How to Measure SEO Performance in the AI Age: What Your Metrics Are Telling You Now

Search has changed. The way people find information, discover brands, and make decisions looks very different to how it did even 18 months ago. And yet most businesses are still measuring performance the same way they always have, opening GA4, looking at sessions and organic traffic, and drawing conclusions from numbers that no longer tell the full story.

The metrics have not broken. The environment around them has. And that means we all need to rethink what good performance actually looks like, and what we should be measuring to see it.

This is what those numbers actually mean right now, what to track instead, and what to expect next.

What you are seeing in GA4 right now

If your organic traffic is falling, your direct is up, and a small channel called AI Assistant has started appearing in your acquisition report… you are not imagining it, and nothing is necessarily broken.

Lily covers the full picture of why organic click volumes are declining in our piece on the rise of zero click searches. The short version is that AI Overviews and AI chat tools are answering more queries before anyone clicks, and that behaviour is structural, not a blip. Ahrefs analysed 300,000 keywords and found that position one CTR drops by 34.5% when an AI Overview is present. Seer Interactive found organic CTR on those queries fell from 1.76% to 0.61%. The click, increasingly, is optional.

What is worth understanding here is what that shift is doing to the other numbers in GA4.

Direct traffic is rising in part because AI-driven discovery does not always leave a trail. When someone reads your brand name in a ChatGPT response, closes the tab, and navigates to your site directly… GA4 records it as direct. The visit is real. The source is invisible. The Digital Bloom estimates that around 70.6% of AI-driven traffic is currently misclassified this way, sitting unattributed in the direct bucket. If your direct traffic has grown materially in the last 12 months, a meaningful share of that is almost certainly AI-influenced discovery you have no visibility into yet.

The AI Assistant channel is where GA4 captures the traffic that does pass its referrer correctly: clicks coming directly from ChatGPT, Gemini, Perplexity, Copilot and similar tools. Volume is still small, less than 1% of total web traffic across all AI platforms. But one study found AI search visitors converting at 14.2% versus Google organic at 2.8%. These visitors arrive having already read a summary of what you do. They are warm in a way that most organic traffic simply is not.

Ranking on page one no longer means what it used to. The click is increasingly the exception, not the default.

The metrics that actually matter now

Sessions and clicks are not gone, but they are no longer the main story. Here is what to look at alongside them. If you are not sure whether your GA4 is set up to track any of this properly, our common GA4 problems guide is a good place to start.

Engaged sessions by channel

An engaged session lasted more than 10 seconds, triggered a conversion event, or included at least two page views. It filters out noise automatically. Engagement rate by channel tells you far more about traffic quality than raw volume does, and it holds its value even as total session counts shift.

Conversion rate and revenue per user by source

Traffic is an input. What it does is the output. Breaking conversion rate and revenue per user down by channel quickly shows which sources are actually contributing to the business, and which are inflating the session count without delivering anything meaningful.

Brand search volume alongside GA4

Growing brand search in Google Search Console while organic clicks decline is one of the clearest signals that AI tools are surfacing your business upstream. Users are discovering you through an AI answer and searching your name to find you. It shows up as branded organic or direct in GA4, but the intent behind it is AI-driven. Tracking both together gives you the fuller picture.

Direct traffic to interior pages

Real navigational direct traffic lands on your homepage. Direct sessions landing on blog posts and service pages is almost always dark traffic arriving from somewhere GA4 cannot see. Filtering by landing page separates the two and gives you a cleaner read on what is actually happening.

Predictive metrics: purchase probability and churn probability

GA4 now runs machine learning on your user data and generates three forward-looking metrics automatically once your property hits the data thresholds: purchase probability (likelihood to convert in the next 28 days), churn probability (likelihood not to return within 7 days), and predicted revenue.

The practical use is building remarketing audiences around users with high purchase probability who have not yet converted. The platform has already identified who is most likely to buy. The question is whether you are acting on it.

What to expect next

The shifts happening now are structural. They are not going to reverse. But the measurement picture will get clearer as GA4 and the broader analytics ecosystem catch up.

AI referral attribution will improve

Right now, most AI-driven traffic lands as direct because AI interfaces do not consistently pass referrer data. That is changing. GA4’s AI Assistant channel is already a step forward, and as AI platforms mature, more of that traffic will be correctly labelled. The businesses tracking it now will have a baseline to measure against when that happens.

Predictive signals will become standard

As traditional traffic metrics become less reliable indicators of business health, predictive metrics like purchase probability and churn risk will move from a nice-to-have to a core part of reporting. GA4 is already generating this data. Most businesses have not looked at it yet.

Brand visibility will matter more than position

If AI Overviews and AI Mode continue expanding, whether you rank in position one will matter less than whether you are cited, referenced, and trusted as a source. GA4 cannot measure citation share directly, but it can show you the downstream effects: branded search lift, direct traffic to content pages, and stronger conversion rates from AI-referred sessions. This is also where AI SEO becomes relevant, optimising not just for traditional rankings but for how AI tools discover and reference your content. Tracking those signals now gives you a head start on understanding what SEO performance will look like in 12 months.

So, what now?

Most businesses already have the data to understand what is happening to their traffic. They just have not rebuilt their reporting around what the numbers mean in 2026.

If your organic is down, that does not automatically mean your SEO is failing. If your direct is up, that does not automatically mean brand awareness is growing. And if a small channel called AI Assistant has started ticking upward… that is worth paying close attention to.

If you want someone to look at your GA4, make sense of what the numbers are telling you, and help you build a measurement framework that actually reflects how your customers are finding you today, that is exactly what we do.