AI Search UTM Parameters: The Complete 2026 Reference

← Back to AI Search Attribution

UTM parameters are the most reliable attribution signal — when they exist. The problem with AI search: most engines don’t auto-append UTMs, and the ones that do are inconsistent. This reference covers every known pattern as of May 2026.

Which Engines Auto-Append UTM Parameters

The short answer: almost none.

Engine	Auto-appends UTM?	Pattern when present	Notes
ChatGPT	❌ Never	—
Claude	❌ Never	—
Perplexity	❌ Never	—
Gemini	❌ Rarely	`source=aiovs`	Only on some AI Overview clicks
Copilot / Bing	✅ Sometimes	`msclkid=...`	Microsoft click ID, not a UTM
Google AI Overviews	✅ Sometimes	`source=aiovs`	Not a UTM, but detectable
You.com	❌ Never	—
Grok	❌ Never	—

Practical takeaway: Don’t rely on AI engines auto-appending UTMs. The only reliable way to get UTM-based AI attribution is to pre-tag your own URLs before AI engines surface them.

How to Pre-Tag Your Content for AI Discovery

If you’re creating content targeted at AI citation — comparison pages, FAQ guides, buying guides — embed UTMs in the canonical URLs before you publish:

https://yourstore.com/blogs/moissanite-vs-lab-diamond
  ?utm_source=ai-organic
  &utm_medium=organic
  &utm_campaign=comparison-pages
  &utm_content=moissanite-vs-lab-diamond

When an AI engine cites this URL in a response and a user clicks through, GA4 reads the UTMs and attributes the session correctly.

Limitations of pre-tagging:

Only works for pages where you’ve embedded UTMs before publication
AI engines sometimes rewrite or truncate URLs in their responses — the UTMs may be stripped
Users who copy-paste rather than click lose the UTMs
Organic citations of your existing (non-tagged) pages aren’t captured

Pre-tagging is a best-effort supplement to referrer-based detection, not a replacement.

Complete Detection Pattern Reference

ChatGPT

Referrer domains:
  chatgpt.com
  chat.openai.com          (legacy endpoint, still active)

URL parameters:
  (none auto-appended)

Note: User agent Mozilla/5.0 (compatible; ChatGPT-User/1.0) is
the CRAWLER, not user sessions. Filter separately.

Claude

Referrer domains:
  claude.ai
  claudeusercontent.com    (artifact preview pages)

URL parameters:
  (none auto-appended)

Note: claude.ai referrer reliability ~50% on desktop, ~15% on iOS

Perplexity

Referrer domains:
  perplexity.ai
  www.perplexity.ai

URL parameters:
  (none auto-appended)

Note: Most reliable referrer of the major engines (~85% desktop,
~60% iOS). GA4 default misclassifies as Organic Social — add
custom channel rule before trusting Perplexity numbers.

Gemini

Referrer domains:
  gemini.google.com        (conversational — distinguishable)
  bard.google.com          (legacy, still active)
  google.com               (AI Overviews — NOT distinguishable
                             from organic without source= param)

URL parameters when present:
  source=aiovs             (AI Overview clicks)
  sca_esv=...              (Google experiment parameter)

Detection gap: AI Overview clicks using google.com referrer are
indistinguishable from organic without source=aiovs.

Copilot / Bing Chat

Referrer domains:
  bing.com
  www.bing.com

URL parameters when present:
  msclkid=...              (Microsoft click ID — use for order-level attribution)
  form=MA13FV              (Bing Chat context)
  form=HDRSC1              (Bing header search)

Note: msclkid when present enables stronger order-level attribution
similar to Google's gclid.

You.com

Referrer domains:
  you.com

URL parameters:
  (none auto-appended)

Note: ~90% referrer reliability — behaves like a standard link.

Patterns That Look Like AI But Aren’t

Shopify Shop App

shop.app
checkout.shopify.com

Shopify's own discovery app — real purchase traffic, but not AI search.
Track separately as "Shop App" channel, not AI Search.

AI Crawler Traffic (Not User Sessions)

AI crawlers index your site but don’t buy anything. Filter from session analysis:

User agents to exclude:
  GPTBot
  ChatGPT-User (crawl mode)
  ClaudeBot
  anthropic-ai
  PerplexityBot
  Google-Extended
  Amazonbot
  CCBot

In GA4: Admin → Data Streams → configure bot filtering. If you’re seeing inflated AI traffic numbers with zero purchases, crawler sessions may be leaking into your user-session reports.

Building a UTM Taxonomy for AI Content at Scale

If you’re publishing AI-targeted content systematically, use a consistent taxonomy from day one:

Parameter	Value	What it tracks
`utm_source`	`ai-organic`	All AI-targeted content
`utm_medium`	`organic`	Organic AI (not paid)
`utm_campaign`	`comparison-pages`	Content type cluster
`utm_content`	`[page-slug]`	Specific article

This lets you report on all AI-targeted content in aggregate (utm_source=ai-organic) while drilling into specific campaigns or articles (utm_campaign=comparison-pages). It also gives you a clean signal to separate AI-SEO investment from regular content performance.

Next: Measuring AEO ROI — the metrics that tell you whether schema, FAQ blocks, and comparison pages are actually working.

← Back to AI Search Attribution