Customise Job Descriptions with AI

Job descriptions do two jobs at once: they attract candidates and they tell the AI matching engine what to look for. Most JDs do both badly. AI recruiting platforms can rewrite them to do both well, and the lift on candidate quality and quantity is meaningful, but the optimisation has to be done with the recruiter and hiring manager, not by the AI alone.

The five-step JD optimisation process

1. Audit

Score the current JD against three dimensions: clarity (does a candidate understand the role), specificity (does the AI have enough signal to match well), and inclusivity (does the language exclude qualified candidates unintentionally). Most JDs score 4 to 6 out of 10 on this baseline.

2. Rewrite

The AI proposes a tightened version: clearer skill requirements, fewer must-haves that are actually nice-to-haves, more honest seniority signal, less marketing-speak. The recruiter reviews and edits before publishing. The rewrite is a draft, not a publish-as-is.

3. Calibrate

Run the new JD against the existing talent pool and look at the top 30 returned candidates. If the new ranking matches your gut better than the old, the rewrite is a win. If it produces strange results, tune further.

4. Localise

For roles posted in multiple regions, the AI generates region-specific versions: comp ranges in local currency, language tone matched to local norms, statutory disclosures included. Saves recruiters from maintaining 6 versions of the same JD by hand.

5. Iterate

Once the role closes, the AI compares the candidates who actually performed well against the JD signal. Feedback loops back into the rubric for next time. The JD gets a little sharper with each cycle.

A well-tuned JD does double duty: it attracts the right candidates and gives the matching engine clear signal. Most teams treat the JD as marketing copy and lose the second job entirely.

Common rewrite improvements

• Tighten skills section: 4 must-haves not 12; nice-to-haves clearly separated
• Replace vague seniority language (“5+ years”) with concrete signal (“led a team through one rewrite”)
• Remove gendered or exclusionary language (the inclusivity audit catches more than recruiters expect)
• Add comp band; transparency lifts both quantity and quality of applicants
• Cut marketing copy that does not help the candidate decide

The trade-offs

Optimised JDs attract more candidates and match more accurately, but they can also feel less branded. The hiring manager who liked their version of the JD may push back on the AI’s tightening. The fix is collaboration, not capitulation; show them the data on candidate quantity and quality, then refine together.

How JDs flow into the matching engine

The structured fields in the JD (must-haves, nice-to-haves, seniority signal) become the rubric the AI scores against. The narrative section feeds the semantic-match layer. The clearer the JD, the cleaner the match. See how AI matches candidates to job descriptions for the full pipeline.

What good vendors offer

• JD audit and rewrite as a first-class feature, not a hidden setting
• Side-by-side preview of how candidates rank under old vs new JD
• Inclusivity language flagging built in, not as a separate tool
• Localisation that handles statutory disclosures by region
• Closed-role feedback loop that feeds back into rubric tuning

For the broader productivity context, see AI summarising candidate information. For the matching engine, see the full matching pipeline.