85% of Employers Record Interviews. Answer for the AI Scorecard Draft

Your interview confirmation says the call will be recorded by BrightHire or Metaview, and there's an opt-out link at the bottom. That link is the least important part of the email. What matters is whether the AI draft your interviewer turns into a scorecard still shows what you actually did.

Recording is now the normal setup. Kyle & Co's benchmark report on interview intelligence, released Oct. 14, 2026, found that 85% of employers already run some kind of interview intelligence tool. So skip "should I opt out" and ask a better question: what makes it into the notes?

Should I opt out of interview recording?

For most candidates, no. If you opt out, you join a 2-4% minority, and instead of a full transcript, the panel gets whatever your interviewer could type while also trying to listen.

BrightHire says that across Greenhouse customers, about 2-4% of candidates opt out of recording. Its pricing page puts the range at "fewer than 1-4%." Employers that publish candidate FAQs (eBay, SoFi, Sourcegraph, Pariveda) describe the same fallback. In eBay's words: "If you choose not to be recorded, interviewers will take notes manually instead."

That fallback costs you. An interviewer typing live notes catches the gist of your answers and loses the details, and numbers are the first to go. A recruiter on Blind said recording gives you a better shot if you're "in a technical role and have to share a lot of specific info." Your 40% latency cut turns into "improved performance" in someone's rushed notes.

You also don't opt out one interview at a time. BrightHire's help center says that if you opt out before your first interview, none of your interviews will be captured. That covers the whole loop.

When opting out is still reasonable

• You can't get a straight answer on data use. eBay says it doesn't let BrightHire use recordings to train AI. If an employer can't tell you the same, it's fair to ask, and fair to decline.
• You'll need to discuss sensitive current work in detail. If you can't describe a project without naming unreleased work, a recording raises the stakes of a slip.
• You've weighed it and you'd simply rather not. That's allowed. If you go this way, speak more slowly and repeat your key numbers so the interviewer can write them down.

One more point in favor of recording: it protects you too. The same Blind interviewer said they'd heard of a manager reviewing a recording after a candidate complained about a poor interview experience. Without a recording, a complaint like that comes down to one person's word against another's.

What Kyle & Co's report actually found

Nearly every employer uses interview intelligence, and few of them analyze what it captures. 85% of employers run some kind of interview intelligence tool, 97% are at least piloting one, and only 14% analyze that data or check it against how hires later perform.

Interview intelligence is software that records, transcribes, and summarizes interviews, then feeds those notes into the hiring team's scorecards. BrightHire, Metaview, and Pillar are three of the vendors in the category.

The report's author is Matt Charney, principal analyst at Kyle & Co, the firm founded by Kyle Lagunas. Read the sample carefully. Of the 431 respondents, 257 were employers and 174 were job seekers, surveyed in August and almost all based in the U.S. It's mostly an employer survey, and it shows a large gap between how employers see their interviews and how candidates experience them.

Sources: Kyle & Co via Globe and Mail coverage (rows 1-3); BrightHire (row 4).

Why the 37-point gap on structure? Employers build a rubric and believe that makes the interview structured. Candidates often never see the rubric: nearly one in four said they got a vague explanation, or none, of how they'd be evaluated. The company runs a structured process while the candidate experiences a conversation, and those two get scored differently.

The 14% figure also tells you who reads your interview. At most companies, no one is analyzing patterns across hires. The reader is a tired interviewer writing a scorecard, followed by a panel skimming summaries in the debrief.

How interviewers use interview transcripts

Interviewers increasingly write their feedback from the transcript instead of from memory. With tools like Metaview, the scorecard starts as an AI draft built from the call, and the interviewer reviews it and submits.

The Blind thread says this directly. One interviewer wrote that BrightHire "makes a transcript of the interview that I use to write my interview feedback," and said they get annoyed when candidates opt out.

Metaview describes the same process from the vendor's side. It turns calls into "structured notes mapped to your rubric," then "drafts the scorecard for the interviewer to review and submit." The interviewer's job is to "add the observations nobody said out loud."

That ratio is the whole argument. When about four in five scorecards start as a draft, the easy move for a busy interviewer is to tidy the draft and approve it, not to add things from memory. Whatever the draft leaves out probably stays out, and the draft only knows what you said.

Interview intelligence AI notes follow the rubric, not your story

Tools like Metaview don't summarize your interview in the order things happened. They file what you said under the scorecard's competencies, so an answer that never names or clearly matches a competency may not get filed anywhere.

A rubric is the list of competencies an interview round scores, such as "stakeholder management," "technical depth," or "ownership," with a description of what strong looks like for each. Metaview picks its suggested video snippets by "reading the role's scorecard questions and the job description." So your words get matched against two documents: the internal rubric and the public posting.

That changes how you should finish your stories. Compare these two:

• Unlabeled: "So we got everyone in a room and eventually shipped it."
• Labeled: "That's how I handle stakeholder conflict: I got finance and product to agree on one metric, and we shipped two weeks later."

Both describe the same event. Only the second gives the AI a competency name, a first-person action, and a result it can file in the right section.

The job description is the half of the rubric you can actually see, so read it like an answer key. Pull out the five or six skills it keeps repeating and match each one to a story. Refolk does that mapping for your resume: paste the posting, and you get your own history rewritten against that posting's criteria. You walk in already knowing which project answers "ownership" and which answers "technical depth."

What the transcript can't see

The transcript only captures what you say. Anything you draw, point at, type silently, or show on screen without describing it is invisible to the AI, and Metaview leaves it to the interviewer to add those things back.

With roughly 78% of scorecards starting as drafts, don't count on that happening. These are the things that tend to drop out:

1. Whiteboard diagrams. The transcript catches "so this goes here, and then this connects to that," which means nothing without the drawing.
2. Screen shares and portfolio walkthroughs. A designer clicking through a prototype in silence leaves a gap in the record.
3. Silent coding. Four minutes of typing is four minutes of nothing in the transcript.
4. Nods and gestures. "Yeah" and "mm-hm" don't show agreement with a design tradeoff.
5. Pointing to an attachment. "It's on page three of my deck" keeps the numbers on page three out of the notes.

The fix is to narrate what you're doing. Say what you're drawing and why: "I'm putting a cache in front of the read path because 90% of traffic is reads." Say the number on the slide out loud. When you go quiet to think, say so, then sum up what you decided.

Recorded job interview tips: talk for the transcript

Answer as if the panel will only read your interview, because many of them will. Lead with the result, say the number, use "I," and name the competency.

1. Result first. Open with the outcome, then explain how you got there. The panel is skimming summaries, so put the important part where a skimmer will see it.
2. Say the numbers, with units. "Cut onboarding from 14 days to 5" makes it into a summary. "Made onboarding a lot faster" doesn't.
3. "I," not "we." A transcript full of "we" gives the AI no way to tell what you did. Say "we" for the team, then "I" for your part.
4. Name the competency. "This is an example of how I prioritize" tells the AI where to file the answer.
5. Close each answer with one sentence. "So the short version: I owned the migration, and it saved $200K a year." That hands the interviewer a ready-made line for the scorecard.

Here's what the change looks like in practice:

Most of your best spoken answers are resume bullets you've already written, said out loud. A bullet with a number in it is a sentence that will make it into the transcript. If your bullets are vague, your answers will be too. Refolk's fit score shows where your history is thin against a specific posting, so you know which competency needs a rehearsed story before the recording starts.

Ask for the rubric before you sit down

Ask the recruiter which competencies the round scores. The answer tells you how the AI notes will be organized, and almost a quarter of candidates say nobody told them how they'd be judged.

The person to ask is usually whoever sent the recording notice. Recruiters and coordinators are the people who typically set up these tools and send the opt-out emails.

Some simple ways to ask:

• "What competencies does this round score?"
• "Is there a particular project or skill the panel wants to hear about?"
• "Who will be reading the feedback from this round?"

Asking costs you nothing, and even a bare list of competency names is enough to work with. Map one story to each, and the AI draft will have a clear answer under every heading.

FAQ

Does opting out of BrightHire recording hurt my chances?

The research doesn't show a direct penalty, and employers like eBay say interviewers take notes manually instead. The indirect cost is real, though. You're in a 2-4% minority, an interviewer on Blind said opt-outs annoy them because they write their feedback from the transcript, and typed notes lose detail, especially numbers. If you opt out, speak slowly and repeat your key figures.

Will my interview recording be used to train AI?

That depends on the employer. eBay states in its public FAQ that it doesn't let BrightHire use recordings to train AI, and SoFi, Sourcegraph, and Pariveda also publish candidate FAQs on BrightHire. If the company you're interviewing with hasn't said, ask the recruiter in writing before the interview.

Is this only a BrightHire and Metaview thing?

No. Interview intelligence is a whole category, and Pillar is another vendor in it. The details differ, but the basic mechanics are the same: record, transcribe, map what you said to a rubric, and draft notes for the interviewer. The advice holds whichever tool is on the call: say it out loud, give the number, and name the competency.