> Documentation index: https://usertold.ai/llms.txt

# Unmoderated Usability Testing on Your Own Site

Unmoderated usability testing is a test where a person does a task in your product with no researcher present, and the session is recorded for you to watch later. This page explains when it fits, how it differs from a moderated test, and how to run one on your own site with your own users.

## Moderated and unmoderated tests compared

| | Moderated | Unmoderated |
|-|-|-|
| Who is present | You and the participant, in a room or on a call | Only the participant |
| Scheduling | A time slot for each session | None; the participant starts when they choose |
| Follow-up questions | You ask them live | Written in advance, or asked by a voice model after the task |
| Sessions per week | A few | As many as people start |
| Risk | You may lead the participant without noticing | The participant may misread the task and nobody corrects it |
| Best for | Early ideas, complex workflows, sensitive subjects | One clear task in a working product |

Both kinds can be remote. “Remote usability testing” means only that the participant is not in the room with you.

## When an unmoderated test fits

- The product or prototype works well enough for a person to finish a task alone.
- You can describe the task in two sentences.
- You want to see the same task from several people.
- Your users are in other time zones or will not book a call.

Choose a moderated session when you do not yet know what to ask, or when the workflow takes more than ten minutes.

## How to run one on your own site

1. **Choose one outcome.** For example: a new user sends a first invoice.
2. **Write the task.** Name the outcome, give a stopping point, and say what the person must not do. The [usability test script and task examples](/guides/usability-test-script) has wording you can copy.
3. **Choose who sees it.** Offer the test on the page where the task starts, to the users it applies to.
4. **Record the screen and the voice.** Ask the person to say what they are looking for while they work. Without the voice you see a pause and have to guess the reason.
5. **Ask about the task right afterwards.** Use the [usability testing questions](/guides/usability-testing-questions) for after the task.
6. **Review five sessions before changing anything.** A problem that two or more people met in the same place is ready to fix.

## The weak point of unmoderated tests and how to cover it

In a moderated test you can ask “What were you looking for just now?” at the right second. A written list of questions at the end cannot do that, because it was written before the session happened.

UserTold covers this with a voice conversation after the task. The widget shows your task and records the screen and the person's voice while they work. When they finish, the voice model asks about the pauses, errors and returns it captured, one question at a time. Nobody from your team has to be online.

You get the full recording, and each moment where the person had trouble is marked with their words, the page and the time. A coding agent can read those moments over [MCP](/docs/mcp) or the [CLI](/docs/cli).

## Your own users or recruited testers

An unmoderated test can run with people from a tester panel or with your own users. Panel testers are available at once and suit a product with no users yet. Your own users have real accounts, real data and their own reasons to finish the task, so their problems are the ones your customers have. UserTold works with users you can already reach and does not recruit participants. The longer comparison is in [UserTesting alternative](/guides/usertesting-alternative).

## Consent and safety

Tell the person what is recorded before the recording starts, and let them stop at any time. Write the task so it ends before a payment, an invitation to a real person, or anything that cannot be undone. [Participant consent](/guides/participant-consent) covers the disclosure in detail.

[Sign in](/login) and choose **New study**, or read the [Quickstart](/docs/quickstart) first.
