# Interview Copilots Cost Up to $299 a Month. The Follow-Up Breaks Them

_AI interview copilots cost $55 to $299 a month, and recruiters spot the reading. Build a one-page evidence sheet that survives the follow-up question._

- Canonical URL: https://www.refolk.ai/candidates/blog/interview-copilots-follow-up-question-evidence-sheet
- Published: 2026-11-12
- Reading time: 9 min
- Keywords: AI interview assistant, how recruiters detect AI in interviews, LockedIn AI interview, is using AI during an interview cheating, how to prepare for follow-up interview questions

You can pay $55 to $299 a month for a copilot that listens to the interviewer and puts a suggested answer on your screen. It will write a decent first answer. It can't write the second one, because the follow-up asks about things only you remember.

Recruiters have noticed. The Globe and Mail reports that they can see candidates' eyes move across the screen as if they're reading, and that the answers sound robotic and even scripted. The piece names LockedIn AI. A better-hidden copilot won't fix this. A one-page evidence sheet built from your own resume will, because it gives you the second and third answer for free.

## What does an AI interview assistant actually do, and what does it cost?

An AI interview assistant listens to the interviewer's question, transcribes it, and shows a suggested answer on your screen while you're still on the call. Listed plans run from about $55 to $299 a month, and LockedIn AI's Stealth Mode, which hides the app during a screen share, needs a paid plan.

| Tool / plan | Listed price | What you get |
|---|---|---|
| LockedIn AI, cheapest unlimited | $54.99 to $119.99/mo | Unlimited sessions |
| LockedIn AI, Unlimited Professional | $239.98/mo (from month two) | Unlimited plus Stealth Mode |
| Interview Coder Pro | $299/mo | Competitor copilot |
| Interview Coder Lifetime | $799 one-time | Competitor copilot |

LockedIn's own materials don't agree on its prices. One review found official sources quoting $69.99, $54.99 and $41.24 a month, and found that monthly and quarterly plans are non-refundable. Interview Coder and Cluely compete for the same buyers.

LockedIn's founder and CEO, Caesar Gui, says the company's mission is "to level the playing field for applicants." The pricing points the other way. Stealth sits behind the higher tiers and the plans can't be refunded. The people most likely to pay $239.98 a month to hide a tool are the most anxious candidates in the market. If it fails in the one interview that mattered, the money is gone.

## How do recruiters detect AI in interviews?

Mostly by watching and listening, not with software. The Globe and Mail's sources describe eyes moving across the screen as if reading and answers that sound robotic and scripted. One said many recruiters end the interview once they have a strong feeling the candidate is using AI.

The signals that give it away:

- **Reading eyes.** A steady sweep across the screen looks nothing like someone thinking.
- **Uniform lag.** The tool has to hear the question before it can draft anything, so every answer starts after the same delay.
- **Scripted cadence.** The answer has neat structure and no hesitation or self-correction. That's how written text sounds when someone reads it aloud.
- **Missing texture.** No names, no awkward details, nothing that sounds like it came from someone who was actually there.

Some recruiters now deal with it up front. One source in the Globe piece advises asking candidates at the start of the call to confirm that answers are "being delivered personally from the candidates' own recollection" and without AI. Keep the word *recollection* in mind, because the rest of this piece depends on it.

Usage is common. A Resume Genius survey of 1,000 active US job seekers found 22% already use AI during real-time interviews. Flags are common too.

> **38.5%** of candidates flagged for AI-assisted behaviour
>
> Greenhouse analysed 19,368 live interviews between July 2025 and January 2026.

The two figures come from different samples, so you can't subtract one from the other. But a flag rate that close to the admitted usage rate suggests some ordinary nervous candidates who glance at their notes are getting flagged along with copilot users. You don't avoid the flag by hiding better. You avoid it by answering from memory, with details nobody could have scripted for you.

## Why does the follow-up question break a script when detection software can't?

The follow-up breaks the script because the copilot only knows what's in your resume and the question it just heard. It can answer "tell me about a time you failed" with a believable story. It can't answer "what did your manager say the next morning?", because only you know that.

Each follow-up also restarts the tool. It transcribes the new question, generates a new answer, and you read it, so the lag and the reading happen again every round. One scripted answer can slip by. Three in a row, each a little more specific, is a pattern.

Who asks those questions matters too. In Refolk's index of professional profiles, 122,621 US profiles hold the title Recruiter, Technical Recruiter or Talent Acquisition Partner. Only 381 of them list "Behavioral Interviewing" as a skill.

| Segment (Refolk index, US) | Count |
|---|---|
| Recruiter, Technical Recruiter or TA Partner | 122,621 |
| Of those, tagged "Behavioral Interviewing" | 381 |
| Share with that tag | ~0.31% |

Skill tags are self-reported, so plenty of recruiters who run good behavioral screens never added the tag. Still, the number tells you something. Recruiters mostly check whether you qualify. The hard probing ("how did you measure that?", "what would you change?") tends to come from hiring managers, who know the work well enough to dig. That's an inference, not a measured fact. If it holds, the copilot is weakest in the round that decides the offer.

> A copilot can write your first answer. It can't remember what your manager said the next morning.

## Does Stealth Mode solve the problem?

No. Stealth Mode hides the copilot's window during a screen share. It doesn't hide your eyes, your timing, or how your answers sound, and those are what the Globe's recruiters said they noticed.

It may not even hide the window. Interview Coder reports that the most common Trustpilot complaint about LockedIn is that the other side was "able to see the screen," and that the browser extension is visible to interviewers. Interview Coder is a competitor, so treat that with some skepticism. But even if stealth worked perfectly, the $119.99 to $239.98 tiers only protect the screen share, and the interviewer is mostly watching your face.

## Is using AI during an interview cheating?

If the employer asks you to confirm your answers come from your own recollection and you're reading from a copilot, yes. You've broken a condition they set. If nobody asks, it's still a misrepresentation, because the employer is judging what you can do on your own, and that's how you'll do the job.

The practical risk is bigger than the ethical one. Even if the copilot gets you through the video round, there's often an in-person round after it.

> **72%** of recruiting leaders run at least one in-person stage to counter AI-assisted fraud
>
> Google, McKinsey and Cisco have reintroduced mandatory face-to-face rounds.

Sundar Pichai said Google would introduce "at least one round of in-person interviews for people, just to make sure the fundamentals are there." So a copilot doesn't remove the risk of failing. It moves the failure to a later, more expensive round. If you're going to prepare anyway, prepare something that still works when you're sitting across a table.

## How to prepare for follow-up interview questions: the one-page evidence sheet

An evidence sheet is a single page that lists every claim on your resume next to the facts a follow-up question will go after. You build it before the interview, rehearse it out loud, and then put it away. Its job is to load your memory, not to sit on screen as a script.

Start with the resume you're actually sending for this role. If it's generic, you'll prepare for the wrong questions. [Refolk](https://www.refolk.ai/candidates) writes your resume from your own history and tailors it to each posting, so the claims the interviewer reads are the ones that matter for that job, and each one came from something you actually did.

### Build one row per claim

Each bullet on your resume becomes a row. Fill in these fields:

1. **The claim.** Copy the bullet exactly as written. The interviewer is reading this version.
2. **The number.** The figure in the bullet, or the closest one you can honestly back up.
3. **Baseline and timeframe.** What it was before, what it was after, and over how long.
4. **Your part vs. the team's part.** Which pieces you did yourself and which the team did. This answers "what did *you* do?", the follow-up that catches the most people.
5. **How it was measured.** The dashboard, report or tool the number came from, and who checked it.
6. **What went wrong.** One real problem: a missed deadline, a wrong assumption, a pushback you lost.
7. **What you'd change.** A specific thing you'd do differently, not "communicate more."

### What each field protects you from

| Field | Follow-up it answers |
|---|---|
| Baseline and timeframe | "Compared to what?" |
| Your part vs. the team's | "What was your specific role?" |
| How it was measured | "How did you measure that?" |
| What went wrong | "What broke?" |
| What you'd change | "What did you learn?" |

A copilot can't fill in fields 3 through 7. It doesn't know your baseline, it doesn't know a teammate built the pipeline, and it doesn't know why the launch slipped. Those details make an answer sound remembered, and they're what an interviewer means by "own recollection."

### Find the claims they'll probe hardest

You don't need equal depth on every row. Spend most of your time on the claims that match the posting's top requirements, because that's where a hiring manager will push. Refolk scores how well you fit each posting, which shows you which requirements are strong and which are thin. The thin ones get probed hardest, so give those rows the most detail.

### Rehearse it, then put it away

Practice the sheet out loud until you don't need it, then keep it face down or closed during the call. If you read from it, your eyes will make the same movement that gets copilot users flagged.

- **Read each row aloud once**, then answer the bullet out loud without looking.
- **Have a friend ask "why?" three times** on your two strongest claims. If you run out of answers on the third one, add a row.
- **Say your numbers the same way every time.** Interviewers compare what you say to your resume, and a figure that changes between rounds hurts you more than a modest one.
- **Allow yourself one glance** at a single keyword on a sticky note if you blank. Looking away for a moment reads as thinking. Reading a block of text reads as a script.
- **Pause on purpose before you answer.** If it's clear you're remembering something, the pause sounds nothing like waiting for a transcript.

When a follow-up comes, answer with a detail from the sheet: the baseline, a name, the thing that broke. Specific answers sound remembered. They also tend to stop the probing, because you've already given the interviewer what they were digging for.

## FAQ

### Can recruiters tell if I'm using LockedIn AI in an interview?

Often, yes, and not through software. They notice reading eye movements, the same delay before every answer, and answers that are polished but say nothing specific. Stealth Mode only hides the window during a screen share, and Interview Coder (a competitor) reports that some users say interviewers could still see it. A Greenhouse report found 38.5% of candidates across 19,368 live interviews were flagged for AI-assisted behaviour.

### Is it cheating to have notes during a video interview?

I think one keyword you glance at once is fine, and a brief look away reads as thinking. Reading full answers is not fine, whether they come from a document or a copilot, and it's worse if the interviewer has asked you to confirm your answers come from your own recollection. The safest approach is to rehearse your evidence sheet until you only need it as a backup.

### How do I answer a follow-up question I didn't prepare for?

Use the structure of the evidence sheet even if the exact question is new: what it was before, what you did yourself, how you measured it, what went wrong. Most follow-ups are some version of one of those four. If you really don't know, say so and give the closest thing you do know. That comes across better than a smooth answer that doesn't actually address the question.

---

_From the Refolk job search blog. More posts: https://www.refolk.ai/candidates/blog/index.md_
