AI interviewers, notetakers and interview intelligence: what's the difference?

Why the labels blur in 2026

Two years ago the categories were tidy. Notetakers took notes, video platforms ran one-way interviews, and the applicant tracking system held the pipeline. In the last twelve months, several of the best-known vendors bought, built or announced an AI that talks to candidates:

• BrightHire. Zoom announced it was acquiring BrightHire on 13 November 2025 and closed the deal in December. Zoom calls it an interview intelligence platform that also runs first-round screens through an AI interviewer.
• HireVue. In March 2026 it acquired Hireguide's technology and named a voice-based AI interviewer as the first milestone.
• Greenhouse. In May 2026 it agreed to acquire Ezra AI Labs, a voice AI interviewer, and now sells Greenhouse Voice AI.
• Ashby. Two days later it announced an AI Interviewer, voice and text, developed with design partners and offered through early access.
• Metaview. On 9 September 2026 it launched Metaview Screening in every workspace, with the first 25 screens free. Its product page describes a natural voice agent.

So one vendor can now be a notetaker, an analytics suite and an interviewer at once. Buyers need definitions based on what the software does in the interview, not on what the company used to sell.

Five categories, defined

AI notetaker

Software that joins a video call, records it, transcribes it with speaker labels and writes a summary. A person runs the interview; the notetaker listens. Its value is memory: feedback written from a transcript instead of from recall three days later. Its limit is that a summary is not an evaluation. Unless the notes map to the criteria for the job, you get tidier vibes.

Interview intelligence

A notetaker plus a layer of evaluation and analytics across many interviews. Typical features: scorecards drafted from the transcript, a check of which competencies were covered, talk-time and question patterns, and coaching for interviewers. A person still leads every interview. The software helps the team judge it more consistently.

AI interviewer

A voice, video or chat agent that conducts the interview itself, usually an early screen. It asks prepared questions, follows up on answers, and produces a transcript and a draft score against criteria the employer sets. Candidates can usually take it at any hour. The newest wave is voice-first: Greenhouse describes Voice AI as "a two-way voice interview using the questions and scoring criteria your team defines".

One-way video interview

The older format. The candidate records answers to fixed prompts, usually against a timer, and nobody is on the other side. Reviewers watch later, and some products score the videos with AI. There is no follow-up question, which is the main thing an AI interviewer adds. Candidates often dislike it: CareerPlug reports that 33% of job seekers have abandoned applications that required one.

AI recruiter or agentic recruiting platform

A set of agents across the funnel: finding candidates, reviewing applications, scheduling, screening and writing up interviews. Metaview, for example, says its platform deploys AI agents to automate sourcing, application review and interviewing, with people setting the criteria and making final decisions. Here the interview is one agent among several.

Side by side

The laws column deserves a closer look. The EU AI Act lists AI used "to evaluate candidates" as high-risk, with the main duties now due from 2 December 2027 under the Digital Omnibus. The duty to tell people they are interacting with an AI system applies from 2 August 2026, and inferring emotions in the workplace has been banned since 2 February 2025.

In New York City, Local Law 144 requires a bias audit within a year before an automated employment decision tool is used, a public summary of the results, and notice to candidates. Illinois requires notice, an explanation and consent before AI analyzes a recorded video interview. Our guide to AI interview laws covers the rest.

What candidates notice

Candidates do not care about category names. They care whether they were told, whether they were heard and whether anyone replied. In Greenhouse's May 2026 survey, 63% of job seekers had faced an AI interview, 70% were not clearly told upfront that AI would evaluate them, and 38% had left a process because it included one.

The top reason for walking away was a pre-recorded, AI-scored video with no human present (33%). That is a warning for one-way video, and for any AI interviewer that feels like one. The safeguard candidates asked for most was the option to request a human interview (46%). We cover the rest of the evidence in what candidates really think about AI interviews.

The frame we believe in: one rubric, whoever runs the interview

The research on interviews is about structure, not the medium. In the largest re-analysis of selection methods, structured interviews predicted job performance at .42, against .19 for unstructured ones. Structure means the same job-related questions, scored against a written scale. An AI interviewer with no rubric is an unstructured interview at scale. A human interview with a good rubric is a structured one.

So the category you buy matters less than the rules you hold it to:

1. Write the rubric first. Four to six competencies per job, each with an anchored scale. Every interview, AI-led or human-led, scores against it.
2. Keep the evidence. Every score should point to the moment in the transcript that earned it, so anyone can check it.
3. Compare like with like. A candidate screened by an agent and one screened by a recruiter should reach the debrief with comparable scorecards.
4. Name the decider. A person reviews the evidence and makes the call, and the record shows who it was.
5. Tell candidates, and offer a person. Say when AI is involved, what it evaluates, and how to choose a human instead.

We call software built on these rules a structured interview platform. It is a frame rather than a sixth category. Notetakers, interview intelligence and AI interviewers can all sit inside it, as long as the rubric stays constant and a person stays accountable.

Seven questions to ask any vendor

1. Can AI-led and human-led interviews for one job use the same rubric and scale?
2. Does every score link to a quote or timestamp in the transcript?
3. What happens by default after a score: a recommendation to a person, or an automatic action? Can anyone be rejected without a person reviewing it?
4. Does it analyze faces, voice tone or emotion? In the EU, inferring emotions at work is banned.
5. How is the candidate told, and can they choose a person without penalty?
6. Has the tool had a bias audit, and is the summary public?
7. Where are recordings stored, for how long, and who are the subprocessors?

A vendor that answers these clearly, in writing, is easier to defend to candidates, to your legal team and to the hiring manager who has to live with the decision.

Where Itya fits

Itya is a structured interview platform. Today it sends a notetaker into Google Meet, Microsoft Teams and Zoom, runs its own browser interview room, and holds one rubric per job. Scorecards and debriefs are drafted with a link to the transcript moment behind each claim, and candidates give consent before recording. By default, Itya's AI recommends and a named person confirms; it never rejects a candidate automatically. The product tour shows how it works.