Ask most marketing teams in Malaysia whether their business shows up in ChatGPT or Google’s AI Overviews, and the honest answer is usually “we’re not sure.” Google gave that question a partial answer in mid-2026 by adding a Generative AI performance report to Search Console. It is a genuine improvement. It also raises a new problem: the data it provides is easy to misread, and the traffic it implies rarely shows up cleanly in Google Analytics either. Measuring AI search visibility properly means understanding what each tool actually tells you, and building a process around the gaps.
What Search Console’s AI report actually shows
Search Console’s Generative AI performance report sits alongside the standard Performance report and tracks when a page appears in AI Overviews and AI Mode, with historical data running back to mid-May 2026. It can be filtered by page, country, device and date, in the same way as ordinary search performance data.
What it does not include is more significant than what it does. There is no click data, no click-through rate, no average position, and no query-level breakdown. A business can see that its pages were surfaced inside an AI-generated answer a certain number of times, but not what was searched, where the page ranked within the answer, or whether anyone acted on it. Google is, in effect, confirming something SEO practitioners had already suspected from indirect evidence: a citation inside an AI Overview is not the same event as a search result click, and the two need to be measured separately rather than assumed to move together.
Why impressions and clicks are decoupling
Once AI impressions are visible, the more useful exercise is comparing that trend against organic click trends for the same pages, rather than reading either number in isolation. It is common for AI impressions to rise while organic clicks for the same query set stay flat or fall, and it is tempting to read that as a ranking problem. Often it isn’t. If an AI Overview answers the question directly on the results page, a searcher who would previously have clicked through to compare two or three sources may no longer need to, regardless of where the underlying page ranks.
This matters for how a Malaysian business should react to a dip in organic traffic on informational content. The right first question is no longer just “did we lose rankings,” but “did the AI Overview take the click instead of a competitor.” Search Console’s new report is the first data source that lets that question be checked directly, rather than inferred from a plausible-sounding theory. It does not settle every case — a genuine ranking drop and an AI-answer substitution can happen at the same time — but it removes some of the guesswork.
Why GA4 usually hides the traffic that does arrive
Even when an AI tool does send a visitor through, GA4 frequently fails to label that session correctly. ChatGPT, Perplexity, Microsoft Copilot and similar tools often strip or alter referrer information before a click reaches the destination site, which means GA4 tends to file the session under Direct traffic, or under a generic referral bucket with no clear source, rather than attributing it to the AI platform that actually sent it. GA4 has no default channel grouping built for this category of traffic, so without deliberate setup, a business can be receiving a steady trickle of AI-referred visitors and never see it as a distinct line in its reporting.
Fixing this is a configuration task, not a platform limitation that has to be lived with. A custom channel group built around known AI referral domains — chatgpt.com, chat.openai.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and similar — will start separating this traffic out from generic referral and direct sessions going forward. It won’t recover historical data that was already miscategorised, and it won’t catch every session, since some AI tools pass no referrer at all. But it turns an invisible traffic source into a visible, trackable one, which is the minimum needed before anyone can make a sensible decision about whether AI search deserves more attention in the wider analytics and tracking setup.
A practical measurement framework
For a Malaysian business trying to get a genuine read on AI search visibility rather than a hopeful guess, three steps do most of the work.
- Treat the current AI impressions figure as a baseline, not a verdict. Because Google’s historical data only extends back to mid-May 2026, there is no long run of history to compare against yet. The useful comparison is forward from today: is the count of pages and impressions growing, shrinking, or shifting toward a different set of pages over the coming months.
- Fix the GA4 channel grouping before drawing conclusions from current AI referral numbers. Any business still using GA4’s default channel groups is almost certainly under-counting AI-referred sessions right now, which makes any claim that “AI isn’t sending us traffic” premature until the traffic that is arriving can actually be seen.
- Use manual or third-party AI-visibility checks as a supplementary layer, not a replacement. Periodically querying ChatGPT, Perplexity and Google’s AI Mode directly with the questions a prospective customer would plausibly ask is a reasonable way to sense-check whether a brand appears where it should. It is a useful signal, but it samples a narrow set of queries and can’t see into logged-in or personalised AI conversations, so it works best alongside platform-level data rather than instead of it.
Where this fits with what actually drives citations
MRVS looked previously at whether adding schema markup improves the odds of being cited by AI search tools, and the evidence pointed the other way: schema correlates with citation because it tends to appear on sites that already invest broadly in SEO, not because it causes citations on its own. The same caution applies here. A rising AI impressions count is worth watching, but it is a symptom of underlying content and search-visibility quality, not something to chase by making isolated technical changes and hoping the number moves. The businesses most likely to see durable AI citation are the ones already doing the fundamentals well — clear, direct answers to real questions, consistent entity information across the web, and organic rankings that were solid before AI Overviews existed at all.
For most Malaysian businesses, the practical priority isn’t choosing between Search Console’s new report, GA4 configuration and third-party AI-visibility tools. It’s making sure all three are actually reporting accurately before treating any single number as the answer to “are we visible in AI search.” MRVS works across SEO and the underlying analytics and tracking setup together, precisely because a visibility number that can’t be trusted is not much more useful than not measuring it at all.
