AI Traffic Quality: Do ChatGPT Visitors Actually Convert?

2. What Counts as AI Referral Traffic?

Quick Answer: AI referral traffic is any visit that arrives from an assistant such as ChatGPT, Gemini, Copilot or Perplexity, carrying that assistant as the referrer. Google Analytics now groups these under an AI Assistants channel. Google’s own AI Overviews and AI Mode are excluded. Those still count as organic search.

The definition matters, because two things people call “AI traffic” are measured in completely different places.

• Assistant referrals. The visitor was inside a chat window, clicked a link in the answer, and arrived with that assistant recorded as the source. This is the only kind you can count directly.
• AI Overview clicks. The visitor was on a Google results page and clicked a link inside the AI summary. Analytics records this as organic search, not as AI. See our look at the click-through data behind AI Overviews.
• AI-influenced visits. The buyer got a recommendation from an assistant, then typed your brand into Google or straight into the address bar. Nothing in the visit says AI was involved.

Google’s default channel group documentation now defines an AI Assistants channel for arrivals from sources like ChatGPT, Gemini, DeepSeek, Copilot and Grok, and states plainly that it excludes AI Overviews and AI Mode. That single line explains most of the confusion in AI traffic reporting. A dashboard that lumps them together is measuring two different things.

3. Does AI Referral Traffic Convert Better?

Quick Answer: On enquiry rate, yes. Across Malaysian SME sites we track, visitors arriving from an assistant enquire more often than non-branded organic visitors, though less often than people who typed the brand name. The channel sits closer to branded search than to cold search.

The external evidence points the same way. Search Engine Land reported a 12-month analysis of 94 e-commerce sites in which ChatGPT traffic converted at 1.81% against 1.39% for non-branded organic, about 31% higher, while contributing under 2% of organic revenue. Ahrefs, looking at its own site, found AI search traffic at roughly 0.5% of visits but 12.1% of sign-ups.

Our own numbers are lead-based rather than e-commerce, but the ranking holds.

Read the last column before the first. The AI channel wins on rate and loses badly on volume: 1.6% of sessions against 39% for cold organic. A channel converting at nearly five percent still produces fewer enquiries than one converting at two, when it is twenty-four times smaller.

4. Why AI Visitors Behave Differently on Arrival

Quick Answer: The comparison already happened. A searcher opens five tabs to work out who to call; an assistant user has been handed a shortlist and clicks to check one name on it. They arrive later in the decision, which raises enquiry rate and shortens the visit.

That single difference explains most of the behaviour. Three consequences follow, and each one changes what your page needs to do.

• They arrive to verify, not to browse. The visitor is checking that you are real, reachable and priced sensibly. Slow proof costs the enquiry.
• They land deeper than you expect. Assistants cite specific pages, so entry is often a service or guide page rather than the home page. Every page needs its own contact path.
• They ask fewer, sharper questions. The enquiry that follows tends to be specific — a scope, a date, a budget — because the general questions were answered before the click.

None of this is a reason to build a separate landing experience. It is a reason to make ordinary conversion rate optimisation work harder on the pages assistants actually cite. If you are unclear on the baseline, our plain explanation of what conversion rate means is the place to start.

5. Which Malaysian Industries Get the Most AI Traffic?

Quick Answer: Considered, research-heavy purchases send the most assistant referrals. Professional services, B2B suppliers and education sit at the top; walk-in and impulse categories sit at the bottom, where buyers use maps and social rather than an assistant.

The gap between top and bottom is nearly sevenfold, and it tracks how long the buyer spends deciding. A company sourcing an industrial supplier will describe the specification to an assistant; someone choosing dinner will not. Language matters too. Many Malaysian queries mix English with Bahasa Malaysia or Chinese, which is where multilingual SEO in Malaysia starts to matter for visibility.

6. How Fast Is the Volume Actually Growing?

Quick Answer: Quickly, from a very low base. Assistant referrals have moved from a rounding error to low single-digit percentages of sessions on Malaysian SME sites in two years. The enquiry rate has held steady while volume climbed, which is the healthier of the two patterns.

Watch the last column rather than the first. When a channel grows and its enquiry rate stays flat, the new visitors are as good as the early ones, and the channel is genuinely widening. A sagging rate would mean assistants had started sending browsers instead of buyers. That has not happened.

7. Where the AI-Influenced Enquiries Actually Land

Quick Answer: Most of them are filed as something else. When Malaysian buyers told us an assistant sent them, analytics had recorded only a minority as AI referrals. The rest arrived as direct, branded search or WhatsApp, because the buyer left the chat before clicking.

The WhatsApp column is the Malaysian part of this story. A buyer reads the answer on a phone, saves the number, and messages later, so the assistant gets none of the credit and the session may never exist. A swollen direct bucket hides more than this. See our guide to too much direct traffic in GA4.

8. How to Measure AI Referral Traffic Honestly

Quick Answer: Pair the analytics number with a self-reported one. Analytics tells you what clicked through; a “how did you hear about us” field tells you what influenced the decision. Neither is complete alone, and the gap between them is the useful figure.

How to set up AI referral traffic measurement

Five steps, in the order that gets a Malaysian SME to a trustworthy number fastest.

1. Separate the assistant channel. Confirm assistant referrals are reported apart from ordinary referrals, and apart from organic search where AI Overview clicks sit.
2. Add a self-reported source field. Put an open “how did you hear about us” question on every form and ask it on every phone enquiry. This is what catches the invisible majority.
3. Tag your WhatsApp entry points. Use a distinct link or number per page so a message can be traced back to the page the assistant cited.
4. Record the landing page, not just the channel. Which pages assistants cite matters more than the session total, because that list tells you what to strengthen.
5. Review monthly, judge quarterly. Volumes are small enough that a single week means nothing. Only draw conclusions once you have a few hundred sessions.

Step one is where most setups fail, and it is the mechanical part. Our walkthrough on tracking AI search traffic in GA4 covers the configuration. Steps two and three are the ones that actually change the picture, and almost nobody does them.

9. What to Fix Once You Can See the Traffic

Quick Answer: Work on the pages assistants already cite before chasing new citations. Strengthening a page that is being quoted lifts both the enquiry rate on today’s AI visitors and the odds of being quoted again tomorrow.

Four fixes, in the order that pays back fastest.

• Make cited pages convert. Put contact details, price context and proof on the page itself, not one click away. The visitor arrived to verify, not to navigate.
• Fix how you are described. An assistant that names you with the wrong service or no city sends the wrong people. Consistent identity work through entity SEO, a Google Knowledge Panel and a clean Wikidata entry is what settles it.
• Widen where you are readable. Assistants lean on public text. That is why AI keeps quoting forums, and why accurate business directory citations still earn their keep.
• Decide your crawler policy deliberately. Blocking the bots that feed these systems removes you from the one input you fully control. Think it through with our note on letting GPTBot read your website.

Coverage differs by engine, so check more than one. Google’s surface is covered in our guide to showing up in Gemini results, the Microsoft side in getting into Copilot answers, and the largest assistant in showing up in ChatGPT answers.

10. How to Judge an Agency’s AI Traffic Reporting

Quick Answer: Ask for the denominator. A report showing AI traffic up 400% without the session count, the enquiry rate and the self-reported figure beside it is describing a small number getting slightly less small.

Percentage growth on a tiny base is the easiest number in marketing to flatter. Four questions separate a real report from a good-looking one.

• How many sessions, in absolute numbers? Twelve to sixty is a 400% rise and still sixty sessions.
• What is the enquiry rate, and has it moved? Rising volume with a falling rate means worse visitors, not better visibility.
• What does the self-reported data say? If nobody asked buyers, the report only covers the visible third.
• Which pages earned it? Every gain should trace to a specific page an assistant started citing.

The same evidence standard applies to the visibility work upstream. Ask how AI search share of voice is measured against named rivals, and what a scope of AI SEO services actually commits to each month.

11. Conclusion

Quick Answer: ChatGPT visitors do convert, at rates closer to branded search than cold search. The channel is still small and badly under-recorded, so treat it as a quality signal to protect rather than a volume channel to replace search with.

The honest position sits between the two loud ones. The channel is not a fad; its enquiry rate has held while volume tripled. It is also not yet a channel that pays the bills for most Malaysian SMEs.

What it deserves is measurement. Add the self-reported question, separate the channel properly, and watch the pages being cited. That costs almost nothing and tells you when the volume finally justifies budget.

12. Frequently Asked Questions

1. Does AI referral traffic convert better than organic search?

Usually better than non-branded organic search, and usually not as well as branded search. The visitor has already compared options inside the chat, so they arrive later in the decision. Published analyses report gaps ranging from around 30% higher to several times higher, depending on the business model.

2. Why is my AI referral traffic so low in Google Analytics?

Because most AI-influenced visits never carry the assistant as a referrer. Buyers copy the name into Google, type the address, or message on WhatsApp instead of clicking. Those land in direct or branded search. The recorded number is real but partial.

3. Do clicks from Google AI Overviews count as AI traffic?

No. Google’s documentation puts AI Overviews and AI Mode inside organic search, not the AI Assistants channel. The AI Assistants channel covers arrivals from separate assistants such as ChatGPT, Gemini, Copilot and DeepSeek.

4. How much AI referral traffic should a Malaysian SME expect?

Low single digits as a share of sessions in 2026, weighted by how long buyers deliberate. Considered B2B and professional service categories sit at the top of that range, while walk-in retail and F&B sit well below it.

5. Is it worth optimising for AI search if the traffic is this small?

For most businesses, yes, because the same work also serves ordinary search. Clear service pages, accurate listings and consistent descriptions help you get cited and help you rank. The judgement call is how much extra spend AI-specific work deserves on top.