AI in Google Analytics 4: what the Analytics Advisor does and does not do for you

AI in Google Analytics 4 is not an autopilot, it is an assistant. In short: GA4 has two kinds of AI on board. One is diagnostic (it tells you what is happening now and what stands out), the other is predictive (it tries to guess what a visitor will do next). The diagnostic side, such as the Analytics Advisor, is immediately usable. The predictive side requires so much data that it simply does not work for most Belgian B2B sites. In this article we explain honestly what you can and cannot expect from it.

The common thread: AI helps you see faster where to look, but it does not decide for you. Steer on customers and revenue, not on a dashboard that suddenly shows a pretty number.

Which kinds of AI are inside Google Analytics 4?

It helps to keep two categories apart, because they do something completely different.

The first is diagnostic AI: features that search through your existing data and point out what stands out. Think of the Analytics Advisor and the automatically generated insights. They look backwards and at the present: they compare periods and report deviations.

The second is predictive AI (predictive metrics): it tries to look ahead and assign a probability to future behaviour, such as the chance that someone will buy within seven days or drop off instead. Google introduced these predictive capabilities in Google Analytics as an addition to the standard reports.

The difference is crucial. Diagnostic AI already works on your account today, no matter how big you are. Predictive AI needs a minimum amount of training data before it kicks in at all. If you first want to understand how GA4’s entire measurement model fits together, read our piece on GA4. And if you are wondering whether you should manage this yourself or better outsource Google Analytics, we come back to that at the bottom.

What does the Analytics Advisor actually do?

The Analytics Advisor (and the broader “Insights” feature) runs over your data daily and picks up patterns you would otherwise miss. These are ready-made observations in plain language.

A typical example of what such an insight delivers: “Pageviews from organic search +47% versus last week” (see how Google Analytics 4 uses AI to enhance your marketing data). So you do not just get a number, but a comparison with context: what went up or down, on which channel, and by how much.

That is useful for three reasons:

• You spot deviations faster than when you leaf through reports manually every morning.
• It focuses your attention: a 47% jump in organic traffic is a reason to check whether those visitors also convert, or whether it is noise.
• You can follow up on insights and ask the data questions yourself without building a report.

But watch out for the pitfall. A rise in pageviews is in itself a vanity metric: nice to look at, but only valuable once those visitors do something that counts. The Advisor shows you the way, you judge whether it holds up. So always tie an insight to your real goals, such as a completed contact form or a quote request, and to your conversion rate.

Do the predictive metrics in GA4 work for my company?

This is where the honest story matters, because this is where many B2B companies hit a wall.

Google Analytics 4 offers predictive statistics such as purchase probability and churn probability. To calculate them, Google needs training data: at least 1,000 returning users who did show the relevant behaviour (a purchase, for example) and at least 1,000 who did not, within a 28-day window (according to GA4’s threshold requirements for predictive metrics).

Do the maths. For a Belgian B2B service provider with a few hundred quality leads per month, 1,000 conversions in 28 days is often unreachable. And on top of that you also need 1,000 negative examples. The result: the predictive metrics stay empty or never become available. That is not a mistake in your setup, it is simply how the model works.

Our stance on this: do not pretend. We sometimes see proposals to set up “AI-driven audiences” based on purchase probability, while the account never reaches that threshold. That is selling smoke. Predictive metrics are built for sites with many, frequent and similar conversions (think e-commerce with thousands of transactions), not for a B2B funnel with long sales cycles and low volumes. You build usable audiences on purchase intent rather than on volume: read how to segment remarketing with Google Analytics smartly.

So what does work for you? Solid first-party data and clear events. If you measure properly which channels deliver leads and what a lead is worth, you do not need a probability score from Google to make the right decision. Diagnostic AI plus your own conversion tracking gets you further than a prediction that never kicks in.

How do you use AI in GA4 without fooling yourself?

AI in GA4 is at its best as an accelerator, not as a decision maker. A few practical principles:

• Use the Analytics Advisor to look faster, not to think less. An insight is a question, not an answer. “Organic +47%” means: go investigate, not: pop the champagne.
• Ignore predictive metrics if you do not have the volume. Do not waste time on features that will stay empty anyway. Invest that time in clean event tracking.
• Translate every AI signal into a customer question. Does this channel deliver real leads? Does this segment eventually become a customer? AI that does not lead to an action is noise.
• Keep a human in the loop. AI sees patterns, but it does not know your context. A spike can be a successful campaign or a bot attack. You make that distinction. Such a polluting spike is often referral spam in Google Analytics, which you had better learn to recognise and filter.

This fits our whole view on data: it is only valuable if it leads to a better decision. Do you want to set this up and monitor it professionally? Then look at our data & analytics service, where we configure GA4 around your leads and revenue instead of around features that happen to be available.

Frequently asked questions about AI in Google Analytics 4

Is the AI in GA4 free? Yes, the diagnostic functions such as the Analytics Advisor and automated insights are part of the standard version of GA4 and cost nothing extra. Predictive metrics are free too, but they are only calculated if your account meets the data thresholds.

Why do I not see any predictive metrics in my account? Usually because you do not have enough conversions. GA4 needs at least 1,000 returning users who did show the behaviour and 1,000 who did not within 28 days to train a prediction. For many B2B sites with lower volumes that threshold is never reached.

Can I blindly trust the insights the AI shows? No. Treat an insight as a pointer to where you should look, not as a conclusion. The AI does not know the context of your business and can flag an innocent fluctuation as important.

Does AI in GA4 replace a data analyst? No. AI speeds up the spotting of patterns, but the interpretation, choosing the right KPIs and linking them to revenue remains human work. The AI is an assistant, not a replacement.

Want to set up GA4 around customers instead of features?

AI in Google Analytics 4 can take a lot of work off your hands, as long as you know which features fit your volume and goals. We set up GA4 so that you measure what counts: leads, channels and revenue, not vanity figures or predictions that stay empty anyway. Small team, moves fast, honest advice.

Plan your free intake