Automation bias — and why we keep reminding you

What automation bias is

Automation bias is the well-documented human tendency to over-trust an answer because a computer produced it — to skim it, assume it's right, and skip the check you'd have done if a person had said the same thing. It runs both ways: acting on a wrong suggestion (commission), and NOT acting when the tool is silent or wrong (omission).

For job-hunting it matters because the stakes are yours. A match-gate score, a tailored CV line, a cover-letter paragraph — each is a starting point a model generated from limited information, not a verdict on you or the role.

What we do about it

• Reminders at the point of use. Every AI surface carries a short, always-on note — not a one-time popup you dismiss and forget — that the output is a data point, not a decision.
• A fuller card on your first few runs. The first three times you see a match-gate result, we show the long version: this is one model's view of the fit; use the score as a data point; decide for yourself.
• Honesty about limits. A below-gate score says the gap is real and stops — that's a correct outcome, not a failure to route around. We never present a score as permission to apply, or not to. And we never invent a claim about your experience to make a draft fit.

How to read AI output well

• Treat every draft as a first draft: read it, change what's wrong, own what you send.
• When the score is low, ask whether the gap is real before dismissing the tool — and whether it's real before dismissing yourself.
• If a line about your experience doesn't cite something you actually did, delete it. The AI is told to ask you rather than invent — hold it to that.
• The deciding factor is your judgement. The model doesn't apply for the job; you do.

See also About our AI for what the AI does, what it doesn't, and the safety checks every output passes before it reaches you.