Quickstart

Run your first in-product AI interview from the dashboard. Verified new accounts receive a small starting balance. Both Managed AI and BYOK require positive prepaid balance; add an OpenAI API key in Project Settings to use BYOK.

1. Create a project

Sign in to the dashboard and create a Project. It already contains one active First user interview that asks what the participant tried to accomplish, what was difficult, and what they expected instead. It has no Intake, reward, or placement restrictions.

2. Copy the Project snippet

Open Studies and choose Copy snippet. Install this Project script once across the site:

<script
  async
  src="https://usertold.ai/v1/widget.js"
  data-project-key="ut_pub_YOUR_KEY"
></script>

The starter works without visiting the Study editor. This is the same code for every Study in the Project; Visibility chooses which active Study appears. After deployment, open Studies → Website interviews, paste a page where the snippet runs into Page to verify, and run the check.

3. Optionally tailor the interview

Use Web, REST, CLI, or MCP to edit the same Study's title, goals, flow, Invitation, Visibility, websites, or lifecycle. A connected agent can tailor it after you provide product context; see MCP setup, CLI setup, and the Study Design Guide.

See the AI interviewer for contextual debriefs.

4. Review participant disclosure

Keep your identity, purpose, voluntariness, and required privacy information in the host product or invitation. UserTold requires no separate privacy-policy URL or setup step. The widget supplies its recording notice, privacy link, and Allow recording and start action. See Participant Consent.

5. Understand placement

Visibility selects one active Study for the current pathname and widget language; that Study's Invitation defines the launcher. One Project can contain many active Studies. Without limits, the starter is eligible everywhere. See Widget Integration for matching and conflicts.

6. Test one complete Interview

Visit the embedded product as a participant and confirm:

  1. The widget opens First user interview.
  2. The Interview completes.
  3. Interviews shows the source recording, transcript, timeline, capture limitations, and processing state.
  4. Source-linked Evidence appears after processing, or the result says none was found.

Use Study Runtime to diagnose failures.

7. Review the result

Inspect the Interview source before its analysis. Evidence preserves a source-linked moment; a Finding synthesizes one or more Evidence cards. Check a draft Finding's sources, then choose Review. Review changes UserTold only; choose Send to Linear or the GitHub handoff separately.

Continue