Start with buyer problems
I shaped a prompt research set around integration questions, product-owner decisions, and relevant markets, informed by the broader search and customer-priority work.
TRANSAK · AEO RESEARCH & DISCOVERY MEASUREMENT
Understand discovery beyond Google rankings.
TL;DR
Connect buyer-question research with a separate view of AI referral traffic, so a search strategy can account for both.
Delivered: A commercially focused prompt research set and a dated view of AI referral traffic, with a clear distinction between research, visits, and attribution.
Problem
Traditional search reporting shows only part of how someone might discover an infrastructure provider. AI-generated answers introduce a different set of questions, sources, and measurement limits.
The research needed to follow questions a relevant buyer would ask. It also needed to distinguish appearing in an AI answer from receiving a visit, a form submission, or a commercial result.
Process
Make the research reviewable
The delivered sheet contained 100 prompts, their topics, and what each was intended to measure, plus fields for sales feedback. Questions outside serviceable markets were flagged for review.
Establish a separate channel baseline
I examined GA4’s reported AI Assistant channel alongside Google Search data. This made it possible to discuss different discovery channels without treating an AI referral as proof of citation share or revenue.
Results
A commercially focused prompt research set and a dated view of AI referral traffic, with a clear distinction between research, visits, and attribution.
100 prompts prepared
A structured research deliverable for sales review, rather than an undifferentiated list of crypto questions.
Separate channel reporting
AI Assistant sessions could be reviewed alongside Google Search clicks without combining their different signals into one growth claim.
Defined measurement boundaries
Referral sessions describe visits classified by GA4. They do not count every AI answer, establish citation share, or isolate the effect of this work.
Evidence & scope
The September client-review record documents the 100-prompt deliverable. The chart uses the previously verified, read-only GA4 and Search Console extract retrieved on 4 October 2026.
Scope. The traffic shown is sitewide, observed data. It is not lift attributed to the prompt research, and the case does not claim that the proposed tracking programme was fully implemented.
SITEWIDE CONTEXT
AI discovery is a measurable part of the picture.
Transak’s monthly AI Assistant sessions approximately doubled from July to September 2026, growing 100.2%.
RELATIVE AI ASSISTANT SESSIONS · JULY = 1×
View the data table
| Month | Relative sessions |
|---|---|
| Jul 2026 | 1.00× |
| Aug 2026 | 1.72× |
| Sep 2026 | 2.00× |
+100.2%AI Assistant sessions · July to September
−28.4%Google Search clicks · July to September
How to read these results
Over the same months, Google Search clicks fell 28.4%. The AI channel grew alongside a decline in Google traffic. That split supports measuring the channels separately and reviewing content against buyer intent.
These are sitewide observed results, not a controlled test or traffic attributed solely to my work. AI Assistant is GA4’s reported channel classification; it does not measure all AI exposure or prove citation share, leads, or revenue. Other marketing activity and changes in search demand can affect the results.
Data source. Google Analytics Data API, Transak property. Metric: sessions. Dimension: sessionDefaultChannelGroup = AI Assistant. Full calendar months, 1 July to 30 September 2026. Google Search Console domain property, Web search, final data, for the Google click figures. Retrieved 4 October 2026.
Calculation. Each month’s sessions are divided by the July baseline. Percentage change = (September ÷ July − 1) × 100. Calculations use underlying values before rounding.
GET IN TOUCH
Let’s talk about
your search strategy.
Send me your website and what you want to achieve.
We’ll work out where I can make a difference.