Measure AI traffic: Do AI referrals become interested customers?

What counts as AI traffic here?

AI traffic here means identifiable website visits following a link from an AI service. A mention without a click is not a website visit. A visit is not yet an enquiry or a purchase.

Our AI visibility guide examines stored answers. This article starts later, with your own website measurement system. The two views can complement one another, but their figures cannot simply be converted into one another.

An answer might mention your brand without including a link. Someone could type your name later or return on a different device. Your website will not necessarily receive an attributable AI referrer. Describe the result as “identified referrals”, not “all influence from AI”.

Prepare one landing page and a genuine goal event

Choose an important page and a concrete outcome, such as a successfully completed registration. Use an appropriate test to check that the event is recorded after completion. Clicking “Try it” is a different action and should have a different definition.

Use analytics that are already appropriately configured and reviewed for your circumstances. This guide activates no cookies, sends no data to Google and does not replace a consent assessment. Keep personal information out of URLs and event parameters.

Record that definition in your protocol. Without a stable definition, two weeks can look different simply because the tracking changed.

Check ChatGPT referrals in GA4

If GA4 is already set up, open Traffic acquisition. Navigation may vary with the interface language and custom report collection. Select a session-scoped dimension such as Session source / medium; the official report documentation explains its scope.

1. Set a fixed period and display the session source.
2. Look for domains actually present in your data, such as chatgpt.com or perplexity.ai. These examples are not a complete or permanent provider list.
3. Read the actual values, including the medium. Do not classify every unfamiliar source as AI.
4. Use the matching rows as a small comparison group and document your filter.

Decide on a combined filter after reviewing your own data. Check each addition against known rows. A broad term such as “ai” would capture unrelated sources.

Which landing pages help visitors take the next step?

For the identified source group, examine the entry pages. Are people looking for a comparison, instructions or prices? Check whether the page answers that expectation promptly and offers a clear next step.

A measurement guide can point to methodology and a relevant product trial. A technical IndexNow article should first solve the setup problem. An aggressive purchase message before the instructions can miss the reason someone arrived.

Record one possible improvement per page: explain a confusing term, add a missing condition or link to the right follow-up guide. The customer-question content map helps identify unanswered follow-up questions.

A small example: Twelve sessions and one registration

The following figures are invented calculation data, not Nalqerio customer results. During a fixed period, twelve sessions are attributed to an identified AI source. One of those sessions records a completed registration.

The denominator is sessions, not everyone who saw an AI answer. For this calculation, use sessions containing the event, not an unchecked sum of potentially repeated events. With twelve visits, one additional registration changes the percentage substantially.

Also examine whether registrations match your intended customer and whether the product is subsequently used. One outcome establishes neither a stable rate nor a causal effect from a particular content change.

Why some AI-related visits remain unattributed

Referrers can be missing, including during some app transitions, redirects or privacy configurations. Later direct visits and device changes further complicate attribution. A report can identify sources without knowing the full preceding journey.

Users, sessions and events are not interchangeable. Keep your chosen denominator consistent. If necessary, consult the GA4 data-field documentation and the definitions of your specific report.

Mark changes to tracking, filters and goal events. A jump after introducing a new filter is initially a change in reporting. You may also offer a voluntary, open question about how someone first found you; that answer remains a separate source of evidence.

The zero-click guide distinguishes a search without an external click from a visit your analytics did not record.

Read AI answers and website outcomes side by side

Keep two small records: one for answers and citations, another for identified visits and goal events. This helps distinguish an information gap, a landing-page problem and a simple lack of data.

Nalqerio provides the view of released API answers to your customer questions. In this workflow it is not a replacement for GA4, nor an automatic connection between an AI answer and a Stripe purchase. Share of voice explains the competitive share within your answer sample.

Begin with one page and one clear next action. That tells you more than a large figure assembled from incompatible sources.