Generative AI in recruiting
Also called: GenAI in recruiting, LLM in hiring, ChatGPT for recruiting
Where generative AI helps in hiring
The honest list of useful current applications:
- Drafting job postings : a 200-word skeleton from role inputs, then edited by a human. Cuts authoring time from 30 minutes to 10.
- Outbound sourcing messages : a personalized first message based on the candidate’s recent work. Hand-edited before sending.
- Rejection-email language : localized, role-specific, polite. Reviewed once per role, then deployed.
- Interview-note summaries : turning 45 minutes of scattered notes into a structured scorecard. Reviewed and corrected by the interviewer.
- Internal docs : hiring plans, debrief summaries, role briefs.
Pattern: AI drafts; a human ships.
Where it fails
- Verbatim use of generated text : ships content the team doesn’t actually believe. Candidates notice .
- Hallucinated facts : the model invents salary bands, benefits, or company facts that aren’t true. Specific risk in job postings and offer letters.
- Generic personalization : outreach that says “I saw your recent work on [X]” without the model having actual access to [X]. Worse than no personalization.
Disclosure considerations
The EU AI Act requires disclosure when AI-generated content is presented as human-written in candidate-facing communication. Practical: a footer line on AI-drafted candidate emails saying so, or an internal policy that all AI drafts are reviewed before sending.
Where TalentGleam fits
TalentGleam’s generative-AI features (posting drafts, outreach drafts, rejection language) explicitly mark output as AI-drafted and require human review before send. The team always sees what’s about to leave. See the features page .
