Which AI assistants are detected?

ChatGPT, Claude, Perplexity, Gemini, and AI Overviews are the primary surfaces detected via referrer normalisation. Smaller or new AI surfaces are added to the maintained pattern list as they appear. When a referrer cannot be confidently attributed to an AI surface, the visit stays uncategorised rather than being invented.

How is AI search traffic identified?

Server-side referrer detection. The ingest layer normalises the referrer signal and matches it against a maintained list of AI-assistant referrer patterns. UTM parameters take precedence when present; both are stored and the dashboard renders the canonical source.

Is the detection probabilistic?

No. Detection is rule-based against known referrer patterns. Each row carries a confidence label so operators read with the right amount of trust — clean signals score higher, stripped signals lower. There is no black-box scoring.

What about AI assistants that strip the referrer?

Stripped referrers land in 'unknown' or 'direct'. WebmasterID does not speculatively label these as AI. When a surface is known to strip referrers under certain conditions, the FAQ in the product docs records the caveat.

Can I combine AI search analytics with AI crawler analytics?

Yes. Both signals are stored in the same event store; the dashboard joins them on the AI Visibility view. You can see the crawler timeline next to the resulting human-referral timeline per AI surface.

Does AI search analytics work with the Claude / MCP integration?

Yes. The same surface is exposed through MCP so Claude can read the AI search slice on your behalf. The integration is read-only, workspace-scoped, and audit-logged.

AI search analytics

ChatGPT, Claude, Perplexity, Gemini — see who is actually sending traffic

WebmasterID is AI search analytics for the AI-search era. Per-assistant detection of human referrals from AI surfaces, joined back to the underlying crawl coverage on your site. Privacy-first, deterministic, operator-focused.

Start free See pricing AI referrals detail

Use cases

What AI search analytics is for

The directional view of AI-assistant traffic, surfaced honestly. Useful for editorial prioritisation, AI-search-readiness checks, and weekly reporting.

See which AI assistant sent traffic this week

Per-assistant rollup: ChatGPT, Claude, Perplexity, Gemini, AI Overviews. The same week-over-week comparison you would do with Search Console.

Confirm an article is appearing in AI answers

When humans arrive from AI assistants on a specific article, it is a signal the article is being surfaced in answers — not a guarantee, but a useful directional reading.

Cross-check AI traffic against AI crawl coverage

When a crawler from one AI surface appears, does the assistant from that surface eventually send humans? AI search analytics answers the join.

Brief content teams on AI-search readiness

Per-pathname coverage of AI-assistant referrals. Useful for editorial prioritisation in the AI-search era.

Investigate a missing AI surface

If a surface that should be sending humans is absent, the absence is the finding. The product never invents AI traffic that did not happen.

Hand the data to Claude through MCP

Claude can read the AI search slice and summarise. Operator decides what to do next.

How WebmasterID helps

A small, deterministic pipeline for the AI-search signal

Detection is rule-based at ingest. The dashboard is the read surface; the Event Explorer is the drill-down surface; MCP is the AI-assisted read surface.

1. Referrer normalisation. The ingest layer normalises the referrer signal so AI surfaces stay identifiable as their referrer formats change.
2. Per-assistant categorisation. ChatGPT, Claude, Perplexity, Gemini, AI Overviews. One bucket per recognised surface; unknown referrers stay 'unknown'.
3. Confidence labelling. Each row carries a confidence label — clean signal vs stripped referrer. Operators read with the right amount of trust.
4. Joined to crawl coverage. The AI Visibility view joins the crawler timeline with the resulting human-referral timeline per surface.

Feature

The per-assistant breakdown

One panel per AI surface. The product reports what it sees and does not invent the rest.

ChatGPT

Detection of human visits arriving from ChatGPT, including the share-link variants. Joined to GPTBot crawl coverage where applicable.

Claude

Detection of human visits arriving from Claude surfaces (claude.ai and similar). Joined to ClaudeBot crawl coverage.

Perplexity

Detection of human visits arriving from Perplexity. Joined to PerplexityBot crawl coverage.

Gemini and AI Overviews

Detection of human visits arriving from Gemini and from AI Overviews surfaces in Google search results, where the referrer is identifiable.

Privacy & trust

An AI-search signal without surveillance

The signal is the referrer. The product does not need to know who the visitor is to know which AI surface sent them.

No cross-site identifiers
A visit from Claude on site A is unrelated to a visit from Claude on site B. There is no global user concept.
No fingerprinting
Device entropy, canvas, audio, and fonts are never read. The signal is the referrer + UTM.
Honest 'unknown'
Stripped or empty referrers land in 'unknown'. We never invent an AI source for a signal that does not exist.
No model training
We do not train shared models on workspace data. AI search analytics is observational, not extractive.

Questions? info@helperg.com.

FAQ

AI search analytics, answered

Which AI assistants are detected?: ChatGPT, Claude, Perplexity, Gemini, and AI Overviews are the primary surfaces detected via referrer normalisation. Smaller or new AI surfaces are added to the maintained pattern list as they appear. When a referrer cannot be confidently attributed to an AI surface, the visit stays uncategorised rather than being invented.
How is AI search traffic identified?: Server-side referrer detection. The ingest layer normalises the referrer signal and matches it against a maintained list of AI-assistant referrer patterns. UTM parameters take precedence when present; both are stored and the dashboard renders the canonical source.
Is the detection probabilistic?: No. Detection is rule-based against known referrer patterns. Each row carries a confidence label so operators read with the right amount of trust — clean signals score higher, stripped signals lower. There is no black-box scoring.
What about AI assistants that strip the referrer?: Stripped referrers land in 'unknown' or 'direct'. WebmasterID does not speculatively label these as AI. When a surface is known to strip referrers under certain conditions, the FAQ in the product docs records the caveat.
Can I combine AI search analytics with AI crawler analytics?: Yes. Both signals are stored in the same event store; the dashboard joins them on the AI Visibility view. You can see the crawler timeline next to the resulting human-referral timeline per AI surface.
Does AI search analytics work with the Claude / MCP integration?: Yes. The same surface is exposed through MCP so Claude can read the AI search slice on your behalf. The integration is read-only, workspace-scoped, and audit-logged.

Related cluster pages

AI search analytics sits next to AI visibility and attribution.

AI visibility analytics

The crawler side of the same signal.

Attribution analytics

The directional attribution surface AI search analytics sits on.

Bot intelligence

The deterministic bot categorisation that powers AI visibility.

Event Explorer

Drill into the underlying events behind a per-assistant rollup.

AI-assisted analytics

Read AI search data with Claude alongside you, through MCP.

Multi-site analytics

Roll AI search coverage up across the workspace's full site list.