What is AI user testing, and how does Menso run it?

What happens in a Menso test

1. You give it an address and a task. Paste your product's URL and pick a template: Purchase intent, Sign up & log in, or With materials. Or start from the address alone, and Menso researches the product and writes a persona and a task for it first; that research does not use a test.
2. An AI user works through the task. It gets its own browser session on your live site. At each step it looks at the screen, decides what to do, acts, and says what it expected and what happened. It can sign up or log in with an email and password you give it. Saved logins are stored on Menso's server and used by Menso's cloud browser to run the tests you ask for. The web app shows a saved login's email or username, never its password.
3. The run ends at a result. It stops when the task is done, when the product blocks it, or when its step budget runs out (the Product Hunt board run below had 16 steps). If a run is interrupted on Menso's side, the test stays held for one free retry: a retry that finishes counts once, and an unused or interrupted retry gives the test back.
4. You read the evidence. The replay shows every step with the AI user's think-aloud. The TRACES score sums the run up in six dimensions, each with a written reason.

What you get back

• A replay of every step: what the AI user saw, did and thought. Every plan.
• A TRACES score from 0 to 100 across six dimensions, each with its reason. Every plan. How the score works.
• A share link for a replay, for your team or a client, once Menso has checked that it shows none of your logins or secrets. Pro and Max.
• A UX report on every run, with a fix prompt per friction point for your AI coding tool. Free shows the first 3 friction points on screen; Pro and Max show all of them and export a PDF.

How deep does a test go?

Every test now runs at the Deep setting: the AI user gets time to think at each step, the same setting every run on the Product Hunt board used. There is no cheaper, shallower tier to pick; one test is one test.

What it costs

Pro: 12 tests a month for $29, about $2.42 a test; Free: 3 tests a month, no card.

• Free: 3 tests a month, there from signup; unused tests reset. A one-time Welcome pass in the sidebar adds 3 more that do not expire.
• Pro: $29 a month for 12 tests, plus replay sharing and two products at once.
• Max: $69 a month for 30 tests, six products at once and the front of the queue.
• Extra tests on any plan, Free included: 5 for $15; they do not expire with the month.

Every plan and pack is on the pricing page.

A real run, as it happened

Menso puts the top launches from Product Hunt's weekly board through one shared task. Here is one of them; the full board has the rest of its week.

Every run here used Menso's Purchase intent template: the persona Zainab Salazar, a freelance UX consultant, reads the homepage, the pricing page and the FAQ, decides whether she would buy or sign up, and stops before paying. It is a route-specific diagnostic, not a product-wide score.

Made one of these products? Write to support@menso.io and we will correct or remove the example as soon as we can.

Who Menso is for, and who it is not for

A good fit

• Your small team is about to ship, or has just changed its signup or purchase path.
• You want an AI user to try it first, before real visitors arrive.
• You cannot pay for a panel of human testers yet and want to decide where research with people should go first.

Not a fit

• You need to understand how customers feel or whether they trust you.
• You need a rate across many visitors, or scripted tests that pass the same way after every deploy.
• You need phone layouts or native apps checked.
• You need an accessibility audit or a security test.
• You need verification codes and CAPTCHAs handled without your help.

What it cannot tell you

• One AI user, one route, one run. The result describes that route, not your whole product, and a different persona or task can go differently.
• An AI user is not your customer. It has no money at stake and no history with your product. Studies of AI-simulated users in conversation tasks report that they can miss friction people feel and make results look more optimistic (Liu et al., Yoon et al.).
• It has no inbox or phone of its own. When Menso recognizes a signup or login page asking for a verification code, sent by email or by SMS, the run pauses and asks you to paste the code (or the emailed link), and waits up to two hours; if nobody answers, the run ends there. If it does not recognize the page, the AI user decides on its own what to do. A CAPTCHA that Menso recognizes also pauses the run. You can skip the step or stop the test, and the run ends after ten minutes if you do neither.
• It is not an accessibility audit. The board run cited on this page used no keyboard navigation and no screen reader.

Use it to find likely friction before real users arrive, and to decide where your research with people should go next.

If visitors already arrive and do not sign up, start with your analytics: the guide to finding out why visitors don't sign up shows where a task test fits after a funnel and recordings.