TL;DR
Paste the JD and your resume into ChatGPT or Claude, and ask for an honest 1-10 fit score plus a per-gap prep plan — not a “you should apply!” pep talk. Add the line “be honest, do not score generously to encourage me,” because the default behaviour is to inflate. A fit score answers “can I bridge the gaps in time?”, which a keyword/ATS score (Jobscan-style, target 75-80%, as of June 2026) cannot. Use Claude Opus 4.7 or Sonnet 4.6 for the reasoning (it hallucinates fewer skills you don’t have); copy the prompt below.
The task
You are weighing a role and are not sure the fit justifies the interview-prep time. You want an honest 1-10 fit score — not the cheerleading AI defaults to — plus a prioritised gap-bridging plan you can run in 3-5 days before you submit or start the loop. Without it, you over-prepare on irrelevant gaps and under-prepare on the dealbreakers.
This is a fit score, not an ATS keyword score. They answer different questions:
| Fit score (this prompt) | ATS / keyword score (Jobscan, Teal, Enhancv) | |
|---|---|---|
| Question answered | Can I do the job, and bridge gaps in time? | Does my resume’s wording match the JD’s? |
| Scale | 1-10 with reasoning per gap | 0-100% keyword overlap; target 75-80% as of June 2026 |
| Best for | Deciding whether to apply and what to prep | Polishing the resume you’ve decided to submit |
| Blind spot | No view of how the parser reads your file | No view of whether you can actually do the work |
Run the fit score first to decide whether to apply; run an ATS check after you’ve decided to submit. Note that the often-cited “75% of resumes are auto-rejected by ATS” figure is largely a myth — a 2025 study of US recruiters found 92% don’t set content-based auto-reject rules. What actually gates you before a human are knockout questions (work authorization, hard minimum-years requirements) and resumes that fail to parse (tables and columns cause a large share of failures). A fit score surfaces the knockout-question risk early.
When AI helps — and when it does not
AI is excellent at structured scoring, surfacing partial-match requirements (you have something, but not exactly), and proposing a prep task per gap. It is poor at calibrating fit: it scores generously by default, and ChatGPT in particular will invent skills or certifications you never claimed if you let it write the resume side too. Tell it explicitly: “be honest, do not score generously to encourage me.” Keep it scoring only the evidence you paste in. If the stakes are high, cross-check with one person actually in the industry.
What to feed the AI
- The full JD
- Your resume / portfolio summary (concise)
- Your career stage (junior / mid / senior / staff / director)
- The interview process you know about (rounds, panels, take-home)
- What you cannot move on (location, comp floor, hours)
- Time budget for prep (days, hours per day)
Copy-ready prompt
Score my fit for this role and build a prep plan.
JD:
"""
<paste>
"""
My resume / portfolio summary:
"""
<paste>
"""
Career stage: <line>
Interview process I know about: <line>
What I cannot move on: <list>
Prep time budget: <days, hours/day>
Be honest. Do not score generously to encourage me.
Return:
1. Fit score 1-10, with one-sentence reasoning
2. Three must-haves I clearly meet — with evidence from my background
3. Three must-haves I partially meet — what I have and what's missing
4. Three nice-to-haves I have
5. Three gaps — each ranked dealbreaker / coachable / fluff
6. A prep plan: for each gap, the specific task I should do this week
7. Recommendation: apply / apply if I bridge gaps / pass
8. The single thing in the JD I should over-prepare for
If a gap is a dealbreaker I cannot reasonably bridge in my time budget, say so.For competitive roles, append: “Add a ‘storytelling angle’ — given my background, what story should I tell about why my path leads to this role?”
Which model. As of June 2026 either flagship works, but they behave differently here. Claude Opus 4.7 or Sonnet 4.6 (free tier covers limited Sonnet 4.6) tends to stay grounded in the evidence you paste and pushes back more readily, which is what you want from an honest scorer. ChatGPT GPT-5.5 (use the Thinking mode in the picker) is faster but more eager to please — watch for inflated scores and any skill it credits you with that you didn’t paste in. Both hold a 1M-token context as of June 2026, so a long JD plus a full resume fits with room to spare; you do not need to trim either input.
What a usable result looks like
A score with one-line reasoning at the top, then the four lists (meet / partial / nice-to-have / gaps with each gap tagged dealbreaker / coachable / fluff), then a prep plan with one specific task per gap, then a single apply/bridge/pass recommendation. The “one thing to over-prepare for” is your rehearsal priority — the item most likely to decide the loop. If the model returns one undifferentiated wall of “strengths,” it skipped the honest part; re-prompt with “rank the gaps, and tell me which one is a dealbreaker.”
How to check the output is usable
- The score has reasoning, not just a number
- Partials are honest (half-evidence is not full evidence)
- Gaps are ranked dealbreaker vs coachable, not all “important”
- Prep tasks are specific (“write 3 STAR stories on system design”) not generic (“study system design”)
- The single over-prepare item is rehearsal-ready
Common mistakes
- Asking only for the score, not the prep advice. Score without action is just procrastination
- Ignoring the gap-bridging suggestions. They are the value
- Applying without bridging known gaps. They surface in round 2
- Trusting a generously-scored AI. Push back if scores feel too high
- Skipping the over-prepare item. That is the single most valuable rehearsal
FAQ
- Should I apply with a 6/10 fit? Yes, if you can bridge the top gaps inside your prep window. No, if a gap is structural (“5 years of platform X you don’t have”) or a knockout requirement you can’t meet (a security clearance, a specific license, work authorization).
- What if my resume looks weak next to the JD? Either pass, or address the gap directly in the cover letter. Don’t let it surface as a surprise in the interview — a named, explained gap reads better than a discovered one.
- How often should I re-score? Once per JD. A fit score is JD-relative, not an absolute rating of you: the same resume can score 8/10 for one team and 4/10 for a near-identical title elsewhere, because each JD weights different must-haves.
- Is a fit score the same as an ATS match score? No. The fit score tells you whether to apply and what to prep; an ATS/keyword tool (Jobscan, Teal) tells you whether your resume’s wording matches the JD, targeting roughly 75-80% as of June 2026. Run the fit score first, the ATS check after you’ve decided to submit.
- Will the AI just tell me to apply? Often, unless you stop it. The “be honest, do not score generously” line is doing the real work — without it, both ChatGPT and Claude lean encouraging. If a score still feels high, ask “what would make this a 5 instead of an 8?”
For a sanity check on what recruiters actually screen on, Harvard Business Review’s coverage of hidden-worker hiring filters is a good primer on how knockout requirements gate applications.
Related
- Resume bullet rewrite AI — sharpen the bullets that get scored
- Job description analysis — JD decoding without scoring
- Company role research AI — research before applying
- AI resume writing — tailor resume per JD
- Cover letter customisation prompts — cover letter per JD
- Mock interview AI — practise on the over-prepare item
- STAR interview answers — behavioural prep for matches