AI referral traffic, no enquiries: inspect the path

The AI sessions chart is rising. The enquiry inbox is not. A marketing team adds ChatGPT, Perplexity or Copilot referrals to a dashboard, notes the upward line, and assumes the next task is to acquire more of the same traffic.

That is often premature. For a Dubai, UAE or UK service business, high AI referral traffic with few enquiries can mean the visitor found the wrong page, reached a page with no clear commercial route, abandoned a difficult contact process, or completed an action that analytics failed to record. The referral is an observation. It is not proof that the landing page did its job.

The chart records arrival, not the ability to act

AI referral traffic is useful because it shows that a platform sent a measurable visit to your website. It cannot, by itself, show why the visitor came, whether the page met their need, whether they found a relevant service, or whether an enquiry event was captured correctly.

The tempting assumption is simple: a referral session proves the landing page produced, or should have produced, an enquiry. In practice, that conclusion skips the part where the visitor has to understand what you offer and choose a route to contact you.

A common example is a visitor landing on a portfolio article after asking an assistant for examples in a particular sector. The project looks credible, but the page offers only related projects and a general footer link. It does not identify the relevant service, explain the next step, or make an enquiry route visible before the visitor runs out of patience. The source may have sent a sensible visitor. The destination has not made a sensible response easy.

Diagnose the page before judging the referral source

Start with the actual landing URLs, rather than the total AI referral number. Grouping every AI platform into one acquisition channel is useful for a high-level report, but it can hide a very different problem on each destination.

Compare the landing intent with the page’s job

Read the landing page as the visitor would. What question might have led an assistant to send someone there, and does the page answer that question without requiring a second hunt around the site?

A portfolio page may satisfy evidence-seeking intent. It may be a poor page for someone who now needs to know whether your agency provides technical SEO, AI visibility monitoring or structured-content work. Add a relevant path when it is genuinely useful: a short contextual service link, a clear next action, or a section explaining the work behind the example. Do not turn every case study into a collection of desperate buttons.

Also inspect basic machine-readable relationships while you are there. A project page that names a vague service in body copy but has no clear internal link to the corresponding service URL makes the route less obvious for people and weaker for crawlers. That is a mundane implementation detail, but it regularly matters.

Trace every contact route on the page

List the actions a visitor can take from the landing page: form submission, telephone link, email link, calendar booking, WhatsApp route, downloadable brief, or a click through to a relevant service page. Then test each one on desktop and mobile.

This is where apparently healthy traffic often meets a boring fault. A form may submit inside an iframe while GA4 listens only for a button click on the parent page. A thank-you page may not load. A telephone link may work perfectly but never count as an event. Or the main contact link may sit in a navigation menu that disappears behind a mobile menu control.

Use engagement as a clue, not a verdict

Engaged sessions can help separate an immediate mismatch from a page that held attention but did not produce a measurable action. Look at engagement in context with the landing page and device, rather than treating it as a quality score.

• Very short visits with no meaningful scroll or onward clicks can suggest an intent mismatch, a slow page, an unexpected destination or a weak opening.
• Engaged visits that repeatedly move from a portfolio page to a relevant service page may show early research behaviour rather than a failed conversion.
• Engaged visits with contact-link clicks but no recorded enquiry point first to measurement or form completion issues.
• Engaged visits with no visible contact interaction can indicate genuine friction, but only after you have checked whether the page presents a credible route to act.

Do not infer a person’s motives from a single anonymous session. Use patterns across comparable landing pages, and retain the source, device and route data that makes a pattern testable.

Verify the conversion event, not just the event name

A GA4 conversion called enquiry is only as reliable as the condition that fires it. Confirm that it records a completed form submission or other defined action, rather than a button click, page view or form-start event. Then complete a test enquiry and check the path through the tag manager, analytics property and CRM or inbox.

The measurement setup deserves its own careful review. Our guide to tracking ChatGPT traffic in GA4 without confusing referrals with visibility covers the acquisition side of the problem. This diagnosis starts where that report stops: whether the visitor could take, and complete, the action the business needs.

High AI traffic with few enquiries should be assessed by comparing landing intent, available contact routes, engaged sessions and verified conversion events. If those inputs align, the lack of enquiries may indicate genuine commercial friction or low-intent research traffic. If they do not align, the first problem is usually measurement or page routing, not the AI referral source.

Separate a tracking gap from actual friction

You do not need to rebuild the site to make this distinction. Use a small evidence record for the affected landing pages and write down what each check proves.

Signs the measurement may be incomplete

• A test form reaches the inbox or CRM but no conversion event appears in GA4.
• Email, phone, calendar or WhatsApp actions are available but are not tracked.
• A third-party form, cross-domain booking tool or consent setting interrupts the expected event.
• Contact clicks occur, but the destination URL or completion state is not captured.

Signs the page may be creating friction

• The referral lands on an article with no prominent route to the service implied by the topic.
• The page gives evidence of past work but no explanation of scope, fit or next action.
• The primary contact route is hidden on mobile, slow to load, or asks for more information than the initial enquiry reasonably requires.
• The visitor is sent to a generic contact page that loses the context of the service or case study they were viewing.

One useful repair is to preserve context. If a visitor moves from a hospitality portfolio piece to a service enquiry, pass the project or service name into the form where practical. The sales or delivery team can then see what prompted the contact, and the website team can verify whether the route worked. This is more useful than adding another broad AI metric to a monthly slide.

A diagnostic checklist for high AI traffic and few enquiries

1. Export the AI-referred landing pages and review the page, device and platform separately.
2. Write the likely visitor question beside each landing page, without pretending it is known fact.
3. Check whether the page answers that question and links clearly to the relevant service or contact route.
4. Test every contact path on mobile and desktop, including forms, mail links, booking tools and telephone links.
5. Submit a controlled test and verify the event, thank-you state and business-side receipt.
6. Compare engaged sessions, onward service-page visits and contact interactions before deciding whether the issue is traffic quality or page friction.
7. Make the smallest defensible change, then review the same landing-page paths again rather than celebrating a general traffic rise.

When the evidence warrants a wider review

Escalate beyond a page-level fix when several important AI-referred destinations have unclear service routes, conversion tracking cannot be reconciled with actual leads, or the same intent repeatedly lands on informational content with no commercial continuation. At that point, the issue may involve site architecture, measurement design and content roles together.

Do not claim that an AI platform caused the commercial shortfall when only referral sessions are known. Separate what was observed from what is plausible and what remains unmeasured. If the pattern survives verified tracking and a clear enquiry path, it is worth prioritising in a wider AI visibility audit implementation plan. Until then, the rising chart is merely a reason to inspect the route.