What it does
UserTold.ai runs AI-powered voice interviews with real users at the moment they get stuck inside a product. Using GPT-Live-1, it records a voice debrief triggered in-session, capturing the user's quote, screen recording, and page navigation path together as linked evidence. Product teams and coding agents can then review these evidence cards—each with a confidence score and separate fields for what was said, observed, and inferred—before grouping them into draft findings and pushing those findings to Linear or GitHub.
Use cases
- Diagnosing onboarding drop-off by observing where users hesitate during setup and asking them to explain their expectation at that exact moment
- Identifying navigation confusion, such as users repeatedly returning to the wrong page when looking for billing settings
- Understanding pricing page confusion by capturing how users interpret plans and charges before they decide
- Connecting a feature request to the underlying task and workaround a user actually has
- Letting a coding agent run an end-to-end study via MCP or CLI, collect real-user evidence as JSON, and propose fixes grounded in that evidence
Who it is for
- Product teams who need to understand why users drop off rather than just where, based on in-product behavior
- Developers and coding agents that want structured, machine-readable user research evidence linked to specific pages and moments
- Early-stage SaaS founders who cannot get users to schedule separate research calls but want qualitative insight from existing users
- UX researchers who want timestamped, source-attributed evidence rather than AI-generated summaries without receipts
Alternatives and comparison
No named competing products are identified on the page. UserTold positions itself against three general patterns: session replay tools that show behavior without explanation, outreach-based interview scheduling that users ignore, and context-free feedback surveys. Its stated differentiation is combining voice debrief, screen recording, and page path into a single timestamped evidence card with explicit confidence scoring and separate fields for observation, quotation, and inference.
- Session replay tools (e.g., general category)
- Session replay tools show where users navigate but do not capture the user's spoken explanation of why they struggled; UserTold pairs the replay timestamp with a voice debrief from the same session.
- Scheduled user interview platforms (general category)
- Outreach-based interview scheduling depends on users accepting calendar invites, which the page describes as a common failure point; UserTold triggers the interview in-product at the moment of struggle without a separate booking step.
- In-product survey tools (general category)
- Text surveys collect feedback without screen or navigation context; UserTold attaches the quote to the exact page path and screen state, keeping context intact.
