Guide

How to test your website
with AI.

You should not need a developer to find out whether people can sign up, buy something or send you a message. An AI testing tool can open your site in a real browser, do what you describe, and tell you what happened, with a video to prove it. This guide shows how that works in Testorim, what to ask for, and where it stops.

Join the waitlist

What you type

“Fill in the contact form with a test message and check that a thank-you message appears.”

What comes back

Passed · every step

video · screenshots · accessibility scan

console and network errors: none

Three steps, no code.

01

Give it your site's address

Paste the address people type to reach your site, or a staging copy of it. The site has to be reachable on the public internet: a copy running on your own computer or behind a company network cannot be opened.

02

Describe what a visitor should be able to do

Write it the way you would explain it to a new colleague, and say what success looks like: “and check that a thank-you message appears.” One goal per test keeps the answer clear. The AI turns your sentence into steps and runs them in a real Chromium browser while you watch.

03

Read the report

Every run ends as passed, failed or needs review. Failed means a step did not work, and the report says which one and what the page showed. Needs review means the steps ran but something could not be confirmed, so a person should look. A run you are happy with can be saved and replayed later without the AI, on a schedule if you like.

Things you can ask for.

Each of these is a complete test. Swap in your own pages and wording; the more plainly you say what should happen at the end, the clearer the answer.

Sign-up

“Create an account with a new email address and check that the welcome page opens.”

Log in

“Log in with the test account and check that the account page shows its email address.”

Checkout

“Add any product to the cart, go to checkout, and check that the order summary lists it with a total. Stop before paying.”

Contact form

“Fill in the contact form with a test message and check that a thank-you message appears.”

An empty form

“Submit the contact form with every field empty and check that it asks for an email address.”

Search

“Search for “blue shirt” and check that at least one result appears.”

Broken links

“Open each link in the footer and check that every page loads.”

Booking

“Open the booking page, pick the first available date, and check that time slots appear.”

What a run gives you back.

Not a score: evidence. If something is broken you can send the report to whoever built the site, and they will have what they need to fix it without asking you to reproduce it.

  • A video of the whole visit, so you can watch exactly what happened.
  • Step-by-step results each step marked passed or failed, with the reason in plain words.
  • Screenshots of the final page and of any step that failed.
  • Console and network errors the problems a developer will ask for, captured around the step that failed.
  • An accessibility scan of the pages it visited, listing issues such as missing labels.
  • A written report that you can share, plus a Playwright trace if a developer wants to dig in.

What it cannot do.

Native mobile apps

It tests websites in Chromium. An iPhone or Android app from an app store is out of reach.

CAPTCHAs

A “prove you are human” check is built to stop exactly this kind of visitor. Turn it off on a staging copy, or test the steps around it.

Codes sent by email or SMS

A live run can pause and ask you to type the code in. A scheduled run cannot, because nobody is there to answer.

Private sites

Anything on your own computer, a company network or a private address is blocked on purpose.

Your whole site at once

It checks the pages and links you describe. It does not crawl every page looking for problems.

Other browsers

Runs use Chromium, the engine behind Chrome and Edge. On the Team plan, saved tests can also replay in Firefox and in WebKit, the engine behind Safari.

Not the same as an AI website checker.

Many tools sold as AI website checkers work like this: you paste an address and get a grade for page speed, search tags, image sizes or mobile layout. They read the page. They are quick, and worth running.

They cannot tell you whether a visitor can finish anything. A page can score well and still have a checkout button that does nothing or a contact form that never sends. Testing with AI in the sense of this guide means a browser that uses the site: it types, clicks, waits for the answer and reports whether the thing you described worked.

Quick answers

  • Do I need to know how to code?

    No. You write what a visitor should be able to do in ordinary English, and the report comes back in ordinary English too. A developer can open the Playwright trace or export a saved test as code later, but nobody has to.

  • Will a test change anything on my live site?

    It can, because it does what a visitor does. A sign-up creates an account and a submitted form sends a message. Use a test account, stop checkouts before payment, or point the test at a staging copy of your site if you have one.

  • How is this different from an AI website checker?

    A website checker grades a page: its speed, its search tags, its images. This uses the site the way a visitor does, filling in forms and clicking through, and tells you whether the thing they came to do actually worked. The two answer different questions, and many sites benefit from both.

  • Is my data used to train AI?

    No. Test descriptions, screenshots, reports and run data are used to run and explain your test, not to train AI models. Passwords typed during a test are masked in the report.

  • What does it cost?

    Plans are $19, $79 and $199 a month. Each starts with a 3-day trial that needs a card, and there is no free plan. Sign-up is currently by waitlist.

Join the waitlist