AI tagging scans your application, identifies pages and features, and suggests tags for you to review and apply, including improvements to existing tags and fixes for broken ones.
Overview
AI tagging is available inside the Visual Design Studio (the designer). When you launch it, AI tagging scans your application, generates tag suggestions, and loads them into the designer when they're ready. Each suggestion includes a proposed tag name, rule, description, product area, and page-to-feature link. You review and accept suggestions before anything is applied.
Better tag coverage benefits your whole subscription. More accurate analytics, more precise guide targeting, and more reliable AI insights all depend on well-maintained tagging.
Prerequisites
A subscription admin must enable AI tagging in Settings (gear, top right) > Subscription settings > AI access, then check the box for AI tagging.
AI tagging uses Gemini by default and is available to all data regions. Claude is also available as an alternative AI model, but only in the US data region.
Tag with AI
Open AI tagging
- Go to Product > Features and select Tag Features. You can also start from Product > Pages > Tag Pages, AI tagging behaves the same way from either entry point.
- Select Launch Designer to open your application in the designer.
- AI tagging opens by default.
Select More options (...) to change the panel layout between Popover (default), New Window, or Fullscreen, and to view available keyboard shortcuts.
Tip: To switch between AI tagging and classic tagging, select the AI tagging dropdown at the top of the panel and choose Classic tagging, or select More options (...) and choose Switch to classic mode. To return to AI tagging from classic mode, select AI Enriched Tagging.
Set up before you start (recommended)
When the designer opens, two recommended actions appear before you start tagging. These are optional but will improve the quality of AI tagging's suggestions. For the full list of configuration options, see Additional settings below.
- Set naming convention: Describe your team's naming conventions in the text box. This helps AI tagging generate suggestions that match your existing structure and terminology. If you already have a lot of manual tags, setting this first is especially recommended.
-
Set tag rule preference: This determines the order in which AI tagging selects the rule for each suggested tag, from strongest to most fragile. Pendo provides a recommended default order. You can reorder selectors by dragging them, or toggle any off that you don't use.
Two things worth noting:- Selectors marked as recommended but without retroactive data are custom HTML attributes that haven't been added to your Pendo subscription yet. When you accept a tag using one of these, Pendo automatically adds it to your subscription. However, these selectors only capture data if the attribute is already present in your application's code. If it isn't, no data will be collected. For more information, see Custom HTML attributes.
- Selectors marked as not recommended (such as text :contains() and positional :nth-child()) are fragile and disabled by default. They may break when page content changes.
- If your application serves content in multiple languages, enable Localization aware to prevent AI tagging from relying on text-content selectors, since visible text changes per locale.
Tip: The heatmap is also available in AI tagging. It highlights popular elements in your application that should be tagged, giving you a visual starting point before you begin capturing.
Tag pages and features
- Navigate to the page in your application that you want to tag.
- Select Start tagging. AI tagging captures the DOM of the current page and generates suggestions in the designer for the page tag and its features. Suggestions may take a few moments to load.
To tag multiple pages in a single session, enable Multi-page capture in Manage settings before you start. See Multi-page capture in Additional settings for details.
Review suggested tags
Suggestions are grouped by page. The page tag itself appears first, either as a new suggestion or as a recommended update to an existing tag (for example, a better name, description, or rule). Feature suggestions for that page follow. New tags are indicated by a green band under the Create heading. Tags that need updates are indicated by a yellow band.
Tip: AI tagging identifies and suggests page tags automatically alongside feature tags. You don't need to tag pages separately.
While the designer is open, hover over a highlighted element in your application and the designer will automatically scroll to the related suggestion in the panel.
- Expand a tag to see the suggested name, element type, rule, description, and product area.
- Select the checkbox next to any page or feature you want to tag. Use Select all to select everything in the list.
- Select Accept selected to apply your selections.
- To start over, select Clear selection, and then start the AI tagging process again.
Review tags for deletion
AI tagging surfaces broken or unused tags as deletion recommendations alongside creation and update suggestions.
- Review deletion recommendations in the designer.
- To accept a deletion, select the tag, then select Accept selected.
Deleting a tag removes it from your subscription and any reports or guides that reference it. Review deletion suggestions carefully before accepting.
Additional settings
Select Manage settings at any time to access the full settings panel. Settings are organised into the following sections.
Capture
- Suggest product areas: AI tagging suggests product areas for tagged features and pages. On by default.
- Fast mode: Uses AI to suggest features from the current page without a full analysis. On by default.
- Multi-page capture: Lets you capture multiple pages in a single session before reviewing suggestions. Off by default. By default, AI tagging captures and suggests tags for one page at a time. When enabled, the button label changes to Start capturing pages. Browse through the pages you want to tag. AI tagging captures each page's DOM as you navigate. When you're done, select Stop and tag with AI to generate suggestions for all captured pages at once.
- Sitewide mode: Suggests which tags should be applied sitewide, if they’re repeated multiple times and/or across multiple pages. On by default.
- Create-only mode: Only suggests features to create — skips existing tags AI tagging would match, update, merge, or remove. Off by default.
AI model
Choose between Claude or Gemini. Claude is only available in the US data region.
Naming convention
Also accessible here, in addition to the recommended action on the landing screen. Describe how AI should name pages and features in your application.
Tagging context
Add any extra context to help AI tag accurately, for example, product terminology, areas to ignore, or what matters most. This is distinct from the naming convention, which controls format and structure. Tagging context controls scope and intent.
Example: Ignore the admin debug bar. Treat the left rail as navigation, not features.
Selector preference
Also accessible here, in addition to the recommended action on the landing screen. Drag to reorder selectors from best to worst. AI tagging tries each enabled selector from top to bottom and uses the first stable match. Toggle off any selectors you don't want used.
You can also add a custom selector using the text field at the top of the list.
Advanced
- Dry run: Always on. No tags are created or modified until you confirm your selections. Full control will be available in a future release.
- Debug: Shows verbose logging to help troubleshoot tagging issues. Off by default.
How AI tagging works
AI tagging captures the DOM of your application's pages through the designer and analyzes the page structure to generate tag suggestions. Suggestions are organized by page and grouped by action type: tags to create, update, or delete.
AI tagging uses a prioritized rule set to determine tag quality: custom HTML attributes first, then element IDs, then weaker selectors. If only a weak selector is available, the suggestion is flagged. You can still accept it, but it's worth asking your engineering team to add a more stable identifier.
By default, AI tagging captures one page at a time. If you enable multi-page capture, AI tagging runs up to five pages simultaneously, with additional pages queued and processed after the initial set completes. Suggestions may take a while to load.
If you've added a naming convention, AI tagging applies it when generating suggestions. Otherwise, Pendo's recommended best practices are used by default. Dismissed suggestions won't be offered again in future sessions.
Frequently asked questions
Does AI tagging work with tag maintenance?
Yes. AI tagging builds on tag maintenance. If AI tagging detects a broken tag or one with no recent DOM match, it surfaces it in the suggestions panel as an update or removal recommendation alongside suggestions for new tags.
Will AI tagging only suggest tags with strong selectors?
No. AI tagging uses a prioritized rule set: custom HTML attributes first, then element IDs, then weaker selectors. If only a weak selector (such as an nth-child) is available, the suggestion is flagged rather than skipped. You can still accept it, but it's worth asking your engineering team to add a custom HTML attribute or element ID for a more reliable rule.
Can I choose which suggestions to apply?
Yes. Use the checkboxes next to each suggestion to select what to accept or reject, or use Select all to apply everything. Nothing is committed to your subscription until you select Accept selected.
Does AI tagging have access to sensitive data in my application?
No. Before any DOM content is sent to the AI for analysis, Pendo sanitises it client-side — in your browser — so sensitive data is removed or masked before it ever leaves your application.
Can I update my naming convention without changing existing tag rules?
Yes. When you update your naming convention, AI tagging gives you the opportunity to review and apply updates to existing tags before accepting.
Does AI tagging work on extension apps?
Yes.
Does AI tagging work on Mobile apps?
Not currently.