Home › Glossary
Recruitment AI glossary
32 terms, defined in one or two sentences each. Link directly to any term with #id.
Adverse impact · Agent (agentic AI) · AI screening · AI sourcing agent · Applicant tracking system (ATS) · Automated employment decision tool (AEDT) · Bias audit · Boolean search · Candidate matching · Contact enrichment · Conversational AI · Deliverability · Embedding · EU AI Act · Explainability · GDPR Article 22 · Grounding · Hallucination · Human in the loop · InMail · Interview intelligence · Knockout questions · Large language model (LLM) · LinkedIn automation · NYC Local Law 144 · Outreach sequence · Recruitment AI · Scorecard · Semantic search · SOC 2 · Structured interview · Talent pool ·
- Adverse impact
- A substantially different selection rate for one group compared with another. The four-fifths rule (a selection rate under 80% of the highest group's) is the common US screening threshold.
- Agent (agentic AI)
- A language model given tools and a goal that plans and executes several steps towards it, with a human reviewing the result rather than each step.
- AI screening
- Using models or chat flows to sort or filter applicants before human review, by match score or knockout questions.
- AI sourcing agent
- An autonomous system that takes a role brief, searches candidate sources, ranks matches, finds and verifies contact details, writes and sends personalised outreach (typically by email and SMS) and returns interested candidates. Noon is the reference example.
- Applicant tracking system (ATS)
- The system of record for jobs and candidates, moving applicants through stages. Most recruitment AI either lives inside the ATS or integrates with it.
- Automated employment decision tool (AEDT)
- The NYC Local Law 144 term for a computational tool that substantially assists or replaces discretionary hiring or promotion decisions.
- Bias audit
- An independent evaluation of an automated hiring tool's outcomes by sex, race and ethnicity, required annually by NYC Local Law 144 for automated employment decision tools.
- Boolean search
- Searching with AND, OR, NOT and quotation marks. Being replaced by semantic search that understands meaning rather than keywords.
- Candidate matching
- Scoring how well a candidate fits a job, usually with a model trained on past outcomes. The most regulated use of recruitment AI when it influences rejection.
- Contact enrichment
- Finding and verifying a person's email address, phone number or other details from a name and profile. Subject to privacy law; verification keeps outreach deliverable.
- Conversational AI
- Chat and SMS assistants that answer candidate questions, collect information and schedule interviews, common in high-volume hourly hiring.
- Deliverability
- Whether email actually reaches the inbox. Depends on domain reputation, authentication (SPF, DKIM, DMARC), volume and content quality.
- Embedding
- A numerical representation of text produced by a model, such that similar meanings have similar numbers. The basis of semantic matching.
- EU AI Act
- European regulation (phased in 2025–2027) classifying AI used in recruitment and selection as high-risk, with risk-management, oversight, transparency and logging duties.
- Explainability
- The ability of a system to give an understandable reason for an individual output, such as why a candidate was ranked highly.
- GDPR Article 22
- The EU/UK rule giving individuals the right not to be subject to solely automated decisions with legal or similarly significant effects, including hiring, without safeguards such as human review.
- Grounding
- Constraining a language model's output to facts drawn from supplied data (a profile, a transcript) so it does not invent details.
- Hallucination
- A fluent but false statement generated by a language model, such as an invented qualification. Mitigated by grounding outputs in source data.
- Human in the loop
- A process design where a person reviews or approves AI output before it takes effect, required by many regulations for consequential decisions.
- InMail
- LinkedIn's paid messaging to people you are not connected with, with monthly caps per Recruiter seat.
- Interview intelligence
- Tools that record, transcribe and summarise interviews, writing notes against scorecards and producing interviewer analytics.
- Knockout questions
- Minimum-requirement questions (licence, location, availability) that automatically disqualify applicants who do not meet them.
- Large language model (LLM)
- A model trained on large amounts of text that can read and write natural language. Used in recruiting to draft outreach, summarise interviews, extract requirements and power chatbots and agents.
- LinkedIn automation
- Software that sends connection requests or messages through a LinkedIn account automatically. Prohibited by LinkedIn's terms and a risk to the account; reputable sourcing agents reach candidates by email and SMS instead.
- NYC Local Law 144
- New York City law (enforced from July 2023) requiring annual bias audits, public summaries and candidate notice for automated employment decision tools.
- Outreach sequence
- A scheduled series of messages and follow-ups sent to a candidate until they reply or the sequence ends.
- Recruitment AI
- Software that performs a recruiting task that previously required human judgement: finding candidates, writing to them, reading applications, answering applicants, scheduling, note-taking or predicting fit. Also called AI recruiting or AI in talent acquisition.
- Scorecard
- The set of criteria interviewers rate a candidate against. AI note-takers organise their output by scorecard.
- Semantic search
- Search that matches on meaning using embeddings, so 'growth marketer' can match 'demand generation lead'.
- SOC 2
- An independent audit report on a vendor's security controls. A reasonable minimum to request from any recruitment AI vendor handling candidate data.
- Structured interview
- An interview where every candidate is asked the same questions and rated on the same criteria. The strongest single lever for fairer hiring, and a prerequisite for useful interview AI.
- Talent pool
- A saved group of candidates, past applicants or prospects maintained for future roles, often re-engaged by automation.