Working note · October 2026 · Work
The Invisible Part of the Funnel, AI Influence Beyond Referral Traffic
Some of the research and shortlisting may happen before a person visits the website. The later visit can look direct, branded or unremarkable in analytics.
The visit may be late in the decision
Referral traffic is useful when it exists. But it does not show every moment in which a brand becomes known, compared or quietly ruled in. Someone may ask an AI system for options, inspect a few answers, remember one name and only later search for that brand directly. The web analytics record begins at the visit, not at the earlier encounter.
This is not proof that every direct visit was influenced by AI. It is a research question: when discovery happens inside an answer interface, what trace remains once the person reaches a website through another route?
Measure the gap, then test it
A reasonable starting point combines signals rather than treating any one of them as conclusive. Branded search patterns, direct traffic, assisted conversions, location or query changes, customer interviews and controlled comparisons can all contribute evidence. The work is to define a hypothesis before looking at the dashboard.
For example: if a brand is made more intelligible for a specific set of needs, does the later mix of branded discovery, direct visits and qualified enquiries change differently from a comparable baseline? The answer will depend on timing, audience, seasonality and the rest of the marketing environment. That is exactly why this should be treated as a test design, not a shortcut to attribution.