On this page
What the integrity report is#
Every interview room keeps a short, timed list of moments a person may want to look at, such as a paste or a switch to another window. That list is the integrity report. It sits in the finalist packet next to the scores.
What it notes today#
Four notes are built. Each one is In testing until it passes the false-flag check below.
| Note | What it notices | Status |
|---|---|---|
| Leaving the window | The candidate left the interview tab, and for how long. | In testing |
| Pasted text | What was pasted, how much, and when. Paste is off by default in coding rounds. | In testing |
| AI-copied code | Code that matches an answer an AI gave for the same problem. | In testing |
| Typing bursts | A lot of text appearing at once, faster than typing. | In testing |
| ID check | Only when the role turns on the optional ID check. We keep pass or fail, the document’s country and its type. | Off by default |
| Camera and gaze | Where the candidate looks during the room. | Building next |
Camera and gaze are switched off today, so they are in no report. The ID check has not been tested with the live provider yet. Separately, the scorer flags phrases that try to give the AI instructions. A person reviews those too; they never reject anyone on their own.
What it never does#
- It never changes a score.
- It never moves a candidate to another stage.
- It never rejects anyone on its own.
- It never scores emotion or facial expressions.
- It says what happened and when, never why.
How to read it#
- Open the finalist packet and go to the integrity section.
- Each note shows its time in the interview. Jump to that moment in the transcript or the clip.
- Ask whether the note explains anything in the answer. A paste right before a perfect answer matters more than a paste of the problem text.
- If something still worries you, ask about it in your final round. You run that round yourself.
What candidates are told#
Before each room, the candidate agrees to have activity such as leaving the window, pasting and typing bursts noted. On their “Your data” page they see the same notes, in the same words. That page is built, not live yet. They can also ask a person to re-check their result, or ask for a human interviewer. Read the AI interview notice.
When the role allows an AI assistant#
Coding rounds can include an AI assistant: off, hints or full. The default is off. When it is on, using it is part of the round, and an “AI collaboration” skill is added to the scorecard.
False flags#
A false flag is a note on an honest session. Every check like this has some. A check stays In testing until at least 300 honest test sessions, in each kind of room, show it notes 1 honest person in 100 or fewer. While a check is in testing, it never counts toward the review level.
We have no false-flag numbers yet. When we do, they go on the Trust page, with the number of sessions behind each rate.
Was this helpful?
