---
title: QA experiments with Optimizely Mark
slug: web-experimentation/docs/qa-experiments-with-optimizely-mark
createdAt: 2026-11-02T21:28:09.219Z
---

Mark (Optimizely Agent Platform) is an agent orchestration platform that helps you work smarter in Web Experimentation. Use Mark to run quality assurance (QA) on the variations of an A/B test. A QA run checks for issues before visitors see them, so you do not click through each variation yourself. Start a QA run from the experiment's **Variations** page. Mark checks each variation in turn and reports what it finds in Mark chat.

## Choose a QA method

Pick the QA method that matches how your pages load. Start each method from the **QA** menu or from the **More Options** menu of a variation:

- **Mark QA** – Runs the QA in Mark chat in Web Experimentation. Mark QA needs no browser extension. Use it to QA every variation or a single variation.
- **Browser QA** – Hands the QA run to the Optimizely Web Companion browser extension. The extension opens Mark chat in the Chrome side panel and loads your pages in your own browser session. The checks see your pages with your cookies and logged-in state. Use Browser QA when a page needs your session, such as a page behind a login.

## Prerequisites

Confirm the following before you start a QA run:

- Your administrator gave you access to Mark, and you log in to Optimizely with Opti ID. See [Configure Web Experimentation for Mark](docId\:hlHMQy1FYLEFZuoQwEAep).
- Generative AI is on for your Optimizely account. See [Turn generative AI off across Optimizely applications](docId\:X5U0kKD2azV5cIdNqwGNj).
- The experiment is an A/B test. Mark QA and Browser QA do not display for multivariate tests or multi-armed bandit optimizations.
- For Browser QA, you have Optimizely Web Companion in Google Chrome. Log in to the extension with the same email address you use for Optimizely. See [Optimizely Web Companion](docId:8FWbsdOqQEXiJzTmZULdN).

## Run Mark QA

Run Mark QA to check an experiment from Web Experimentation. Each run opens its own thread in Mark chat, so the results of earlier runs stay in their own threads.

### QA every variation

Check every variation of the experiment in one run.

1. Go to the **Variations** page of your experiment.
2. Select **QA** > **Mark QA**.

Mark chat opens a thread that asks the QA agent to QA the experiment. Mark checks each variation in turn and reports what it finds in the thread.

### QA one variation

Check a single variation without running QA on the whole experiment.

1. Go to the **Variations** page of your experiment.
2. Select **More Options** for the variation.
3. Select **Mark QA**.

Mark chat opens a thread that asks the QA agent to QA only that variation.

### Prompt suggestions in Mark QA threads

A Mark QA thread does not display the prompt suggestions that Mark chat usually shows under its greeting. Hiding them keeps the QA results in view. The suggestions return when you select **New Chat** or switch to another thread.

## Run Browser QA

Run Browser QA to check your pages as they load in your Chrome session, including pages behind a login. Browser QA checks the whole experiment, even when you start it from the **More Options** menu of a variation.

1. Go to the **Variations** page of your experiment.
2. Select **QA** > **Browser QA**.

Optimizely Web Companion opens Mark chat in the Chrome side panel and starts a QA thread for the experiment. The extension loads your pages in a tab that it opens for the run. For what the extension does during the run, see [Run QA with Mark](docId:8FWbsdOqQEXiJzTmZULdN).

If Browser QA does not start, Web Experimentation displays an error message. For each message and its fix, see [Browser QA does not start](docId:8FWbsdOqQEXiJzTmZULdN).