AI referral traffic · Citation tracking · Fixed test questions
Look beyond rankings.
Track useful signals.
Measure what matters.
Understand how AI-related discovery connects to your business. KOAT monitors identifiable referral traffic, observed brand mentions and source citations, then reviews those signals alongside relevant enquiries.
How do you measure results beyond traditional rankings?
Combine identifiable website visits, customer actions and qualified enquiries with documented checks of AI answers. Track mentions and citations separately, keep test conditions consistent and compare equivalent periods. Report the limits of each data source alongside the findings.
A practical measurement framework
Connect discovery
to customer actions.
Mentions and citations
Record whether your business is named or your website is cited in the agreed set of AI-answer tests.
Identifiable referral traffic
Review sessions attributed to verified AI-related sources and the pages those visitors enter.
Enquiries and lead quality
Investigate relevant customer actions and, where reliable records exist, whether enquiries fit your services.
Keep the measures distinct
An observed citation is not a recorded visit, and a visit is not a qualified lead. Report each stage separately so changes have a clear meaning.
AI referral traffic
Review the visits
you can actually identify.
Inspect traffic sources
In GA4, session source and medium help investigate where recorded sessions came from. We review actual values before defining the sources included in an AI traffic report.
ChatGPT search referral URLs include utm_source=chatgpt.com. Whether this produces
a usable record on your website depends on the link journey
and your measurement setup.
| Question | What to review |
|---|---|
| Which sources are recorded? | Verified session source and medium values. |
| Where do visitors arrive? | Landing pages for the defined traffic segment. |
| Do they interact with relevant content? | Engagement measures and appropriate recorded events. |
| Do they take an enquiry action? | Verified form, booking or contact events. |
| Are those enquiries suitable? | Lead qualification records, where available. |
| What changed during the period? | Content, campaigns, tracking and other relevant changes. |
Attribution limits
Explain what your
analytics cannot establish.
Some discovery happens without a tracked click
Someone may read an answer, remember your business and visit later through another route. Website analytics does not provide a complete record of that earlier exposure.
Direct traffic represents sessions without a clear referral source. It should not automatically be classified as AI traffic.
Use customer feedback as additional context
An optional “How did you hear about us?” question can help explain discovery. Keep self-reported answers separate from analytics attribution.
Describe uncertain journeys honestly. Avoid assigning a specific source when the available evidence cannot support it.
Fixed test questions
Build a repeatable
view of observed visibility.
Select a core set of questions relevant to your customers, services and markets. Repeat them using documented conditions and keep the original responses available for review.
Discovery questions
Test questions about problems, services and providers without inserting your brand into the prompt.
Brand questions
Review how your business is described when someone asks about it directly. Check the factual information.
Market and language groups
Separate Dutch and English questions and any relevant market-specific tests so comparisons remain understandable.
| Field | Recorded information |
|---|---|
| Question | Exact wording, language and intent group. |
| Environment | Platform, visible mode, search status where observable, session context and relevant settings. |
| Date | When the check was performed. |
| Evidence | Saved answer or screenshot and visible source URLs. |
| Outcome | Brand mention, website citation and relevant accuracy findings. |
| Test status | Completed response, error or unavailable feature. |
Keep errors outside the visibility denominator
Record failed tests separately. An error or unavailable feature is different from a valid answer that does not mention your business.
Citation and mention tracking
Use clear definitions
for every reported metric.
| Measure | Definition | Interpretation limit |
|---|---|---|
| Observed mention rate | Valid tested responses naming the business divided by all valid responses in the defined group. | Describes that sample, rather than platform-wide exposure. |
| Observed citation rate | Valid tested responses linking to your website divided by all valid responses in the defined group. | A citation does not establish that anyone clicked it. |
| Identifiable AI sessions | Recorded sessions matching the verified source definition. | Does not capture every AI-influenced visit. |
| Enquiry events | Verified enquiry actions under an agreed counting rule. | Events require validation and may need deduplication. |
| Qualified enquiries | Enquiries meeting documented business criteria. | Depends on reliable lead records and source linkage. |
Show counts alongside percentages
Five citations in twenty valid tests is a 25% observed citation rate for that sample. Publishing both figures makes the size of the sample visible.
Keep the comparison consistent
Compare like-for-like question groups and conditions. Document changes to prompts, platforms or measurement settings that affect interpretation.
Reporting that supports decisions
Explain what changed
and what needs attention.
Evidence
Show the relevant counts, source records and test responses behind the reported observations.
Context
Record website updates, campaigns, measurement changes and known data gaps during the reporting period.
Actions
Identify what should be checked or improved next, with an owner and a practical verification step.
Investigate before attributing a change
A rise in observed citations after publishing new content is an observation worth reviewing. It does not, by itself, prove that the content change caused the increase.
Our approach
Establish a baseline.
Review changes over time.
-
Define
Agree on business goals, source definitions, platforms, questions and reporting frequency.
-
Validate
Review available analytics, customer-action tracking and the evidence needed for the baseline.
-
Monitor
Repeat the agreed checks and collect comparable traffic, citation and enquiry observations.
-
Review
Interpret changes, explain limitations and agree on the next actions supported by the findings.
Scope and investment
What does AI visibility
monitoring cost?
The scope depends on the number of platforms, questions, languages and repeat checks, together with analytics setup and reporting requirements.
We define the baseline work, monitoring schedule and deliverables before preparing a quote.
What helps us get started?
- Your website and priority services.
- The markets and languages you want to monitor.
- Available analytics access or relevant exports.
- Your existing enquiry and booking tracking.
- Lead qualification criteria and available records.
- Recent changes that may affect the baseline.
Frequently asked questions
Monitoring beyond rankings,
explained clearly.
How do you measure results beyond traditional rankings?
Combine documented visibility checks with identifiable website visits and relevant customer actions. Keep citations, traffic and qualified enquiries separate.
Can GA4 identify all traffic influenced by AI?
No. It can help analyse recorded visits with usable source information. Discovery without a tracked click and later visits through other routes may remain unattributed.
Should direct traffic be counted as AI traffic?
No. Direct traffic lacks a clear referral source. Classifying it as AI traffic would require additional supporting evidence.
What is the difference between a mention and a citation?
A mention names your business. A citation links to a source. Your website may be cited without the business being named, so we track the two separately.
Why use fixed test questions?
A consistent core question set helps compare observations over time. Documented conditions make changes easier to interpret.
Will the same question always produce the same answer?
No. Answers can vary. Repeat checks and retain the responses so the report shows observed patterns and variation.
Can you monitor Dutch and English questions?
Yes. We define separate language groups and interpret results according to the intended audience and market.
Can you connect AI traffic to qualified enquiries?
Where the measurement setup and lead records support it, we can investigate that connection. We document gaps and avoid assigning unsupported sources.
How often should monitoring take place?
Set a schedule according to your priorities, budget and pace of change. Agree on a baseline and review frequency before collecting recurring results.
Does monitoring guarantee improved visibility?
No. Monitoring provides evidence for decisions. Improvements require appropriate action, and platform selection remains outside our control.
