AI Search Visibility

How to track website enquiries from ChatGPT and other AI platforms

The most useful question about AI visibility is not simply “Did my business appear?” It is “Did that discovery contribute to a worthwhile enquiry?”

You will not be able to identify every journey perfectly. Some visitors arrive through a link, some search for your business later and some tell you how they found you. A sensible measurement setup combines the evidence available without pretending it is complete.

By David Gee ·

Separate the things you are measuring

What mentions, citations, website visits and enquiries tell you; definitions, not reported results
SignalWhat it tells youWhat it does not prove
Brand mentionYour business appeared in an observed answerThat someone visited or contacted you
CitationA page was linked or referencedThat the citation produced a click
Website visitSomeone reached the siteThat they were a suitable prospect
EnquirySomeone completed a meaningful contact actionThat they became a customer

Keep those categories separate in your reporting. Calling every citation a lead makes the numbers look better while making business decisions harder.

Review identifiable referral traffic

Ask whoever manages your analytics to inspect the source and landing page of relevant sessions. Some AI referrals may be identifiable from the referrer or campaign information available; others will not be.

Use actual observed data to build your reporting groups. Keep a record of which sources you include, and review the list as platforms and referral behaviour change. Do not automatically label all direct traffic as AI traffic.

For Google, its AI Search documentation describes AI-feature traffic within overall Search Console Web reporting. That does not justify treating every Google organic visit as an AI-generated referral.

Track successful enquiries, not just button clicks

A click on “Get a quote” shows interest. A successfully submitted form shows that an enquiry was accepted. Those should be different events.

Check that form tracking happens after confirmed success, fires once and does not also fire when the submission fails. Keep phone or email link clicks separate because a click does not prove a conversation took place.

The useful follow-up is lead quality. Was the enquiry about a service you offer? Was the person in an area you serve? Did it lead to a quotation or booking? A simple lead log can be more useful than a crowded dashboard.

Ask customers how they found you

An optional “How did you hear about us?” question can fill some gaps. Offer a few straightforward choices and an “Other” option. Keep it easy to skip.

If a customer says ChatGPT, record that as a customer-reported source. It may differ from the last source shown in analytics because they visited several times or searched your name after seeing a recommendation.

Keep both pieces of information where useful. Neither has to erase the other.

Observe visibility consistently

If you check AI answers manually, use a small set of questions linked to your services. Record the exact wording, date, platform and whether search was used. Save the cited page and note obvious inaccuracies.

Avoid changing every prompt between reviews. Equally, do not interpret a handful of questions as a complete view of the market. These are observations that may help you investigate, not a universal ranking score.

Microsoft’s AI Performance guidance describes citation reporting for supported experiences including Copilot. Where available, it can add context, but citations remain different from visits and leads.

Keep measurement proportionate

Use your existing consent and analytics setup properly, and have a competent implementer review any changes. Do not send names, email addresses or the contents of enquiries into analytics events.

You usually need to know which page and contact action contributed, not the personal details of the person asking. Your enquiry system can handle the information needed to respond.

What should a useful monthly review contain?

Keep it focused: identifiable AI visits, the pages they reached, successful enquiries, lead quality where known and the improvements made during the period.

Compare results over a meaningful period and note other changes, such as advertising, a new offer or seasonal demand. With small numbers, percentages can be misleading. One enquiry becoming two is not enough evidence to claim a repeatable growth rate.

Use what you learn to improve the next page or the next step in the customer journey. Our AI search action plan explains how to prioritise that work, and the AI Search Visibility Hub brings the platform guides together.

If you want a website that connects useful content with a clearer enquiry journey, contact Gee Web Design. Tell us what you currently track and where you feel you are missing information, and we can discuss the appropriate scope of work.

For the wider website work involved, see our local SEO services.