> ## Documentation Index
> Fetch the complete documentation index at: https://docs.humanize.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Participant quality

> How Humanize keeps bad participants out of your data — and what happens when one slips through.

Bad participants are the quiet failure mode of online research. Someone signs up twice under different names, clicks through a screener in nine seconds, joins from a country your study doesn't cover, or sits silent for forty minutes and collects the incentive. None of it looks obviously wrong in a transcript, and all of it ends up in your findings.

Humanize checks for this continuously, at three points in a participant's journey. You don't configure any of it.

## Three checkpoints

<Steps>
  <Step title="When someone applies">
    Before an applicant is accepted, Humanize checks that they're a real, distinct person applying from where your study expects — and that they answered your screener like someone actually reading it, rather than racing through or picking the same option down the page.
  </Step>

  <Step title="Before they join a session">
    On the way into a session, participants pass device and environment checks: camera and mic working, face clearly lit, connection stable enough for video. Anyone who fails is prompted to fix it before they can enter.
  </Step>

  <Step title="After the session">
    Once a session ends, Humanize reviews how it actually went — whether each participant showed up, stayed for the discussion, and genuinely took part, rather than sitting silent or talking over everyone else.
  </Step>
</Steps>

## What you'll see

Applicants who trip a check carry a **quality flag** on the Recruitment tab, with the reason. Flags are visible on the applicant list and on each applicant's detail view, so you can see for yourself who was caught and why.

Most of the time you won't need to act on them. Flagged participants are handled before they reach your data — the flags are there so the sample you're reading is one you can account for, not a black box.

<Info>
  Checks are automatic and always on. There's nothing to switch on, and no setting that makes a study skip them.
</Info>

## If a session doesn't pass

Quality problems are our problem, not yours. **If a session doesn't meet our quality bar, we re-run it** — recruiting replacement participants and running another session at no additional cost to you.

You're paying for usable research, not for seats filled. A session where two participants were fraudulent or three never spoke isn't research you can use, so we don't ask you to keep it.

<Tip>
  A screener that's easy to game makes every downstream check work harder. Avoid questions that telegraph the answer you want — "Do you own a dog?" invites the obvious yes. [More on writing screeners](/study/screeners).
</Tip>