Skip to main content
Experience faster, smarter testing with BrowserStack AI Agents. See what your workflow’s been missing. Explore now!
No Result Found
Connect & Get help from fellow developers on our Discord community. Ask the CommunityAsk the Community

Visual Review Agent - Customization

Tailor the Visual Review Agent’s behavior with customizable prompts and visibility rules.

You can customize the Visual Review Agent to refine how it analyzes and presents visual differences. This feature allows you to define prompts and rules that guide the AI’s behavior, helping you focus on meaningful visual changes and reduce unnecessary noise.

Benefits

  • Deliver targeted visual feedback that aligns with your visual testing goals.
  • Focus reviews on meaningful visual changes by hiding irrelevant updates and highlighting critical ones.
  • Reduces noise by ignoring acceptable visual differences.
  • Maintains consistency across builds through reusable rules.

Example prompts

Here are sample prompts for customizing how the Visual Review Agent handles different types of visual changes:

Visual change Sample prompt
Minor shifts in logo or padding Ignore small pixel or layout adjustments in static logos.
Dynamic monetary values Ignore changes in order totals, prices, or other dynamic amounts.
Icon updates Ignore changes to social media or interactive icons.
Text updates in critical UI areas Always show text changes in headers and call-to-action buttons.

Access the prompt manager

When AI is enabled, you can access the Prompt manager in two ways:

  • AI Prompt icon: Click the prompt icon directly while reviewing a build.

Prompt icon for AI customization

  • Via NL description text option: Click an NL description to open the Prompt manager.

NL description option to open the prompt manager

This automatically generates a suggested prompt, which you can choose to edit or discard. You may also provide your own custom prompt.

Prompt window with editable prompt field on review screen.

Once a prompt is submitted, you must confirm it with the AI to run it.

Confirm AI response to continue option on review screen

After reviewing the AI’s response, you need to confirm again for the response to be saved for the entire build. The AI processes the prompt using the browsers enabled for that snapshot to deliver accurate and consistent visual insights.

AI results summary across browsers on review screen

Key capabilities

Explore the main customization features that help you customize, manage, and apply visual review rules effectively:

Manage Ignore rules and Always show rules

You can manage two types of rules in the prompt manager:

  • Ignore Rules: Hide specific types of visual changes, such as “ignore color shifts in icons.”
  • Always Show Rules: Ensure critical changes are always displayed, such as “always show text updates in the header.”

  • Rules can be defined at different levels:
    • Project level: Apply to all snapshots within a project.
    • Snapshot level: Override project rules for that specific snapshot.

Snapshot-level rules take precedence over project-level rules.

Edit prompts directly from the build review page

You can open the prompt manager directly from any build’s review page to:

  • Add, edit, or delete prompts and rules.
  • Save changes to automatically regenerate AI diffs.
  • Revert to the previous filter if needed.

Apply rules inline

When reviewing a diff, you can create or apply rules inline:

  • Select Ignore changes like this to hide similar visual differences in future builds.
  • Select Always show changes like this to ensure they remain visible in future reviews.

These inline actions automatically generate suggested rules, which you can edit and save at either the snapshot or project level.

Prompt application across builds

  • Newly created or edited rules are automatically applied to future builds.
  • Existing or in-progress builds continue using the previously active prompts to maintain result consistency.

Prompt limits

To maintain optimal performance and fair usage, the following limits apply:

  • Each rule can be up to 400 characters long.
  • A daily quota applies for regenerating comparisons after prompt edits. Paid groups can perform up to 200 regenerations per day.

If the quota is exceeded, you can try again the next day when the limit resets.

View audit information

Each rule includes detailed audit information to support collaboration and transparency. The audit log displays:

  • The name and email of the last person who edited the rule.
  • The timestamp of the last update.
  • The build number where the change was made.

Revert to a previous filter

After applying a new set of rules, you can revert to the previous filter directly from the Prompt Manager.

  • This restores the diff state and prompts from before the latest edit.
  • Useful when testing new prompt configurations or rolling back unhelpful rules.

Best practices

  • Keep prompts short and specific to one type of visual behavior.
  • Use project-level rules to maintain consistency across Percy builds, and snapshot-level rules to handle specific exceptions.
  • Review and refine prompts periodically to align with evolving design or testing standards.

We're sorry to hear that. Please share your feedback so we can do better

Contact our Support team for immediate help while we work on improving our docs.

We're continuously improving our docs. We'd love to know what you liked





Thank you for your valuable feedback

Is this page helping you?

Yes
No

We're sorry to hear that. Please share your feedback so we can do better

Contact our Support team for immediate help while we work on improving our docs.

We're continuously improving our docs. We'd love to know what you liked





Thank you for your valuable feedback!

Talk to an Expert
Download Copy Check Circle