Search logs should ignore OpenSearch placeholder terms and optionally tag AI-agent searches

Feature
,
1

Summary

The Trending Search Terms report is polluted by OpenSearch URL template placeholders (e.g. {searchTerms}) that are not real searches. This makes the dashboard misleading (often showing {searchTerms} as the top “search term” with 0% CTR).

Additionally, as AI-mediated browsing grows, it would be useful to optionally tag AI-agent initiated searches separately in search analytics.

Problem 1: OpenSearch placeholder noise in SearchLog

On my site, {searchTerms} appears as the #1 trending search term, with thousands of entries and 0% CTR. These entries come from crawlers/bots (e.g. Googlebot, Bingbot, etc.) crawling /opensearch.xml and calling the search endpoint with the literal placeholder string instead of a real query.

This has been discussed before:

But the placeholder terms still show up in analytics.

Steps to Reproduce

  1. Enable/serve OpenSearch (e.g. /opensearch.xml) for the site
  2. Allow crawlers to access it (default public behavior)
  3. Wait for crawlers to fetch the OpenSearch template and hit the search endpoint
  4. View Admin → Reports → Trending Search Terms
  5. Observe placeholder values like {searchTerms} dominating the report

Expected Behavior

Placeholder/template strings used by OpenSearch clients should not be recorded as real searches and should not appear in Trending Search Terms.

Actual Behavior

Placeholder strings (e.g. {searchTerms}) are saved into SearchLog and show up as real search terms, polluting analytics.

Proposed Fix

Filter known OpenSearch placeholder strings before logging to SearchLog, for example:

  • {searchTerms}
  • {search_term_string}

(If there are other common variants, adding them would be fine too.)

This is effectively “bot noise,” never a legitimate human query, and it breaks the usefulness of the report.

Bigger Opportunity: AI-era search analytics (Optional / Medium–Long Term)

The {searchTerms} issue highlights a broader gap: a growing portion of searches are performed by AI agents on behalf of users (e.g. when a user asks an assistant “search this forum for X”). Those searches can represent real user intent, but currently they’re mixed with all other traffic and hard to understand.

Medium Term (Optional)

Tag searches likely initiated by AI agents using User-Agent heuristics (examples only):

  • ChatGPT browsing / agent UA variants
  • Perplexity bots
  • Claude-related agents
  • Google AI-related UAs (e.g. GoogleOther, etc.)

This wouldn’t need to be perfect—just good enough to give admins visibility.

Long Term (Optional)

Add an “AI Search” filter/tab in the Trending Search Terms report so admins can see:

  • Human searches
  • AI-agent searches
  • Everything combined

Why This Matters

  • Placeholder pollution makes the dashboard less trustworthy and can dominate “Trending”
  • Admins shouldn’t have to manually clean analytics or ignore broken top entries
  • AI-mediated search traffic is increasing, and site owners benefit from visibility into those intents