Track Which ChatGPT and AI Visits Become Real Inquiries

Write a qualification definition in language the person answering the inquiry can use. A commercial contractor might look for an offered service, an appropriate property type, a genuinely served location, and a request that can move to an assessment. A retail appointment business may instead need the service choice and a workable booking request.

Keep inquiries separate from completed sales. A good conversation can end without a proposal, and a proposal can stay open. Record each stage on its own. We’d rather show a smaller, honest report than count every form as revenue.

Allow an unresolved status. A new form submission may not contain enough information to qualify immediately. Labeling it pending is more accurate than treating it as unsuitable or successful before the team has responded.

Test the contact path from the page a referred visitor is likely to reach. Confirm that the form can be completed, the intended recipient receives the request, and the customer sees an accurate confirmation. A button click and a received inquiry are different events, so we send a labeled test request and check that it reaches the right inbox.

Google Analytics describes key events as important measured actions. Choose an event that corresponds to the action the business cares about, then verify its behavior. A misconfigured event can count repeated clicks or page views while the team assumes it measures completed requests.

Document test submissions so they can be excluded from operational review. Use a clearly labeled test record, check the handoff, and confirm that the record does not later become a reported customer opportunity.

A source associated with a first visit answers a different question from the source associated with the current session or credited to an important event. Google’s documentation on traffic source scopes explains these distinctions. Name the field you are using instead of calling every source value “where the lead came from.”

For a hypothetical journey, a person might first discover an article through search, return through an assistant link, and later submit a form after typing the website address. Each observation can be relevant. The available records may not connect all three, especially across devices or measurement gaps.

Keep an optional self reported discovery answer separate from observed analytics data. If a customer says an assistant helped them research, record that as their account of the journey. Don’t overwrite the measured source just to make the two fields agree.

Pick fields your team will actually fill in. A short record that someone checks beats a long one that sits empty. Keep contact details in the appropriate business system and avoid sending unnecessary personal information into general analytics fields.

A person may submit a form, call for clarification, and reply to an email. That’s one request with three touches, and we count it once. Link them to the same inquiry when the business can confidently identify that relationship.

Keep genuinely different projects distinguishable. One customer may request two unrelated services at different times. The record should preserve both the customer relationship and the separate opportunities. Write the rule before reviewing channel performance so the counting does not change opportunistically.

The owned lead measurement playbook explains how to establish that operational record across a main website, campaigns, or a network of service sites.

Show available attributed visits, received inquiries, qualified inquiries, and later outcomes as separate measures. Use the same date convention and explain whether outcomes are grouped by inquiry creation date or completion date. A long sales process can otherwise make a recent period look artificially weak. We apply the same cutoff to our own numbers. Our commercial roofing count covers four inquiries from 209 Indiana sites, June 29 through August 31, 2026. A TPO inquiry on September 2 came after that window, so it isn’t in the count.

A hypothetical report might show eight attributed visits and two received requests, with one still awaiting clarification. Those numbers describe the recorded process. They do not establish the total audience exposed to an AI answer or the causal contribution of the assistant to the eventual sale.

Use the small dataset guide before interpreting early conversion percentages. Review citation observations alongside the funnel when useful, while preserving the distinction between an appearance and an arrival.

If suitable visitors repeatedly ask whether a service is available in their area, clarify that information on the destination page. If inquiries lack a detail needed to respond, improve the form explanation or add a relevant optional field. If qualified requests sit unanswered, fix the response assignment before investing in more exposure.

These changes can improve the customer experience even when AI traffic is still small. The service inquiry design playbook provides a practical companion for form and response work.

Can every direct visit be assigned to AI research?

No. Preserve the source information the measurement system actually provides. Unknown or direct activity should not be relabeled to support a preferred explanation.

Does a form submission prove a qualified lead?

It proves the configured action or received request under the system’s definition. Qualification requires the business’s review of service fit and the information needed for a useful next step.

What is the best first improvement?

Verify that one important contact path works and that the resulting record receives a status and owner. That gives the business a reliable foundation before adding more complex attribution.

Discuss your inquiry measurement process with Trojan Digital Marketing. Bring the relevant page, the contact workflow, and the decision your current reporting cannot answer.