Directional attribution for AI, search, and referral traffic

1. 1. Source detection. The ingest layer normalises the referrer signal and looks up its source category against a maintained pattern list.
2. 2. UTM precedence. Explicit UTM parameters take precedence over inferred referrer source. Both are stored; the dashboard renders the canonical source.
3. 3. Confidence labelling. Each row gets a confidence label. Clean signals score higher; stripped or empty signals score lower and are reported as such.
4. 4. Rollup + drill-down. The dashboard shows the workspace-level rollup. The Event Explorer lets you drill into the underlying events. MCP exposes the same surface to Claude.

• No cross-site tracking

  Attribution is computed per workspace, per site. No global identifier follows a visitor across sites.

• No fingerprinting

  The product does not read device entropy, canvas, audio, or fonts. The signal is the referrer + UTM, nothing more.

• No data sale

  Workspace data stays in the workspace. We never sell, license, or share attribution data with third parties.

• Honest confidence

  Stripped referrers land in 'unknown'. We never invent a source for a signal that does not exist.

Questions? info@helperg.com. Privacy detail at /privacy-first-analytics.

What does attribution analytics in WebmasterID actually do?

It categorises each visit by source — AI assistants (ChatGPT, Claude, Perplexity, Gemini, AI Overviews), search engines, social referrers, direct, and unknown — using server-side referrer detection. The result is a directional rollup, with confidence labels per row, that operators read alongside their pipeline and content plans.

Is this paid-ad attribution?

No. Paid-ad attribution involves click IDs, cross-domain joins, and dedicated UTM management. WebmasterID's attribution is referrer-and-UTM-based directional attribution. It is useful for understanding trends and AI-surface coverage; it is not a substitute for an ad-platform attribution report.

How does WebmasterID label confidence?

Each attribution row carries a confidence label derived from the underlying signal. Direct-from-AI-assistant referrals with a clean source pattern score higher; stripped referrers score lower. The label is visible in the dashboard and exposed through MCP so AI clients can explain the directional limits, not gloss over them.

Do you store cross-site identifiers?

No. WebmasterID has no concept of a single user across sites. Attribution is computed per workspace, per site, and joined back to the visit's own session — never to a cross-site profile.

Can I export attribution data?

Yes. The attribution view supports CSV and NDJSON export of the filtered slice. Exports are workspace-scoped and the audit log records them.

How is AI assistant traffic identified?

By matching the referrer signal against a maintained list of AI-assistant referrer patterns. ChatGPT, Claude, Perplexity, Gemini, and AI Overviews are the primary surfaces. When a referrer is stripped or unknown, the visit lands in 'unknown' rather than being speculatively labelled.