Audit a JD for Red Flags
When to use it
Sanity-check a JD — yours or a hiring manager's draft — before it goes live and starts costing you applicants.
The prompt
Read this job description three times, each time as a different strong candidate deciding whether to apply. {{Paste JD here}} STATED SENIORITY AND COMP: {{so you can judge whether the asks are proportionate}} THE THREE READERS: someone who meets about 70% of the requirements and is deciding whether to bother; someone currently employed, well paid and not actively looking; and someone from a background that is under-represented on this team. For each reader, tell me what makes them close the tab, quoting the exact line that does it. Then, across all three: 1. Every requirement that's unrealistic for the stated seniority or comp, and what a realistic version says. 2. The count of must-have requirements, and which ones to cut first. 3. Any phrasing that reads as a warning about the team or the work — "thrives under pressure", "comfortable with ambiguity", "wears many hats" — and what a candidate infers from it. 4. Anything vague enough that a candidate can't tell what the day-to-day actually is. 5. What's missing that a strong candidate needs in order to say yes: scope, who they'd report to, why the role is open, how success gets measured. 6. For every line you flagged above, the exact replacement wording — the original line, the rewrite, and one sentence on what the original was doing wrong. Leave the lines that are already fine alone; I want a short list of real changes, not the whole JD reformatted. Rank your findings by how many good applicants each one is likely costing me. Be honest if the JD is already fine — don't invent problems to fill the list.
Tip
Run this on every JD before it goes live. It catches the same handful of issues nearly every time — too many must-haves, vague scope, an unintended warning in the tone — and hands you the replacement wording rather than a list of complaints.
How to use it
Paste the prompt into ChatGPT, Claude or whichever assistant you use, then replace every {{bracketed}} part with your own detail. The more specific and messier your input, the better the output — a model given raw notes has more to work with than one given a tidy summary you wrote first.
More Job Descriptions prompts
Open the full tool for all 87 prompts, the glossary, and the JD decoder.