Track identifiable AI referrers, landing pages, and conversions in a versioned channel definition. Keep crawler requests, prompt visibility, and direct/unknown traffic separate, and report confidence instead of reallocating ambiguous traffic to AI.
Table of contents
Next step
Use the free audit or a strategy call to map these frameworks to your current website, market, and revenue goals.
Start free auditRelated services
Related sample
Direct answer
AI referral traffic is the website activity that can be attributed to a click from an AI-assisted product through available source/referrer data. Track identifiable referrers, landing pages, conversions, and lead quality in a versioned channel definition. Keep crawler requests, prompt visibility, and direct/unknown traffic separate. Some AI influence is not directly observable, so report confidence rather than reallocating ambiguous traffic to AI.
Key takeaways
| Signal | Where it appears | What it can support |
|---|---|---|
| AI answer mention/citation | Versioned prompt sampling | Visibility and answer accuracy, not site visit |
| Identifiable referral session | Analytics/server request with source evidence | Attributable site visit under the measurement rules |
| AI crawler request | CDN/server log | Machine access, not human referral or conversion |
| Direct/unknown landing | Analytics with no usable source | An unattributed visit; AI origin is one possible explanation |
Prompt visibility and citation measurement belong to dedicated AI-visibility tracking; this framework owns website referral and conversion evidence specifically.
Google Analytics uses collected campaign and traffic-source fields to process and report source dimensions.
Start with observed source/medium values in the site's own data. Map each known referrer to an operator/product and keep the raw value available. Do not publish a 'complete' regex without a date and owner; platform domains and browser behavior change.
Default channel groups cannot be edited, while custom groups can be created for rule-based analysis — keep the default group alongside the custom AI referral view.
Google's current Search Central guidance says sites appearing in AI Overviews and AI Mode are included in overall Search Console traffic under the Web search type. Google also announced in June 2026 that it was introducing and testing new controls and performance insights for generative AI Search, so implementation must verify what is actually available in the property at that time.
Sites appearing in AI Overviews and AI Mode are included in overall Search Console traffic under the Web search type under current guidance.
Google announced in June 2026 that it was introducing/testing new controls and performance insights for generative AI Search.
Do not infer a separate Google AI referral channel from standard 'google / organic' traffic unless an official report or preserved parameter distinguishes it.
Analytics may report direct/none when usable source information is unavailable or lost. Deep-page direct landings, branded-search changes, or an increase in unattributed sessions can support a research question. They do not prove that ChatGPT, Gemini, Claude, Perplexity, or another system generated the visit.
Analytics may report direct/none traffic when usable source information is unavailable or lost, and Google provides dedicated troubleshooting guidance for this.
| Use | Avoid |
|---|---|
| "Known AI referral sessions increased." | "AI drove 40% of direct traffic." |
| "Direct landings to these pages increased; origin is not observable." | "Direct traffic is secretly ChatGPT traffic." |
| "Branded demand moved after the campaign; causal source is unresolved." | "AI search caused the branded lift." |
Server logs may show OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, or other agents. Those are machine requests. They do not create analytics sessions or demonstrate that a human saw, clicked, or converted from an answer.
| Confidence tier | Signal | Conclusion allowed | Conclusion prohibited |
|---|---|---|---|
| A — attributable | Preserved known referrer/source under current rule | Session/conversion was attributed to that source | AI answer exposure or incremental causality |
| B — platform-ambiguous | Source identifies a broad ecosystem but not the AI surface | Activity came from the ecosystem under the rule | Specific AI feature caused it |
| C — correlated | Branded/direct/landing behavior aligns with AI visibility timing | A relationship worth investigating | Visit or conversion originated in AI |
| D — unknown | No usable source evidence | Unattributed activity exists | Any platform credit |
| E — unmeasurable | No observable visit or identifier | State the limitation explicitly | Zero influence |
Do not publish a universal claim that AI referral visitors convert better. Conversion depends on product, task, attribution, period, and sample.
Taskcover builds the versioned channel rules and data dictionary behind defensible AI referral reporting.
No. Analytics may report direct/none when usable source information is unavailable or lost. It is one possible explanation among several, not proof of AI origin.
No. Crawler requests in server logs are machine access and should not be reported as referral traffic. Referral traffic requires an identifiable human session from a known source.
Under current guidance, sites appearing in AI Overviews and AI Mode are included in overall Search Console traffic under the Web search type, though Google has announced new controls and insights that must be verified in-property at implementation.
No. Start from your own site's observed source/medium values, keep the rule dated and owned, and update it as platform domains and behavior change.
Taskcover implements the AI Referral Data Dictionary and versioned channel rules behind accurate reporting.
Google Analytics Help · Accessed: July 13, 2026
Google Analytics Help · Accessed: July 13, 2026
Google Search Central · Accessed: July 13, 2026
Google · Accessed: July 13, 2026
Google Analytics Help · Accessed: July 13, 2026
AI crawler governance means deciding access by operator and purpose — not applying one rule to every AI bot. Separate search/indexing, model training, and user-requested retrieval, and verify what actually happens at the edge.
When organic visibility grows but qualified leads do not, verify the measurement first, then trace the journey from query to sales acceptance through the seven-leakage Search-to-Lead tree.
When an AI answer contains wrong information about a brand, preserve the exact response, classify the severity, trace the likely source, correct owned and third-party facts, escalate where documented, and re-test on a schedule.
Use the free audit or a strategy call to map these frameworks to your current website, market, and revenue goals.
Taskcover uses necessary storage for security and forms. You can choose whether we use preferences, analytics, and marketing measurement.
