GEO measurement and procurement

AI visibility, website traffic and enquiries: an attribution record buyers can audit

Separate AI observations, referred website visits, contact actions and verified enquiries with this eight-field attribution record and supplier acceptance test.

Published by SALIENS · Published 2026-10-04

Treat visibility, visits and enquiries as different evidence

An organisation can be mentioned in an AI answer without receiving a website visit. A person can visit after seeing an answer without the analytics platform identifying that answer as the source. A reader can click an email link without sending a message. These are related stages, but none proves the next one.

A useful GEO report therefore keeps four evidence layers separate: the AI observation, the website acquisition record, the contact action and the verified enquiry. Link them only when a documented identifier or matching rule supports the connection. Otherwise report the stages side by side and leave attribution unknown rather than converting correlation into a lead claim.

Complete an eight-field attribution record

Use one row for each observation or enquiry candidate. Retain raw evidence before classifying it, and record unavailable fields explicitly so a supplier cannot turn missing data into zero performance or full attribution.

AI visibility-to-enquiry attribution record
Field	What to record	What it proves
1. AI observation	Exact question, answer, visible brand mention, own-domain link or citation, date and retained evidence.	What appeared in the tested interface under the stated conditions; not a visit or enquiry.
2. Test conditions	System and interface, market, language, session state, location condition if known, and whether the brand was named.	The boundary of the observation and whether later tests are reasonably comparable.
3. Controlled link	Destination URL and any consistent source, medium, campaign and content parameters on a link the organisation controls.	Which tagged link referred traffic when those parameters reach analytics; not why a person clicked or whether every untagged visit is absent.
4. Website visit	Landing page, session source and medium, timestamp window, consent or collection state, and analytics record ID where available.	A recorded site visit under the analytics configuration; not necessarily an AI-origin visit or a lead.
5. Contact action	Email-link click, form start, form completion event, phone-link click or another defined action, plus event time and route.	That the instrumented action occurred; a click or form start does not prove a message was delivered.
6. Verified enquiry	Inbox, form-receipt or CRM record, received time, subject or need, lawful internal reference and owner.	That an enquiry was actually received; qualification and commercial outcome remain separate stages.
7. Matching rule	The permitted identifier or time-and-route rule used to connect records, exclusions, duplicate handling and reviewer.	How the attribution was made. If the rule cannot distinguish alternatives, the result remains assisted or unknown.
8. Status and confidence	Direct, assisted, unverified or unavailable; evidence source, review date and unresolved gap.	The strength and limits of the conclusion without inventing precision.

Use claims that match the evidence

Google states that traffic from AI Overviews and AI Mode is included within overall Web search traffic in Search Console. Unless the available report or another authorised source provides a supported breakdown, a Search Console click should not be relabelled as a click from a specific AI feature. A retained answer showing a SALIENS link is visibility evidence; a dated Search Console click is search-traffic evidence; they are not automatically the same event.

Use direct attribution only when the evidence chain is specific enough to exclude material alternatives. Use assisted attribution when the buyer reports the touchpoint or several records support a contribution but cannot establish a unique cause. Use unverified when there is a plausible relationship without sufficient matching evidence, and unavailable when the necessary source is not connected or accessible.

Evidence and the strongest supportable statement
Available evidence	Supportable statement	Do not claim
Retained AI answer with a brand mention	The brand was mentioned in this test under these conditions.	The answer generated a visit or enquiry.
Retained AI answer with an own-domain link	The website was linked or cited in this test.	The link was clicked.
Analytics session with a reliable tagged source	The tagged source referred a recorded session.	The visitor saw a particular AI answer unless that is separately evidenced.
Instrumented contact click	The contact route was clicked.	An email, call or enquiry was completed.
Received enquiry plus a documented match	An enquiry was received and attributed under the stated rule.	The lead was qualified, won or caused solely by GEO without further evidence.

Design the website evidence before reporting outcomes

For links the organisation controls, use a stable, case-consistent UTM convention. Google Analytics documents that source, medium and campaign parameters identify referring campaigns in the Traffic acquisition report, while inconsistent naming fragments the data. Record the destination and naming convention in the delivery brief, and preserve untagged or stripped-referrer traffic as unknown rather than reallocating it to a preferred channel.

Define contact events by their real completion point. Google Analytics automatically collects ordinary web interactions such as outbound clicks when configured, and its recommended lead events distinguish an initial generated lead from later qualified, working and converted stages. A measurement plan should mirror those distinctions: contact-link click, form success, received enquiry, qualified opportunity and customer are separate records.

Write the supplier acceptance test

Ask the supplier to provide the measurement dictionary, question and observation register, UTM convention, event definitions, enquiry source of truth, matching rule, duplicate handling, reporting window, unavailable states and a sample audit trail from raw evidence to the final claim. Check at least one direct, assisted and unattributed example.

Reject a dashboard that combines mentions, citations, visits, contact clicks and enquiries into one unexplained conversion number. Also reject retrospective attribution that has no retained source, timestamp, identifier or declared rule. Delivery can be verified through published pages, technical checks and measurement configuration; visibility and commercial outcomes require their own evidence.

Can another reviewer reproduce the observation and trace the claimed source?
Does every number name its system, date range, definition and denominator?
Are clicks, received enquiries, qualified opportunities and customers separate?
Are unattributed and unavailable records preserved rather than forced into a channel?
Does the report state material consent, collection and matching limitations?

How SALIENS can build a defensible measurement chain

SALIENS provides GEO audits, AI visibility monitoring, content and technical improvements, and ongoing GEO delivery for UK and international organisations, including scoped English and Simplified Chinese work. We can define the observation register, review controlled links and contact routes, document event and enquiry stages, and report delivery, visibility, traffic and commercial evidence separately.

Email info@oxfordintelligence.co.uk with your website, priority markets and languages, current analytics and enquiry systems, main contact routes and the buyer questions you need to measure. SALIENS is operated by Oxford Intelligence Limited. Measurement does not guarantee crawling, indexing, mentions, citations, rankings, visits or enquiries.

Sources

Prepared with AI assistance using the linked sources and SALIENS service information.