How can you separate AI-referred visits from brand mentions?
A company name appearing in an AI answer does not establish that anyone visited its website. Conversely, a session attributed to an AI service does not reveal how often that service recommended the company. A useful business report keeps these events separate. This article does not report actual Two Way Medya traffic results. It uses a hypothetical example to explain a measurement record that a business can maintain.
1. Create three evidence rows
The first row records a mention: service, question, language, date and the name observed. The second records a source link: which exact page was cited in the answer? The third records a website visit: which source and destination appeared in analytics? An answer screenshot may support the first two rows, but it cannot establish the third. Several people may see an answer, or nobody may click its link. Do not estimate unobserved answer impressions from a screenshot or turn one citation into a reach figure.
2. Use session scope for the visit question
Google Analytics distinguishes User acquisition from Traffic acquisition: first-user source and session source have different scopes. For “Which sources brought sessions during this period?”, examine session-scoped source/medium. Confirm the property, reporting window and time zone first. Export the source values actually present before applying a classification. A prepared list of AI domains is a starting point, not proof that every possible referral has been captured. Verify what observed sources represent and date the rule used to group them.
3. Review the landing page and next action
Imagine a few sessions classified as AI referrals reaching an English service guide. Look at the landing page, content language and defined key events alongside the session total. A click on a contact button is not proof that a message was sent, and a successfully received form is not a sale. Keep the business’s qualified-enquiry assessment separate. With a small sample, one visitor can change a percentage dramatically. Show raw counts and describe the observed activity without extrapolating a general conversion rate.
4. Leave unknown sources unknown
Google describes direct/none as traffic without a clear referral source. An increase in direct traffic should therefore not be reassigned wholesale to AI activity. Equally, an empty AI source row does not prove that no answer mentioned the business. Measurement setup, permission choices, browser behaviour and redirects can affect what is observed; test these on the actual website. Do not compensate for uncertainty by collecting more personal information. Preserve an explicit unknown category rather than presenting an estimated source as a measured fact.
5. Repeat the comparison with the same rules
Keep period length, source classification, content language and event definitions consistent. If a rule changes, record its date rather than presenting the old and new totals as an uninterrupted series. The AI visibility monitoring guide covers answer samples; this article complements it by examining recorded visits. Connect SEO work to the pages and needs actually observed. A rise in one referral source does not by itself establish that a particular content change caused it, nor does it guarantee future citations.
Frequently asked questions
Does more direct traffic prove better AI visibility?
No. Traffic with an unknown source cannot automatically be reassigned to AI.
Should citations and visits have the same count?
No. A cited link may receive no clicks or several visits. These are different events.