From design to verdict.
AI first, people second, one trail through both — plus three audits that need no testers.
Three moves, in the order a study makes them
Each one is a page of its own.
No study, no tasks, no testers
An expert read of the screen, the copy or the build — back in minutes.
UX Audit
One screen, one goal, ranked findings. Every finding a prediction.
Content Audit
Six registers. Measured and judged kept apart. Rewrites ready to paste.
Fidelity
The design and the built screen side by side, scored on how close they are, with every comparison saved so you can watch a screen improve.
Eight pages, one product
Nothing here is on a roadmap.
AI testing
A cohort of any size attempts your task at once, thinking aloud, streamed live.
OpenHuman testers
One share link. Consent built in. No panel, no scheduling, no SDK.
OpenThe study builder
Missions and eleven question-block types in one ordered list.
OpenTester library
Built-in archetypes and your own — traits that change how a session goes.
OpenUX Audit
One screen, one goal — ranked findings and a blueprint, no testers required.
OpenContent Audit
The words read against their job — six registers, reading level measured, rewrites ready to paste.
OpenFidelity
What shipped against what was designed — scored overall and per category, every comparison kept.
OpenEvidence & reports
Ranked findings, scored tasks, and a report you can hand over.
OpenThe parts that make it a workspace, not a tool
A workspace per client, roles per area, and a bill you can read line by line.
A workspace per client
One organization can hold as many as you need. Studies, testers and reports stay inside the workspace they belong to, so nothing from one client is visible in another.
Roles and per-area access
Admin, member or guest, set separately for research, the audits and the library — so a stakeholder can read a report without being able to start a run.
Spend you can read
Every model call is metered and attributed to the session that made it, with a monthly total, a daily trend, and a breakdown of where the money went.
Handing work out
Share a read-only report link, export a PDF, or send the findings to a coding agent as a work order. None of it needs an account on your side.
The rules the product keeps about its own numbers
Enforced in the product, not promised on a page.
Simulated is labelled simulated
Anything the AI predicted is labelled in the results, in the PDF, and on any link you share.
An estimate never wears a measurement's clothes
A predicted number says it is predicted; a measured one says how many sessions it came from.
Unmeasured says unmeasured
A metric nothing recorded says “not measured” — never a zero, which would read as a real result.
Human testers consent first
Nothing is recorded until the tester has agreed, and screen recording is a separate question again.
Test from your pocket
The uTestMe tester app runs a study on a phone. Taps and screens recorded, consent first, results in the same view.
- Tap tracking
- Screen recording
- Consent first