Quick Answer: AI recruiting has its own dialect. This glossary defines the 40+ terms TA leaders, recruiters, and hiring managers use in 2026, from AI-powered ATS and headless ATS to knockout questions, reqs, and AI phone screens. Use it to cut through vendor jargon and figure out what your hiring stack actually does.
Every vendor in AI recruiting invents a new word for the same thing. "Conversational screening." "Candidate agent." "Talent orchestration." A recruiter hears five pitches in a week and ends up with five different names for the same feature. This glossary is the version we wish existed when we started selling to TA leaders. Definitions are written in the language recruiters use, not the language that shows up on a G2 listing.
We grouped the terms by where they sit in the process: the people, the systems, the screening layer, the AI layer, and the metrics. Skim the ones you need. Bookmark the page for the next vendor call.
The People
Recruiter
The person who owns the requisition day-to-day. Posts the job, sources candidates, runs phone screens, schedules interviews, and shepherds the offer. In high-volume hiring, one recruiter covers 40 to 80 open roles at a time. In executive search, one recruiter covers 2 to 5.
Sourcer
A recruiter focused only on the top of the funnel. Sourcers find candidates who haven't applied yet and get them to apply. At high-volume shops, sourcing is often automated or outsourced. AI sourcing tools and AI recruiting platforms now do most of what junior sourcers used to do.
Hiring Manager
The person the new hire will report to. Writes the job description, interviews finalists, and makes the call. Recruiters coach hiring managers through the process. Hiring managers coach recruiters on what "good" looks like. When the two are out of sync, time-to-fill stretches and candidates ghost.
TA Leader
Talent Acquisition leader. Head of Talent, VP of TA, Director of Recruiting. Owns the recruiting team, the tech stack, and the hiring plan. The person a vendor has to win over to close a deal above $50K.
People Ops / HR Business Partner
Adjacent to recruiting but not the same. HR owns the relationship after the offer is signed. TA owns the relationship before.
Recruiting Operations (RecOps)
The behind-the-scenes function that owns recruiting tools, workflows, and reporting. Picks the ATS, manages integrations, runs the dashboards, sets the SLAs. A small team can survive without RecOps. A 50+ recruiter team can't.
People Analytics / Talent Analytics / TA Analyst
The analyst who turns recruiting data into decisions. Sometimes lives inside TA, sometimes inside HR, sometimes inside Finance. Job titles vary: People Analytics Manager, Talent Insights Lead, Workforce Analytics Analyst, TA Operations Analyst. The work is the same. Build the funnel report, find the leaks, model the hiring plan, justify the budget. With AI screening generating structured transcripts and scorecards on every candidate, this role just got a lot more data to work with.
Employee Experience (EX)
The function responsible for what it feels like to work at the company. Onboarding, internal communications, manager training, engagement surveys, exit interviews. EX picks up where TA hands off. The candidate experience and employee experience are the same arc viewed from two sides.
Employer Brand / Employer Branding
The team or person who owns how the company shows up to candidates. Career site, social content, Glassdoor responses, recruitment marketing campaigns. Common titles: Employer Brand Manager, Recruitment Marketing Lead, Talent Brand Specialist. When TA and Employer Brand are aligned, the funnel runs warmer. When they're not, the recruiter pitches a story candidates don't see anywhere else online.
Candidate
The person applying for the job. "Applicant" when they've submitted. "Prospect" when they haven't. "Finalist" when they're in the last round.
The Systems
ATS (Applicant Tracking System)
The database recruiters live in. Every applicant, every stage, every note. The ATS is the system of record for hiring. Common ATS platforms: Workday, Greenhouse, Lever, iCIMS, BambooHR, Bullhorn (for staffing), JazzHR, Indeed Smart Sourcing.
The ATS does four things well: collect applications, track candidates through stages, store notes, and produce reports. Everything else, it does poorly.
AI-Powered ATS
An ATS with generative AI features bolted on. Resume parsing, job description drafting, candidate summaries, interview question suggestions, auto-rejection emails. The "AI" in AI-powered ATS usually means "we added a ChatGPT wrapper to the notes field." Some are deeper than that. Most aren't.
The useful features in an AI-powered ATS: automated screening questions, candidate ranking against the req, and writing job descriptions from a prompt. The less useful features: every AI-generated summary of a candidate you could have read in 30 seconds.
Headless ATS
An ATS that exposes its data through an API instead of forcing you into its UI. You bring your own front-end, your own candidate experience, your own career site. The ATS becomes the database in the background.
Headless is attractive for companies who want to control the candidate experience. Apploi, Ashby, and Manatal all ship API-first workflows. The downside: you need engineering to build the front-end. Most recruiting teams don't have that budget.
Career Site
The public-facing page where candidates find jobs. Lives at yourcompany.com/careers. Powered by the ATS or a standalone job site builder. The career site is the first candidate experience most applicants have with your brand. Most are broken.
Candidate Portal
The logged-in experience candidates see after they apply. Shows status, schedules interviews, uploads documents. Used more in enterprise. Rare in high-volume hiring, where candidates apply on Indeed and never see a portal.
HRIS
Human Resources Information System. Where employees live after they're hired. Workday, ADP, Rippling, BambooHR. The HRIS and the ATS are usually different products, and getting them to talk to each other is one of the hardest parts of the recruiting tech stack.
The Job
Requisition (Req)
An approved, budgeted, open role. "We have 12 open reqs" means the company has 12 approved positions to fill. Reqs get opened, put on hold, reopened, and closed. The req number is the ID that follows a role through the ATS.
Job Description (JD)
The document that describes the role. Responsibilities, requirements, qualifications, compensation (sometimes). Recruiters and hiring managers argue about the JD more than anything else. Most JDs are copy-pasted from the last open role. AI can draft a first pass in 30 seconds.
Job Posting
The public version of the JD. Lives on the career site and syndicates to Indeed, LinkedIn, ZipRecruiter, and job aggregators. The posting is optimized for search. The JD is optimized for legal.
Pipeline
Every candidate in consideration for a specific req, organized by stage. A healthy pipeline has candidates at every stage. A broken pipeline has 80 applicants stuck in "new" and zero at "offer."
Pass-Through Rate
The percentage of candidates who move from one stage to the next. If 100 people apply and 20 get phone screens, the application-to-phone-screen pass-through is 20%. Benchmarking pass-through rates is how recruiters find leaks in the funnel.
The Screen
Screen / Screening
The step where a recruiter decides who moves forward. Traditional screening means reviewing a resume. Modern screening means a 5-minute phone call or an AI-run interview.
Phone Screen
A short call, usually 15 minutes, where a recruiter confirms the basics. Availability, compensation expectations, why they're looking, relevant experience. The phone screen is the most valuable 15 minutes in recruiting, and the hardest to scale. See our phone screening template for a structured approach.
AI Phone Screen
A voice AI that calls candidates and runs the phone screen automatically. No scheduling, no recruiter on the line, no calendar back-and-forth. The candidate picks up, answers 5 to 8 questions, and the AI generates a summary, scorecard, and transcript.
The honest trade-off: AI phone screens scale infinitely but lose the recruiter's judgment. You get a transcript and a score. You don't get the recruiter's read on tone, energy, or "something felt off." For high-volume, hourly, and skilled trades hiring, that trade-off is usually worth it. For executive search, it isn't. Classet's Joy runs AI phone screens in under 2 minutes, and completion rates land around 80%.
Knockouts / Knockout Questions
The must-have criteria that auto-qualify or auto-disqualify a candidate. "Do you have a valid CDL?" "Can you lift 50 lbs?" "Are you authorized to work in the US?" A "no" is a knockout. Candidates who fail knockouts don't get a recruiter's time.
Good knockout design is the difference between a recruiter screening 10 applicants and 80. Bad knockouts filter out qualified candidates on technicalities.
Video Interview (One-Way)
Asynchronous video. The candidate records answers to prompts on their phone or laptop. Popularized by HireVue. Candidates hate them. Completion rates run 40% to 60%, which means half the pipeline evaporates before anyone sees them.
Chat Screen
A chatbot that asks screening questions in a text thread. Paradox built a business on this. Works well for candidates who prefer text. Works poorly when the role needs voice assessment, like a dispatcher, a customer-facing technician, or anyone who'll spend the day on the phone.
Conversational AI
Vendor shorthand for "our chatbot has a personality." Sometimes real, sometimes marketing. The test: ask the bot a question that isn't in its flowchart. If it handles it, the "conversational" label is earned. If it loops back to "I didn't understand that, please select from the menu," it isn't.
The AI Layer
Voice AI
AI that talks. The model listens to the candidate speak, generates a spoken response, and carries a back-and-forth conversation. Voice AI is the engine behind AI phone screens. Quality varies: some sound like robots reading a script, others are nearly indistinguishable from a recruiter.
LLM (Large Language Model)
The underlying model that powers generative AI features. Claude, GPT-4, Gemini. When a vendor says "AI recruiter," they usually mean "LLM with a recruiting prompt on top." The quality of the recruiting product depends on how the prompt is engineered, what data it pulls from, and how it handles edge cases.
Claude Code
Anthropic's CLI for developers. Not a recruiting product. If your vendor is pitching Claude Code as a recruiting feature, they are confused. What they probably mean: Claude (the model) is doing the reasoning, and the "Code" part is irrelevant to you.
Claude Skills
Anthropic's framework for giving Claude specialized capabilities through instructions and context. Skills are how you get a general-purpose model to act like a recruiting specialist, a phone screener, or a copywriter. In a recruiting context, a "skill" could be "run a phone screen for a CDL driver" or "draft a job description for an HVAC tech."
The difference for recruiters: Claude Code is a tool engineers use to build software. Claude Skills is the pattern that lets a product like Classet turn a general model into a phone-screening specialist. You interact with the skill, not with Claude Code.
Candidate Agent
A vendor term for an AI that handles a candidate end-to-end. Greets them, screens them, schedules them, answers their questions, nudges them if they ghost. The real ones save recruiters hours a week. The fake ones are glorified chatbots with better marketing.
Resume Parser
Extracts structured data (name, email, phone, work history, skills) from an unstructured resume file. Every ATS ships one. Most are bad at parsing non-standard formats, PDFs with images, or resumes in a non-English language.
Candidate Matching / Ranking
The AI ranks applicants against a req based on resume fit, answers to screening questions, and sometimes historical data on who got hired and stayed. Useful when there are 500 applicants and the recruiter has an hour to review them. Less useful when the matching logic is a black box and the recruiter can't tell why one candidate ranked above another.
Bias Audit
A third-party review of whether an AI screening tool produces disparate outcomes by race, gender, age, or disability. Required in NYC under Local Law 144. Done by firms like Warden AI. Classet's bias audit is public.
The Integration Layer
ATS Integration
A connection between your ATS and another tool. Candidates flow in, status updates flow back, and recruiters don't have to copy-paste between systems. The good integrations are bi-directional. The bad ones push data one way and break every time the ATS ships an update.
ATS Sync
Classet's term for the real-time data flow between our AI screening and your ATS. Candidates get screened inside Classet, results push back to Workday, Greenhouse, or whatever you run. Recruiters see scorecards and transcripts in the ATS without leaving it.
Webhook
A callback URL an app hits when something happens. "Candidate completed an interview" fires a webhook, and your ATS updates their stage automatically. Webhooks are how most recruiting tools talk to each other behind the scenes.
API
Application Programming Interface. The technical contract between two systems. If a vendor has a good API, you can build custom workflows. If they don't, you're stuck with whatever they ship.
The Metrics
Time-to-Hire
Days between a candidate applying and accepting an offer. Lower is better. The benchmark for high-volume hourly is 3 to 7 days. For skilled trades, 7 to 14. For salaried professional roles, 30 to 45.
Time-to-Fill
Days between opening a req and filling it. Different from time-to-hire, which starts at application. Time-to-fill includes sourcing lag, whereas time-to-hire starts when the candidate is in the pipeline.
Cost-Per-Hire
Total recruiting spend divided by hires. Includes ad spend, recruiter salaries, vendor tools, and referral bonuses. The number TA leaders report to finance. A useful sanity check. A poor measure of quality.
Completion Rate
The percentage of candidates who finish a screening step. For video interviews, 40% to 60%. For AI phone screens, 75% to 85%. For human phone screens, 60% to 75% (a lot of no-shows). Completion rate is the single most overlooked metric in recruiting.
Ghosting
When a candidate stops responding. Applies, gets an interview invite, and disappears. 40% of hourly applicants ghost after applying. The faster you get to them, the less they ghost. Instant AI phone screens cut ghosting dramatically because the candidate is still thinking about the job when they apply.
Drop-Off
The percentage of candidates who quit the application partway through. Long applications have drop-off rates above 50%. Short applications under 5 fields have drop-off rates under 10%.
Offer Acceptance Rate
The percentage of offers that turn into signed agreements. Below 80%, something is wrong. Either your offers are low, your process is too long, or candidates are getting better offers elsewhere while you deliberate.
Quality of Hire
The hardest metric to measure. How well did the hire perform? Did they stay 90 days? 12 months? Did their manager promote them? Every TA leader claims to track it. Few actually do.
How AI Screening Changes the Process
The old funnel: recruiter posts the job, recruiter reviews resumes, recruiter calls candidates, recruiter schedules interviews, recruiter writes rejection emails. Five hours of a recruiter's day on front-of-funnel work that doesn't require judgment.
The new funnel: job gets posted, applicants flow in, AI calls them within 5 minutes, knockouts auto-filter unqualified candidates, the AI sends a scorecard and transcript to the ATS, and the recruiter opens their morning to a pre-screened shortlist.
The recruiter's job doesn't disappear. It changes. Instead of screening 80 people to find 10 who pass, they spend time on the 10, build relationships with the hiring manager, and close the candidates who matter. The busy work drops away. The judgment work scales up.
What that looks like in practice:
- Monday morning: recruiter opens the ATS and sees 14 new applicants, all screened overnight by AI, ranked by fit, with full transcripts.
- Tuesday: recruiter runs 4 human interviews with the top candidates. No scheduling back-and-forth, because the AI already booked them.
- Wednesday: hiring manager reviews scorecards, picks 2 finalists, and moves them to onsite.
- Thursday: offer out.
That week used to take three weeks. The difference isn't that AI is smarter than the recruiter. It's that the recruiter isn't doing 40 hours of screening to get to a short list.
FAQs
What's the difference between an AI-powered ATS and an AI screening tool?
An AI-powered ATS is the system of record with AI features added. A screening tool like Classet is a specialist product that plugs into your ATS and handles the screening layer. Most teams run both: ATS for tracking, screening tool for the interviews.
Do knockout questions hurt candidate experience?
Only if they're bad knockouts. A question like "Do you have a valid CDL?" for a CDL role is fair. A question like "Are you willing to work 80-hour weeks?" filters out good candidates who'd work hard but have families. Design knockouts around job-required facts, not your cultural wish list.
How do I compare AI phone screens to video interviews?
Completion rate, candidate preference, and cost. Phone screens complete at 75% to 85%, video at 40% to 60%. Candidates prefer phone because everyone already knows how to answer one. Video adds friction. We wrote a full comparison here.
Is headless ATS worth it for a 50-person recruiting team?
Usually not. Headless ATS assumes you have engineering to build the front-end. A 50-person team is better off with an off-the-shelf ATS and a screening tool that integrates with it.
What does a People Analytics or TA Analyst actually do day to day?
Build the recruiting funnel report, find the stage with the worst pass-through, model the hiring plan against headcount targets, and tell the TA leader where to spend the next dollar. With AI screening, they also analyze interview transcripts to spot which questions predict good hires.
How does Employer Branding work with AI recruiting?
Employer Brand owns the story candidates see before they apply. AI recruiting owns what happens after. The handoff matters: if your career site promises a fast, modern process and the candidate gets a 4-week silence, the brand work is wasted. Align the two and conversion climbs.
What's the fastest way to measure if AI screening is working?
Look at two numbers: time from application to first screen, and completion rate on that screen. If AI is working, the first drops from 3 days to under 5 minutes, and the second holds at 75%+. If either number is off, the AI isn't delivering.
Key Points
- AI recruiting terms get invented by vendors faster than anyone can track. A shared glossary makes vendor calls shorter.
- Knockouts are the highest-impact piece of AI screening. Design them around facts, not opinions.
- AI phone screens complete at 75% to 85%. Video interviews complete at 40% to 60%. That gap is the biggest hidden cost in most hiring funnels.
- Claude Code and Claude Skills are different things. A recruiter interacts with Skills, not Code.
- AI doesn't replace the recruiter. It removes the 40 hours of screening that blocked the recruiter from doing actual recruiting.
Next Steps
Want to see what an AI phone screen sounds like? Try a live example or book a walkthrough. We'll show you how Joy runs screens, where she hands off to your team, and how the transcripts flow into your ATS.