Define the observable journey

A referral report measures visits your analytics can identify, not everyone who encountered an AI answer. Start with a simple journey: answer exposure, link click, landing visit, and meaningful action. Your systems may observe only the final stages. Keep that gap explicit when presenting results. Decide which business event matters, such as a completed onboarding step or a qualified inquiry. Page views alone cannot establish commercial value. Write an event definition with a clear trigger, and check that the event is not accidentally emitted twice during navigation or form submission.

Inspect acquisition data

Review source and medium values in your analytics rather than assuming a fixed list captures every AI visit. Inspect landing pages, referral hosts, campaign parameters, and redirects. Preserve raw source values before grouping them into an AI category so classification can be audited. Test representative links yourself where permitted, but do not assume a single browser test explains every device or app. Redirects and browser behavior can affect the recorded visit. Keep direct and unattributed traffic separate rather than automatically assigning unexplained growth to AI discovery or to a recent article.

Distinguish citation and traffic

Bing's citation reporting is not a click report. A cited page can support an answer without receiving a visit. Likewise, an AI referral might arrive on a page other than the one you expected. Compare the datasets by URL and period while retaining their different units. Hypothetical example: a guide receives 300 observed citations and 12 identifiable referral visits in a period. This does not establish a 4% click-through rate because the citation and visit populations may differ. Report both counts independently and explain the available attribution boundaries.

Measure useful actions

Segment identifiable AI referrals by landing page and completed actions. A small stream that consistently brings qualified prospects can matter more than many visits to an unrelated definition. Review the actual conversion path and whether visitors reach an appropriate next step. Use a voluntary discovery question in signup or sales conversations to supplement analytics, keeping it distinct from tracked attribution. Self-reported discovery can reveal influences that referrer data misses, but memory is imperfect. Do not combine the two into a single exact revenue number without a documented attribution model.

Create a conservative dashboard

Show identifiable visits, engaged sessions under your chosen definition, completed events, and attributed revenue where available. Display the date window and explain classification rules. Add an unattributed category instead of presenting complete coverage. Review whether a traffic change coincides with new tracking code, consent settings, or redirects before attributing it to content. Preserve a changelog of analytics configuration. The most useful report shows which articles introduce qualified users and where their journey stalls. It gives the team a testable improvement, such as a clearer next step, rather than a speculative total audience estimate.

PUT IT TO WORK

Connect observable AI referral visits to useful actions while keeping missing attribution visible.

Sources & further reading

Official documentation informs this guide. Examples are illustrative, and product interfaces or limits can change.

THE WORKSPACE WE'RE BUILDING

From search signals to a useful first draft.

SERP.so is in development. Join the waitlist for a workspace that connects research, content review, and publishing.

Join early access ↗