Grok 4.7 for Job Seekers: 12 Prompts to Tailor Your CV (and What It Gets Wrong)

What changed in Grok 4.7 that matters for your CV

xAI shipped Grok 4.7 on September 21, 2026, at the same price and speed as Grok 4.6. The headline changes are a larger base model, longer training on multi-hour tasks, and stronger self-verification: the model checks its own output before it hands it back. For a job seeker, that last point is the one that matters. Earlier Grok versions would happily rewrite your CV and stop. 4.7 will re-read the job ad and flag its own mismatches if you ask it to.

It also handles long context better (xAI lists a 500K-token window), so you can drop your full CV, the complete job posting, your last performance review and a list of your real numbers into a single message without the model losing the thread halfway through.

One thing that hasn't changed: Grok still doesn't know how an ATS parses a PDF. It writes text. Whether Workday or Greenhouse can read that text is a separate question, and it's the one that decides if a human ever sees your CV.

Before you prompt: what to paste in

Grok 4.7 is only as good as its inputs. Give it three things every time: your current CV as plain text (not a screenshot), the full job description copied from the posting (LinkedIn and Indeed block link fetching, so paste the text), and a short fact sheet of your real numbers: team size, budget, revenue, users, tickets closed, percentages improved. The fact sheet is what stops the model from inventing metrics.

Set reasoning to 'high' for the rewrite prompts. 'Low' is fine for extraction and formatting, but a bullet rewrite at low effort reads like a template.

Keep every session in one chat so the model can reference the job ad it already parsed. Starting a new chat per prompt throws away the context that makes 4.7 useful.

Prompts 1–4: understand the job before you touch the CV

Prompt 1 — Extract requirements: "Here is a job description. List every hard skill, tool, certification and soft skill it asks for, grouped as must-have vs nice-to-have. Quote the exact phrase from the ad next to each one."

Prompt 2 — Find the hidden priority: "Based only on this job ad, what are the three problems this team is trying to solve by hiring? Rank them. Do not guess beyond what the ad says."

Prompt 3 — Gap map: "Here is my CV. For each must-have from the list above, tell me whether my CV shows it clearly, shows it weakly, or doesn't show it. Cite the line in my CV you're basing that on."

Prompt 4 — Keyword audit: "Which exact phrases from the job ad appear nowhere in my CV? Give me the list, and for each one tell me if I have real experience that could honestly use that phrase, based on what I've told you."

These four prompts do the thinking. If you skip them and go straight to 'rewrite my CV for this job', Grok 4.7 will produce something fluent and generic.

Prompts 5–9: rewrite with numbers, not adjectives

Prompt 5 — Bullet rewrite with a hard rule: "Rewrite my experience bullets for this job. Rules: start with a strong verb, include a number from my fact sheet in at least every second bullet, use the job ad's phrasing where I genuinely have that experience, never add a metric I didn't give you. Mark any bullet where you were tempted to invent a number with [NEEDS NUMBER]."

Prompt 6 — Summary in 3 lines: "Write a 3-line professional summary for this role. Line 1: who I am and years of experience. Line 2: the two skills from the must-have list I'm strongest in. Line 3: one quantified result. No buzzwords, no 'passionate'."

Prompt 7 — Reorder for the role: "Given the ranked priorities from Prompt 2, reorder my experience bullets and skills section so the most relevant items come first. Explain each move in one sentence."

Prompt 8 — Cut to length: "Cut this CV to fit two pages at 11pt. Remove the weakest bullets first and tell me which ones you removed and why."

Prompt 9 — Skills section: "Build a skills section using only tools and skills I've actually listed or described. Group them the way the job ad groups them. Flag any skill the ad wants that I'm missing so I can decide whether to learn it or leave it."

Grok 4.7's document-creation improvements show up here. It handles structure and length constraints better than 4.6 did, and it respects the 'don't invent' rule more reliably when you make it explicit and ask it to mark uncertain lines.

Prompts 10–12: make Grok check its own work

This is where 4.7 earns its upgrade. Self-verification only kicks in if you ask for it.

Prompt 10 — Honesty pass: "Re-read the CV you wrote against my original CV and fact sheet. List every claim, number or skill in the new version that isn't supported by something I gave you. Be strict."

Prompt 11 — Recruiter's 7-second read: "Pretend you're a recruiter who will spend seven seconds on this. What's the first thing you see? What's the first thing you'd doubt? What would make you close it?"

Prompt 12 — ATS text check: "Output the final CV as plain text with no tables, columns, icons or headers/footers. Tell me if any section heading uses a non-standard name an ATS might not recognise."

Prompt 12 is where Grok's limits show. It can format text cleanly, but it can't tell you how a specific ATS will score the result. For that you need a tool that actually parses the file.

Where Grok 4.7 still gets it wrong

Invented metrics. Even with the fact sheet, 4.7 will occasionally round up or add a percentage that sounds plausible. The [NEEDS NUMBER] marker in Prompt 5 catches most of these; Prompt 10 catches the rest. Never skip both.

Keyword stuffing in the summary. Ask for a 'strong summary' without constraints and you'll get five must-have phrases jammed into two sentences. The 3-line rule in Prompt 6 fixes this.

Confidence about ATS behaviour. Grok will state that a format 'is ATS-friendly' with no way to verify it. Treat that as an opinion, not a test result.

Hebrew and bilingual CVs. Grok 4.7 handles Hebrew text, but it defaults to American résumé conventions (objective statements, one page, 'references available'). If you're applying in Israel, tell it explicitly: two pages is fine, no photo, and keep ATS/LinkedIn/PDF in English.

The step Grok can't do: test the file

Once you have a tailored CV out of Grok 4.7, export it to PDF or Word and upload it to a free CV analysis at https://www.hrlens.io/cv. You'll get a score out of 100, five category scores (experience, tech stack, impact and ownership, clarity and structure, ATS compatibility), your strengths and a visual layout check. Paste the same job description into the job-targeting field and the analysis lists the skills the ad wants that your CV still doesn't show, then generates ATS-friendly rewrites in six templates that mirror the job's keywords.

That's the loop that works in 2026: Grok 4.7 for the thinking and the rewrite, an actual parser to confirm a machine can read what Grok wrote. If the ATS score comes back low, feed the missing-skills list straight back into Prompt 4 and run the cycle again.

If you're hitting a lot of generic ATS rejections, the free ATS resume checker at https://www.hrlens.io/for/ats-resume-checker-free covers the formatting side specifically.

Frequently asked questions