This AI recruiting glossary cuts through the marketing fluff, delivering definitions for 40+ essential hiring terms organized into five operational categories: core technical foundations, top-of-funnel sourcing, automated candidate screening, hiring economics, and regulatory compliance.

1. Artificial Intelligence (AI) in Recruiting

A broad umbrella term describing computer systems and software engineered to perform tasks that historically required human cognitive capabilities. In recruitment, AI systems analyze natural language across resumes, evaluate spoken candidate responses, identify patterns in hiring data, and automate repetitive administrative scheduling.

2. Machine Learning (ML)

A specific subset of artificial intelligence where software algorithms improve their performance over time through exposure to data, rather than following static, hardcoded rules. In talent acquisition, ML algorithms learn from historical hiring outcomes, candidate progression rates, and recruiter decisions to improve candidate matching accuracy.

3. Natural Language Processing (NLP)

The branch of artificial intelligence focused on enabling computers to read, interpret, understand, and extract structured meaning from human language. Modern NLP engines analyze conversational context, transferable skills, and project scope across candidate resumes and interview transcripts, moving far beyond basic character-matching filters.

4. Large Language Models (LLMs)

Advanced deep learning neural networks trained on massive corpora of text data (such as GPT-4, Claude, or Llama). In recruitment technology, LLMs power natural language sourcing queries, draft tailored job descriptions, summarize lengthy interview transcripts, and assist in extracting evidence-backed candidate quotes.

5. Semantic Search

A search methodology that interprets the underlying meaning and intent of a query rather than matching literal keywords or exact text strings. While a legacy keyword search for "software engineer" might miss an applicant titled "backend developer," semantic search understands that both profiles possess overlapping technical competencies.

6. Generative AI

Artificial intelligence architectures capable of creating new, original content including text, audio, images, and code based on user prompts. In modern hiring, recruiters use generative AI to build customized assessment rubrics, while job seekers use it to craft tailored resumes and cover letters.

7. AI Hallucination

An instance where a generative AI model outputs factually incorrect, fabricated, or nonsensical statements while presenting them with high confidence. In recruitment, hallucinations can occur if an uncalibrated LLM fabricates candidate achievements or misinterprets job criteria, underscoring why AI scoring must be grounded in verbatim transcripts and auditable quotes.

8. Training Data

The historical datasets used to train a machine learning model. If a hiring algorithm's training data consists primarily of historical resumes from a homogeneous demographic cohort, the model will inherently replicate and amplify those biases in future candidate evaluations unless rigorous algorithmic safeguards are implemented.

9. Multimodal AI

Artificial intelligence systems capable of processing and synthesizing multiple forms of input and output simultaneously such as understanding spoken voice audio, analyzing written text transcripts, and evaluating structured numerical rubrics.

10. Semantic Resume Parsing

The automated extraction of unstructured candidate data (from PDFs, Word documents, and online profiles) into standardized, structured candidate profiles. Unlike early parsers that looked for explicit section headers, semantic parsers use Natural Language Processing (NLP) to understand context. Accurately distinguishing between a software tool a candidate used in passing versus one they actively architected for years.

11. Boolean Search

The traditional syntax-based search method utilizes operators such as AND, OR, and NOT (alongside quotation marks and parentheses) to query resume databases, search engines, and LinkedIn. While veteran sourcers still rely on Boolean strings, it is an exacting, rigid framework that frequently misses qualified candidates who phrase their experience differently.

12. Natural Language Talent Discovery

A modern sourcing interface powered by Large Language Models (LLMs) that allows recruiters to search talent databases using conversational English rather than complex Boolean strings. Sourcing tools allow a hiring manager to type, "Find senior account executives in Chicago with B2B SaaS experience who have exceeded quota," and the system automatically translates the intent into relevant candidate profiles.

13. AI Sourcing Agent

An autonomous software workflow configured to identify, verify, and engage passive candidates without continuous human prompting. AI sourcing agents operate around the clock searching public directories, verifying candidate contact information, and surfacing pre-qualified profiles directly into a recruiter's daily pipeline.

14. Predictive Talent Matching

Algorithmic scoring models that forecast a passive candidate’s likelihood to consider a new role. By analyzing career trajectories, average company tenure, company funding milestones, and organizational restructuring signals, predictive models surface candidates who are statistically more open to career discussions before they ever update their profiles on job boards.

15. Resume Padding / AI Infill

The growing candidate practice of using consumer generative AI tools to artificially tailor resumes for specific job requisitions. Job seekers use these tools to analyze job descriptions and insert corresponding buzzwords, metrics, and phrasing into their resumes, effectively gaming traditional keyword filters and generating deceptive first impressions.

16. Applicant Tracking System (ATS)

The central software system of record used by HR and recruiting departments to manage the hiring lifecycle. An ATS houses job requisitions, stores candidate records, tracks applicant stages (from application to offer), and facilitates communication between hiring managers and recruiters.

17. Job Syndication Engine

Automated multi-channel distribution technology that allows a hiring team to create a single job requisition and syndicate it across dozens or hundreds of free and premium job boards (such as Indeed, LinkedIn, and ZipRecruiter) simultaneously, consolidating all inbound applicants back into one central dashboard.

18. Asynchronous Candidate Screening

An evaluation process in which candidates complete screening questions or assessments on their own schedule without a live human interviewer present. Asynchronous screening allows candidates across multiple time zones to interview 24/7 while enabling talent teams to review evaluations on demand.

19. Conversational Voice AI

Autonomous spoken-audio systems that conduct structured, two-way conversational interviews directly in a candidate’s mobile or desktop browser. By replacing webcams with natural voice interaction, conversational voice AI removes visual performance anxiety and appearance bias while evaluating candidate verbal communication, technical domain depth, and problem-solving methodologies in real time.

20. Adaptive Follow-Up Probing

A real-time conversational intelligence capability where an AI screening platform actively listens to candidate answers and dynamically generates contextual follow-up questions. Rather than passively playing preset questions from a static script, adaptive systems probe vague statements, ask for specific operational metrics, or prompt candidates to clarify ambiguous technical claims.

21. One-Way Asynchronous Video Interview

A legacy screening format popularized in the 2010s where candidates record video responses into a blank webcam against a strict countdown timer (e.g., Spark Hire, HireVue). While it saves recruiter scheduling time, the format is widely associated with high candidate camera anxiety, impersonal applicant experiences, and elevated drop-off rates ranging between 38% and 52%.

22. Synthetic Video Avatar

An animated digital human face rendered on-screen to simulate an interviewer during an automated assessment. While designed to humanize automated interviewing, synthetic avatars frequently trigger the uncanny valley effect inducing discomfort as job seekers attempt to maintain forced eye contact with an artificial, blinking digital rendering.

23. Conversational Text Chatbot

Automated, text-based messaging interfaces that screen candidates via SMS, WhatsApp, or mobile web chat. Candidates answer untimed, structured prompts via text message. While effective for high-volume, high-turnover frontline, retail, and hospitality roles, text chatbots cannot evaluate spoken verbal communication or live technical articulation.

24. Role-Specific Scoring Rubric

A structured evaluation matrix customized by the hiring team that defines weighted competencies (such as analytical reasoning, domain expertise, conflict resolution, or communication clarity) and assigns objective scoring anchors. The AI uses this rubric to evaluate every applicant against identical criteria, eliminating the subjective inconsistency of unstructured human notes.

25. Automated Quote Extraction / Evidence Trails

A software capability that binds an AI’s numerical evaluation score directly to timestamped audio recordings and verbatim transcript quotes. Instead of forcing hiring managers to trust an opaque algorithmic score, evidence trails allow team leads to click a score and immediately read or listen to the exact statements the candidate made to justify that rating.

26. Attempt Tracking & Logging

Compliance and anti-gaming controls that log every candidate interview session and re-attempt. Advanced platforms track whether an applicant requested a retry, recording the revision history so recruiters can inspect how candidate responses evolved across sessions.

27. Frictionless Web-Link Access

A cloud delivery architecture where candidates complete screening assessments directly in their smartphone or computer browser via a secure link eliminating mandatory user account registration, password creation, or native application downloads that create funnel abandonment.

28. The 93% Screening Rule

An empirical talent acquisition benchmark demonstrating that corporate recruiters spend up to 7 hours and 47 minutes of live calendar time per open requisition conducting manual 15-to-30-minute introductory phone callswith 93% of that time expended on applicants who fail to progress past the initial qualification round. Autonomous screening platforms are deployed primarily to eliminate this administrative sinkhole.

29. Cost of Vacancy (COV)

The daily or monthly financial loss an organization incurs for every calendar day a revenue-critical or operationally vital position remains unfilled. Typically calculated as lost top-line revenue, delayed product milestones, or secondary employee overtime, corporate benchmarks show the monthly cost of an unfilled corporate vacancy ranges between €3,800 and €9,300 (roughly $4,100 to $10,200).

30. Speed Dividend

The quantifiable financial savings realized when an organization compresses its total time-to-hire. Industry recruitment telemetry indicates that reducing corporate time-to-hire by just seven calendar days yields an average of $4,000 in direct savings per hire, driven by reduced recruitment advertising spend, decreased agency reliance, and faster employee time-to-productivity.

31. Pay-Per-Interview / Consumption-Based Pricing

An elastic, software billing architecture where organizations pay strictly for the value delivered completed candidate assessments rather than committing to recurring monthly platform retainers or fixed annual contracts. Modern platforms like SonicHire offer flexible monthly subscription plans starting at $79/month for 20 interviews, $249/month for 70 interviews, or $599/month for 200 interviews, with extra interviews available at 3.95–3.00 each depending on plan. No annual contracts required, and early access customers lock in their price for life. This consumption model allows software spend to dynamically expand during hiring surges and scale down during quiet hiring quarters.

32. Per-Recruiter Seat License

The legacy software-as-a-service (SaaS) billing standard where vendors charge a recurring monthly or annual fee for each named user administrative account. While viable for centralized teams with fixed recruiter headcounts, seat licensing restricts cross-functional collaboration by requiring costly license upgrades whenever hiring managers or department leads need access to review scorecards and candidate transcripts.

33. Headcount-Based Pricing

A vendor pricing model where software licensing costs are calculated on an organization’s total employee headcount rather than actual recruitment activity or software usage. This model heavily penalizes small-to-midmarket businesses with large frontline, warehouse, retail, or manufacturing workforces, forcing them to pay enterprise-level licensing for deskless employees who never interact with the hiring platform.

34. Candidate Drop-Off / Abandonment Rate

The percentage of applicants who initiate a job application or screening assessment but exit before finishing. High candidate abandonment (often reaching 38% to 52% on legacy one-way video platforms) typically indicates excessive platform friction, such as mandatory user account creation, native mobile application downloads, or camera-induced performance anxiety.

35. Time-to-Screen Latency

The total calendar duration elapsed between the moment an applicant submits their resume and the moment their initial qualification assessment is completed and scored. While manual phone screening workflows suffer from latencies of 3 to 5 business days due to calendar scheduling tag, modern autonomous voice screening compresses this latency down to under 24 hours.

36. True-Up Penalty

A contractual clause commonly embedded in enterprise recruiting software agreements where vendors retroactively bill customers for exceeding estimated candidate screening volumes, applicant thresholds, or user seat counts during a contract cycle, frequently resulting in unexpected budget overruns.

37. Unconscious / Affinity Bias

The subtle, subconscious cognitive shortcuts and personal preferences that lead human interviewers to favor candidates who share their educational background, geographical origin, demographic profile, or personal interests. Because human phone screens and unstructured video calls are deeply vulnerable to affinity bias, standardized rubric-driven AI evaluations are increasingly used to ground early evaluations in verifiable evidence.

38. Algorithmic Bias

Systematic, repeatable errors in a computer system that generate unfair, skewed, or discriminatory outcomes against specific groups of applicants. Algorithmic bias in recruiting rarely stems from malicious code; rather, it typically originates from uncalibrated scoring logic or historical training data that reflects pre-existing human hiring imbalances.

39. Adverse Impact (The Four-Fifths Rule)

A foundational employment-law standard used by regulatory bodies (such as the U.S. Equal Employment Opportunity Commission) to determine whether a hiring practice, assessment, or selection tool disproportionately screens out protected demographic groups. Under the standard Four-Fifths Rule, a selection tool demonstrates adverse impact if the selection rate for a protected class is less than 80% (four-fifths) of the selection rate for the group with the highest pass rate.

40. Automated Employment Decision Tool (AEDT)

A specific legal classification (popularized by statutes such as New York City Local Law 144) defining any computational software, algorithm, or AI process that substantially automates, replaces, or directs candidate selection decisions. Platforms subject to AEDT statutes must undergo independent annual bias audits, maintain publicly accessible audit summaries, and provide candidates with advance notification regarding automated evaluation.

41. EU AI Act High-Risk Classification

The regulatory standard under the European Union’s landmark Artificial Intelligence Act that explicitly categorizes AI systems used in employment, worker management, and recruitment as high-risk. This classification mandates that vendors and employers implementing AI recruiting tools enforce strict data governance, maintain detailed technical logging, undergo continuous risk assessments, and establish clear human oversight mechanisms.

42. Explainable AI (XAI) vs. Black-Box AI

Explainable AI (XAI): AI systems engineered to output transparent, interpretable reasoning behind every evaluation or score. In recruiting, an explainable model links its candidate ratings directly to role-specific competencies, verifiable transcript quotes, and timestamped audio recordings. Black-Box AI: Opaque machine-learning models that generate a fit score or candidate ranking without revealing the underlying data points, feature weightings, or semantic reasoning used to arrive at that conclusion. Black-box systems carry substantial regulatory and legal liability for employers.

43. Blind Screening

The practice of systematically stripping identifiable demographic markers such as candidate names, photos, physical appearance, age, gender cues, and socioeconomic indicators from the initial qualification stage. Modern voice-first conversational AI operates as a blind screening mechanism by evaluating candidates purely on spoken dialogue, practical problem-solving, and verbal domain depth without camera exposure.

44. Human-in-the-Loop (HITL) Governance

A foundational ethical and legal framework ensuring that automated recruiting systems function strictly as decision-support tools rather than autonomous decision-makers. Under HITL governance, AI platforms handle administrative data collection, structured voice screening, transcription, and rubric-based synthesis, while final advancement, interview invitations, and employment decisions remain exclusively in the hands of human hiring managers.

45. Compliance Audit Trail

An immutable, chronologically logged digital record that captures every interaction, automated score, interviewer note, rubric modification, and candidate re-attempt across a requisition's lifecycle. A comprehensive audit trail ensures that if a hiring decision is challenged, the employer can produce auditable evidence demonstrating that the evaluation was objective, standardized, and legally defensible.

Frequently Asked Questions (FAQs)

What is the most misunderstood term in AI recruiting?

The most frequent point of confusion is the conflation of legacy ATS keyword parsing with modern semantic AI screening. Many legacy tools claim to feature "artificial intelligence" simply because they scan text documents for specific keywords or job titles.

True modern AI recruiting utilizes Natural Language Processing (NLP) and Large Language Models (LLMs) to evaluate semantic context, project depth, and transferable capabilities. In screening, it conducts dynamic, spoken conversations that actively probe an applicant's reasoning rather than matching static text strings.

Why are modern hiring teams moving away from traditional one-way video interviews?

Early asynchronous video platforms required candidates to record video responses into a blank webcam against a ticking timer. This model created significant candidate friction, driving abandonment rates between 38% and 52% due to intense camera anxiety and the perception that video clips invite appearance, age, and demographic bias.

Modern teams are replacing one-way webcam audits with conversational voice AI, which provides a natural, spoken dialogue via a browser link, eliminating visual bias and reducing performance anxiety while delivering standardized rubric evaluations.

What does Human in the Loop actually look like in everyday hiring workflows?

Human-in-the-Loop (HITL) governance means the machine manages administrative screening and data synthesis, but humans retain full decision-making authority. In practice, conversational voice AI conducts the initial 15-minute qualification screen, generates timestamped transcripts, and scores responses against an objective rubric to build a ranked shortlist.

The AI does not send automated rejection or offer letters without human approval; internal recruiters and hiring managers review the evidence, inspect the transcript quotes, and decide which candidates advance to live interviews.

How do pay-per-interview credits protect recruitment budgets compared to annual SaaS subscriptions?

Hiring needs for small-to-midmarket companies are inherently cyclical: teams experience periods of rapid expansion followed by months of quiet maintenance. Traditional software-as-a-service (SaaS) agreements require fixed annual commitments ($15,000 to $50,000+) or monthly per-seat licenses, forcing employers to pay for software that sits idle during hiring pauses.

Flexible monthly pricing models (such as SonicHire's transparent subscription plans starting at $79/month for 20 interviews, $249/month for 70 interviews, or $599/month for 200 interviews) ensure that software expenses scale in direct proportion to hiring activity, with no hidden fees or long-term contracts. Early access customers also lock in their price for life.

Conclusion

Navigating the AI recruiting landscape does not require an advanced degree in computer science. It simply requires cutting through vendor marketing claims and focusing on the core fundamentals of talent acquisition: speed, evaluation objectivity, candidate experience, and capital efficiency.

When you understand the mechanics behind semantic discovery, conversational voice screening, explainable rubrics, and consumption pricing, you can design a hiring process that screens applicants thoroughly without burning recruiter calendar hours or locking your company into rigid enterprise contracts.

Modernize Your Screening Funnel with SonicHire

Eliminate the 93% calendar waste of manual introductory phone screens and say goodbye to the high drop-off rates of awkward webcam recordings.

With SonicHire, candidates complete an adaptive, conversational voice interview directly in their mobile or desktop browser. No apps to download, no logins to manage, and no webcams required.

• Deploy in under 10 minutes: Generate role-specific rubrics straight from your job description.

• Real-time adaptive probing: Voice AI listens and asks dynamic follow-up questions to explore candidate depth.

• Instant ranked shortlists: Review candidate scores, transcripts, and audio recordings the moment an interview finishes.

• Transparent monthly pricing: $79/month for 20 interviews, $249/month for 70 interviews, or $599/month for 200 interviews.

• Extra interviews available: 3.95–3.00 each depending on the plan.

• No annual contracts or hidden fees. Early access customers lock in their price for life.

• Free 14-day trial: 10 interviews, no credit card required.

Start Your Free 14 Day Trial with SonicHire - Get Started in Under 10 Minutes