The Question Engine · patent pending

We stopped trying to see the phone. We made it useless.

A phone running an AI assistant, an earpiece, an accomplice just off-camera: help that never touches the interview computer, so no software can catch it. Capifiq closes that gap from the other side. Instead of hunting for the phone, we build the questions: interviews where a snapshot of the screen doesn't get you an answer, and where the follow-up was never on the screen at all.

The full picture

Two pillars. One integrity.

Interview integrity has two halves. Capifiq covers both. It catches what runs on the device, and it makes what sits beside the device far less useful.

Detect · on-device

Any assistance on the machine

Capifiq surfaces unauthorized tools running on the candidate's computer: hidden AI assistants, remote-control sessions, virtual machines, virtual cameras, browser assistants, display tampering, and new vectors as they appear. Deterministic, timestamped, sealed evidence.

See on-device detection →

Outpace · off-device

Help that never touches the device

A phone, an earpiece, a person off-camera. No software can see any of it, so Capifiq changes what they'd be helping with. The Question Engine builds the interview itself, and it comes with every account.

You are here →

The gap

What no software can catch.

The newest cheating doesn't run on the interview computer at all. It sits beside it, and that is exactly why software, ours included, can't see it.

A second device

A phone or second laptop runs an AI assistant beside the interview, feeding the candidate answers. It never touches the machine being monitored.

An earpiece

Answers or coaching whispered through an earpiece. There is no software signal at all: the help arrives as audio.

An accomplice off-camera

Someone just out of frame reading questions and supplying answers, or a proxy who quietly takes over the hard parts.

Pre-fed material

Memorized answers, a leaked question bank, or prepared scripts: nothing running live on the device to detect.

This is not a limitation to hide: it's physics. If the help never runs on the interview computer, no software can observe it. Any vendor telling you their software sees a phone on the desk is overselling. So we went the other way: if you can't catch the help, take away what it's useful for.

What it does

  • Writes the interview questions from your brief
  • A person approves every one before a candidate sees it
  • Presents them live, during the interview
  • Hands the interviewer the follow-up to ask next
  • Included with every account

How it works

A snapshot of the screen isn't an answer.

Off-device help runs on a simple loop: read the question, send it out, wait, read the answer back. Every part of that loop needs the question to be sitting there, whole, in one place, long enough to capture.

Capifiq's questions aren't shaped that way. The whole question isn't on the screen at one moment, so a single photograph doesn't contain enough to answer it, getting the full picture means capture after capture, while the candidate is on camera, mid-conversation, being watched. The loop gets slower and more obvious every time round.

Then the interviewer asks the follow-up, which was never on the screen at all.

What's included

You describe the role. It writes the interview.

Nobody has to sit down and design a cheat-resistant interview from scratch. Give it a job description, a pasted paragraph, or a single question you already like. It takes it from there, and a person signs off on everything.

Start from anything

A job description, a rough brief, or one question you want built out. There's no form to fill in and no question bank to maintain.

Interactive, not a list of prompts

Questions a candidate works through rather than reads out, which is what makes them awkward to relay to someone else in real time.

You approve every question

Nothing reaches a candidate unreviewed. You read it, edit it, regenerate it, or throw it out, the interview is still yours.

Grounded in real sources

Factual questions are built from cited material you can open and check, so the answer key holds up when a candidate disagrees with it.

The follow-up, ready to ask

Each question arrives with the probe to ask next: never shown to the candidate, so there's nothing to photograph and nothing to prepare for.

Kept current

As new assistants and tactics appear, the question formats move with them. What worked against last year's tools isn't the bar.

What stays private: how the questions are built is the product, and it's patent pending. Your team gets the questions, the sources behind them, and the follow-ups to ask, without us publishing the recipe for the tool-makers to design around.

Where the line is

What we don't claim.

The tools get better every few months. Anyone promising you a solved problem is selling you one.

It doesn't prevent cheating

Nothing does. A determined candidate with enough patience can still work at it, the point is that it now costs them time and visible effort in a conversation where both are obvious.

It doesn't see the phone

We have no idea whether there's a phone on the desk, and we won't claim we do. What we can change is whether having one gets them through the interview.

It doesn't decide anything

No score, no ranking, no verdict: not even a hint of one. It hands your interviewer better questions and the follow-up worth asking. The judgement stays where it belongs.

Patent pending. Capability described here is subject to our terms and disclaimers: features change, availability varies, and no finding is a substitute for a person's judgement.

Questions

Common questions.

Why can't detection software catch off-device cheating?
Because it never touches the interview computer. A phone, an earpiece, or a person off-camera runs entirely outside the machine, so no software, ours or anyone else's, can see it. Rather than pretend otherwise, Capifiq closes the gap from the other side: it builds questions a phone can't usefully answer.
Does it replace our interviewers?
No, and it isn't trying to. It writes the questions and hands your interviewer the follow-up worth asking; they run the conversation and they make the call. Capifiq produces no score, no ranking and no verdict.
Do we have to use the questions it writes?
Only the ones you approve. Every question is reviewed by a person before a candidate can see it, and you can edit it, rebuild it in a different format, or discard it. You can also hand it a question you already use and have it built out.
Where do the answers come from, can we trust them?
Factual questions are built from cited source material, and the reviewer can open the sources and check them. That matters: an interview question is only as good as the answer behind it, and a confidently wrong answer key is worse than no question at all.
How does this fit with Capifiq's detection?
Two halves of one picture. Capifiq detects unauthorized assistance running on the candidate's device, and the Question Engine handles the help that never touches it. The machine and the desk beside it.
Does it work for non-technical roles?
Yes. It builds from whatever you describe, so it fits technical screens, behavioural rounds and everything in between. What changes is the shape of the question, not the approach.

Get started

Cover both halves of the interview.

Catch what runs on the device. Outpace what sits beside it. See both on your own interviews.