Recagn
Sign in

How it works · Anti-cheating

Interview integrity: the rules behind every cheating flag

Every rule Recagn applies to an interview session, with its level and the check it needs. Recagn's rules list what happened; you decide what it means.

Which rules trigger an interview cheating flag?

Recagn applies a fixed rule table, integrity@1, to every interview with Activity Tracking, once it's scored. It reports the highest level any rule reaches: Strong concerns, Some concerns or, when nothing fires, No concerns. Each row below names the event, its level and the check it needs. You decide what a finding means.

RuleLevelNeeds
More than one face is in a webcam still.Strong concernsWebcam
A webcam still matches the Selfie at under 80% similarity.Strong concernsWebcam + Face match
The ID photo doesn't match the Selfie.Strong concernsFace match
The camera is off for more than 1 min in total.Strong concernsWebcam
Screen Recording is stopped for more than 1 min in total.Strong concernsScreen Recording
No face is in 2 webcam stills in a row.Some concernsWebcam
The head is turned 35° or more to the side in 3 webcam stills in a row.Some concernsWebcam
The camera is off for up to 1 min in total.Some concernsWebcam
The ID can't be read, or no face is found on the ID photo or the Selfie.Some concernsFace match
3 or more pastes from outside the page, each over 50 characters.Some concernsAlways on
3 or more stretches away from the page, each over 10 s, or more than 2 min away in total.Some concernsAlways on
A candidate presses a developer-tools keyboard shortcut.Some concernsAlways on
The connection drops for more than 1 min.Some concernsAlways on
Rules integrity@1, as of

The level follows the rows. If any Strong rule hits, the interview reads Strong concerns. If only Some rules hit, it reads Some concerns. Otherwise it reads No concerns. These are plain rules, not an AI model, so the same evidence always gives the same result, and every result traces back to the rule behind it.

The Needs column says which check a rule depends on. "Always on" is Activity Tracking, which runs on every interview set up since it shipped and can't be switched off. Webcam, Face match and Screen Recording are checks you switch on when you design the interview, and a rule only runs when its check is on. Every Strong rule needs the webcam, Face match or Screen Recording.

Recagn's AI proctoring page lists what each check records, what you see and what it costs.

Integrity has three levels and no score. You can't sort candidates by it or add it to anything.

What is an Integrity finding?

An Integrity finding is one rule in the table that fired, written by Recagn as one sentence and tied to its moment on the timeline, or to the photo it concerns. A notable event is different: Recagn highlights it on the timeline, but it isn't a finding unless a rule counts it. You read both and decide which matter.

Recagn sorts what happens in an interview into three tiers:

  1. Recorded: every activity event goes onto the timeline.
  2. Notable: Recagn highlights the events most worth a look: a long paste from outside the page, a long stretch away, a developer-tools keyboard shortcut, a stopped screen share, a lost camera or a gap in the connection.
  3. Finding: a rule fired. Some rules fire on a single event, others on a count or a total. The table shows which.

A session finding names what happened, how much or how many, and when it first happened on the interview's clock. It links to that place on the timeline. A finding about a webcam still links to the still itself. An ID-check finding has no time: it says the ID photo didn't match the Selfie, or that the ID couldn't be read, and it links to the Selfie or the ID photo.

Leaving the page once to pick a file for a Document answer doesn't count as time away. The rules were written around how candidates answer.

What does the evidence timeline show?

The activity timeline puts everything Recagn recorded in one interview on a single clock: page events, screen and webcam recordings, webcam stills and each Integrity finding. You see it per answer and across the whole interview. Recagn orders it by server time, so you can go from a finding straight to the moment it describes.

The page events on it are:

  • tab switches and focus changes, with how long each lasted;
  • copy, cut and paste, with the text;
  • pastes outside answer fields, which are blocked, with only their length kept;
  • key combinations;
  • the connection dropping and coming back;
  • screen share and camera stopping and starting;
  • files picked for a Document answer.

Long runs of digits are masked before anything is stored.

Webcam stills are taken at random gaps through the interview and analysed after it, when it is scored. The face count, which way the head is turned and the match against the Selfie run on Recagn's side, through AWS Rekognition in Mumbai (ap-south-1).

Integrity's rules read the events from the interview page and, from each webcam still, three things: how many faces are in it, how far the head is turned to the side, and how closely it matches the Selfie. Screen and webcam video sit on the timeline for you to watch.

What a browser reports for each of these signals, and what it can't see, is in what browsers can detect in an interview.

ID check by face match compares the face on the ID photo with the Selfie the candidate takes on camera just before starting. The match is between photos.

Who sees Integrity results?

Only the recruiter sees Integrity, on Recagn's interview review for each candidate, with each finding linked to its moment on the timeline or the photo it concerns. The candidates list and the Company Report never show it, so the hiring company never sees a level, a finding or a recording from the session.

Under the level, the review carries one line: "Recagn's rules list what happened; you decide what it means. Not used in scoring or ranking; the hiring company never sees it."

The level is drawn as a hollow, half or filled ring in ink. There is no red and no score.

No proctoring media, activity event or finding reaches the Company Report. For the candidate's side of the interview, see candidate consent for AI interviews.

Recagn does the screening. You make the calls.

Does Integrity change the score, the ranking or the Company Report?

No: Recagn keeps Integrity out of the interview score, the ranking and the Company Report. The score comes only from the answers, and Integrity is computed separately after the interview is submitted, when it is scored. Its code is fenced, so the scoring and report modules can't import it.

A lint rule stops the scoring, Authenticity, Consistency, comparison, Company Report and Learning Loop modules from importing Integrity, proctoring or activity code. A test plants a marker in Integrity and proctoring data and checks that the published Company Report never contains it.

Because Integrity waits for the submitted interview, no rule can stop a candidate mid-interview, and the ID check result never blocks the start. Recagn recommends. Only you can reject a candidate.

Integrity asks what happened in the session. The Authenticity Detector asks whether an answer is the candidate's own work. The two run separately, and nothing crosses between them.

For how the score itself is built, see how AI interview scoring works.

Which version of the rules is this?

This page shows integrity@1, the rule table Recagn has applied since 15 Nov 2026 to every interview with Activity Tracking, once it's scored. Each Integrity result stores the version it was computed under. When a rule or threshold changes, the version moves to integrity@2 and this page's stamp changes with it.

An interview on a Version locked before Activity Tracking shipped has no Integrity. Its review says so: "No Integrity: this Interview ran on a Version without Activity Tracking".

Features vary by plan. See pricing.

Questions people ask

Can a candidate still cheat with a second phone?

The timeline shows what the interview page and the webcam recorded, so you can check any still or recording for a phone yourself. Integrity's webcam rules read three things in each still: how many faces are in it, how far the head is turned to the side, and how closely it matches the Selfie. You decide what each finding means.

How is this different from Mettl's Credibility Index?

Mettl's remote proctoring page describes a Credibility Index report that captures each candidate's flagged instances and how long they lasted, and it doesn't say how the index is worked out, going by Mettl's remote proctoring page, read 17 Nov 2026. Recagn publishes every rule on this page, reports three levels with no number, and keeps Integrity out of the score and the client's report.

Does Integrity cost extra?

No. Integrity, Activity Tracking, Face match and Screen Recording use no credits. Webcam Snapshots add one credit to an interview, and Continuous adds two for the first hour and one for each further 30 minutes begun; see Recagn pricing.

Does Integrity catch AI-written answers?

Integrity finds pastes from outside the page, and the Authenticity Detector, separately, judges whether an answer reads as written by an AI model. Recagn keeps the two apart, and neither changes the other. You see both on the interview review.

Can a Strong concern stop a candidate mid-interview?

No. Recagn computes Integrity after the interview is submitted, when it is scored, so no rule can stop or end an interview. The ID check result never blocks the start either. You decide what happens next.

See it run on one of your roles.