need to see. This is for that moment, not a general resume tune-up.

What an ATS actually does (worth 60 seconds, because most advice gets it wrong)

The systems behind most corporate job postings (Workday, Greenhouse, iCIMS, Lever, Taleo) are primarily databases, not robot judges. They parse your resume into fields, store it, and let recruiters search and filter the pile by keywords, titles, and answers to the application's knockout questions. Two consequences that change how you use this resource:

1. There is usually no single "ATS score" that auto-rejects you. Rejection mostly happens when a recruiter's keyword search doesn't surface you, or a knockout question filters you. So Prompt 1's 0-100 score below is a rehearsal proxy: a disciplined way to find the keyword gaps a recruiter's search would miss you on, not a number any real system computes.
2. Parsing failures are real, and they're mechanical. Before any keyword matters, your resume has to survive being read by software. The known killers:
   • Multi-column layouts (columns get read in the wrong order or merged)
   • Text inside graphics, logos, or text boxes (invisible to the parser)
   • Contact info in the header/footer region (some parsers skip headers entirely)
   • Tables for layout (cells read out of sequence)
   • Unusual section headings ("My Journey" instead of "Experience")
   • Fancy templates exported oddly; a clean single-column .docx or text-based PDF parses most reliably

Fix those six before running any prompt; no rewrite survives a parser that read your phone number as a job title.

Before you start

• The full job posting, copied as plain text (not a link; paste the actual text)
• Your current resume, either pasted as text or uploaded as a file
• Your last three roles' bullet points, ready to paste separately if you're using Prompt 2 on its own

The prompts

Prompt 1: Score it like the ATS and the recruiter would

Use either ChatGPT or Claude. Upload your resume file (paperclip icon in either tool, or drag it into the chat) and paste the job post text into the same message.

ROLE: You are a corporate recruiter and an ATS keyword-matching
system, evaluated together.

CONTEXT: My resume is attached (or pasted below). Here is the job
posting I'm applying to:
[PASTE FULL JOB POST TEXT]

CONSTRAINTS:
- Score match on a 0-100 scale the way a recruiter's keyword search
  would effectively rank it (literal term matches: skills, tools,
  certifications, job titles). Treat this as a search-visibility
  score, not an official ATS number.
- Separately, score how a human recruiter would react in a quick
  first skim (clarity of seniority, relevance of top bullets, red
  flags).
- Identify exactly 3 keywords or phrases from the job post that are
  missing or under-represented in my resume. Do not suggest more
  than 3. I need the highest-leverage gaps, not a wish list.
- Do not rewrite my resume in this response. Just score and diagnose.

OUTPUT FORMAT:
1. Search-visibility score (0-100) with the 5 terms that matched and
   the 3 that didn't
2. Recruiter skim reaction (2-3 sentences, blunt)
3. The 3 missing keywords, each with the exact sentence in my resume
   where it could honestly be added (no fabricating experience I
   don't have)

Prompt 2: Rewrite your last 3 bullets as outcomes with numbers

Use either model. Do this after Prompt 1 so the AI knows which keywords to work in.

ROLE: You are a resume editor who specializes in converting task
lists into outcome statements.

CONTEXT: Here are my last 3 bullets, verbatim:
[PASTE 3 BULLETS, e.g. "Managed a team that handled customer
escalations"]

The job post I'm targeting emphasizes: [PASTE 2-3 PRIORITIES FROM
POST, e.g. reducing churn, cross-functional leadership, budget
ownership]

CONSTRAINTS:
- Rewrite each bullet in the format: [Action verb] + [what you did] +
  [measurable result]. If I didn't give you a number, ask me for one
  instead of inventing it.
- Keep each bullet under 25 words.
- Do not add skills, tools, or scope I did not mention.

OUTPUT FORMAT: The 3 rewritten bullets, plus a flagged list of any
bullet where you need a real number from me before it's finished.

What the transformation looks like. In: "Managed a team that handled customer escalations." Out:

"Led an 8-person escalations team through a 40% volume spike, cutting average resolution time from 11 days to 4 and saving the two largest at-risk accounts."

Same job, same person. The first version describes a chair someone sat in; the second describes what changed because they sat in it. If you don't have the numbers, that's what the prompt's ask-me-first constraint is for: dig them up or scope the claim honestly, but don't ship the chair.

Prompt 3: Compress 25 years into 2 pages without erasing seniority

Use either model. This is the one people get wrong by either cutting too much (looks junior) or cutting too little (looks unfocused). Paste your full resume, even if it's currently 3-4 pages.

ROLE: You are a resume editor for senior candidates with 20+ years of
experience, targeting a 2-page format.

CONTEXT: Full resume pasted below. I am applying for [JOB TITLE]
roles similar to this one: [PASTE JOB POST TEXT]

CONSTRAINTS:
- Full bullet detail (3-4 bullets each) only for the 3 most recent or
  most relevant roles.
- Roles older than 15 years get one line each: title, company, years.
  No bullets, unless a role contains the single most impressive
  credential of my career, in which case give it one bullet.
- Cut or merge redundant roles (e.g. two similar titles at the same
  company) into one line.
- Do not remove graduation years unless I explicitly ask. Removing
  them selectively looks worse than leaving them, since gaps get
  noticed.
- Total output must fit 2 pages at 11pt, which is roughly 900-1,100
  words including headers.

OUTPUT FORMAT: The full compressed resume, followed by a one-line
note on what you cut and why.

Prompt 4: Generate the 5 questions this job post implies they'll screen for

Use either model. Run this last; it tells you what to rehearse before the call, not what to fix on the page.

ROLE: You are the hiring manager who wrote this job posting.

CONTEXT: Job post: [PASTE FULL JOB POST TEXT]

CONSTRAINTS:
- Infer the 5 questions you would actually ask in a first-round
  screen, based on what the post emphasizes (not generic interview
  questions).
- For each, name which line or requirement in the post it's testing
  for.
- Order them from most likely to least likely to come up first.

OUTPUT FORMAT: A numbered list of 5 questions, each followed by one
line naming the requirement it tests.

When those questions start turning into interviews, the Ageism-Proof Interview Prep picks up where this leaves off.

The "dates back to 1998" problem

If your resume shows work history stretching back 20+ years, the fix isn't hiding the start date. Recruiters know how math works, and an obvious gap or vague date range reads as evasive, not younger. The fix is asymmetric detail: give full bullets to the last 10-15 years, and reduce anything older to a single line with title, company, and years, no bullets. One exception: if an early-career role contains a credential that still matters (you built the system half the industry now uses, you were first in a role that didn't exist yet), keep one bullet for it. Prompt 3 above does this automatically; just don't let the AI talk you into deleting the line entirely.

A warning on keyword stuffing: adding the 3 missing keywords from Prompt 1 means working them into real sentences about real work, not pasting a skills list at the bottom in white text or cramming synonyms into a bullet until it reads like a Boolean search. Recruiters have gotten good at catching this, and a recruiter who catches it will read your whole resume more skeptically. If a keyword doesn't honestly fit anywhere in your actual experience, leave it out. That's a real gap, not a formatting problem.

Where this goes wrong

• You paste the job title instead of the full post. The AI needs the actual language of the posting to find real keyword gaps; "Senior Marketing Manager" tells it nothing. Always paste the full text.
• You let the AI invent numbers for Prompt 2. If you don't have a metric, say "I don't have an exact number, estimate a reasonable range and mark it as an estimate" or leave it as a scope statement instead. A fabricated number that comes up in an interview is worse than no number.
• You apply the same compressed resume to every job. Prompt 3's cuts are specific to the job post you fed it. Re-run it per posting, or at minimum per role type. A resume compressed for a "Director of Operations" post won't emphasize the right 3 roles for a "VP of Supply Chain" post.
• You polish keywords on a resume that doesn't parse. The mechanical checklist at the top comes first. A perfect resume in a two-column template can lose to a mediocre one the software could actually read.

The 2-minute version

Fix the six mechanical parse-killers first. Paste your resume and the job post into one message. Run Prompt 1 for the score and the 3 gaps. If the score's low, run Prompt 3 for the 2-page compression. Skim the Prompt 4 questions on your way to the interview.

---

Want to know where you actually stand with AI? The free AI Readiness Assessment takes 10 minutes and gives you a personalized roadmap.

Weekly, on Tuesdays. The Gen Xcelerator: practical AI intelligence for people who already know how to do the job. Subscribe free.

Experience Is the API. GenXcelerate