How to Track AI Chatbot Traffic in Google Analytics

Marketing teams across Singapore have started noticing the same puzzle in their Google Analytics reports: a small but growing cluster of sessions with sources like chatgpt.com or perplexity.ai, arriving with unusually strong engagement, and no clear paper trail back to a campaign. That traffic is not a glitch. It is the visible edge of a shift search marketers now call AI chatbot traffic, and most Singapore businesses are still measuring it by accident rather than by design.

Google Analytics 4 was never built with generative AI assistants in mind, so it often files this traffic under generic referral, or worse, lets it disappear into direct traffic entirely. That gap has real consequences, because the businesses that can show a finance team exactly how many enquiries came from ChatGPT or Gemini will be the ones defending an AI search strategy with numbers instead of a hunch. Here is a practical way to set that measurement up properly.

Where AI Chatbot Traffic Is Actually Coming From

When ChatGPT, Perplexity, Gemini or Copilot cite a website in an answer and the user clicks through, that visit arrives at the destination site carrying a referrer header identifying the assistant, much like a click from any other website. Analysts have started tracking these visits the same way they once tracked social referrals, by source domain and not by intent, since the assistant itself gives no further signal about why the user was sent there. Most Singapore marketing teams already have the raw data sitting in their property. What is usually missing is a deliberate way to pull it apart from everything else labelled as referral traffic.

The volume is still modest next to traditional organic search, but the citations behind it are becoming more common by the month. Recent Similarweb research covered by Search Engine Journal found that only around 6.8 per cent of ChatGPT answers included an external link by May 2026, up from roughly one per cent a year earlier, while AI platforms overall drew close to 9.5 billion average monthly visits. Search still commands a far larger audience, but a citation rate climbing this fast is exactly why the traffic deserves its own line in a report instead of a footnote.

Why GA4’s Default Channels Hide It

GA4’s default channel grouping was built around search engines, social platforms and paid media, so an AI assistant referral usually lands in the broad Referral bucket alongside partner sites and forums, or in Direct if the link strips its referrer information entirely. Either outcome buries a genuinely distinct audience inside traffic that behaves nothing like it, which makes it hard to answer a simple question: is the content actually getting cited, and is it converting when it is.

Building a Custom Channel Group for AI Referrals

The fix does not require a developer or a data layer rebuild. GA4 lets you define custom channel groups, and Google’s own guidance on custom channel groups in GA4 walks through the mechanics of adding new source and medium conditions to an existing report without disturbing anything else. Anyone with editor access to the property can set this up in a single sitting, and the existing default channel groups keep working exactly as they did before.

In practice, this means going into Admin, then Data display, then Channel groups, and creating a new group with a rule that matches session source contains any of the known assistant domains. A reasonably complete starting list looks like this:

• chatgpt.com
• perplexity.ai
• gemini.google.com
• copilot.microsoft.com
• claude.ai
• meta.ai

Once that rule is saved, GA4 begins classifying matching sessions under a clearly labelled AI Assistants channel going forward, though it will not retroactively reclassify historical data. This sits alongside Search Console’s new AI reports, which show impressions and clicks from Google’s own AI Overviews and AI Mode. Together, the two give a fuller picture of how AI is reshaping the top of the funnel, one from the search side and one from the analytics side.

Reading the Numbers Without Overreacting

Once the channel group is live, resist the urge to treat every session in it as a triumph or a threat. The absolute numbers will likely stay small for most Singapore businesses in the near term, since search still commands roughly five times the combined audience of every AI chatbot put together.

What tends to be more useful than raw volume is engagement quality. Visitors arriving from an AI assistant have usually already had a conversation about their problem before clicking through, so they often land further along in the buyer’s journey than a typical organic visitor lands. Their bounce rate can look different too, sometimes better, sometimes worse, depending on whether the page they land on actually answers the question the assistant implied it would. Comparing conversion rate for the AI Assistants channel against organic search, not just session counts, tends to tell a more honest story about whether this traffic is worth building content around. A handful of enquiries from a channel with a strong conversion rate can be worth more attention than a much larger channel that never turns into a lead.

Making AI Referral Tracking a Reporting Habit

A channel group only earns its keep if someone actually looks at it. Building a short monthly check into an existing SEO report keeps the data useful instead of forgotten:

1. Track the month-on-month trend in AI Assistants session volume.
2. Note which landing pages are receiving the referrals.
3. Compare conversion rate for that channel against organic search.
4. Watch for shifts in which assistant sends the most traffic.

Turning Measurement Into an Advantage

None of this requires new tooling or a bigger budget. It requires deciding, once, that this traffic deserves to be visible, and then actually looking at it each month instead of letting it sit unlabelled in a Referral bucket.

Setting this up takes an afternoon, not a redesign, and it gives a Singapore marketing team language a finance team will accept when the next conversation turns to AI search spend. If your own analytics setup could use a second pair of eyes, feel free to get in touch with our team.