For twenty years, the goal of SEO was simple to state: get to the top of page one. In 2026 that goal still matters, but it no longer guarantees what it used to. Google's AI Overviews now sit above the traditional results on a large share of searches, answering the question before the user ever reaches a link.
If your traffic has softened while your rankings held steady, this is almost certainly why. Here is what is actually happening, and what to do about it.
The numbers behind the drop
Estimates vary depending on who is measuring and which query set they use, but the direction is consistent across every study:
- AI Overviews now trigger on a very large share of searches. Older studies put it in the single digits; more recent measurements put it closer to half of all queries.
- When an AI Overview appears, organic click-through on the top result falls by roughly a third on average. Ahrefs has measured drops of up to 58% for position one specifically.
- Zero-click searches, where the user gets their answer and never leaves the results page, now account for well over half of searches in the US.
Read those together and the picture is clear. The same ranking now produces meaningfully fewer visits than it did two years ago. Nothing about your site got worse. The page around your listing changed.
Ranking is not the same as being cited
The important shift is what "visibility" now means. Traditional SEO optimises for a position in a list of links. AI Overviews optimise for something different: which sources the model pulls from and names when it composes its answer.
That has produced a new discipline, usually called Generative Engine Optimization (GEO), sometimes Large Language Model Optimization. The success metric is not your average position. It is how often your brand gets cited across the range of prompts your customers actually ask.
It is worth being blunt about the state of this field: GEO is young, the tooling is immature, and anyone selling you a guaranteed formula is overselling. What we do know is reasonably well established, and it looks a lot like good publishing practice.
What actually earns citations
Across the studies published so far, the pages that get pulled into AI answers tend to share a few traits.
They answer the question directly, early
A page that spends four paragraphs warming up before reaching the point gives a model very little to extract. Lead with the answer, then explain it. This is also better for human readers, which is the usual sign that a tactic will age well.
They are structured
Clear headings, short paragraphs, lists where a list is genuinely the right shape, and tables for comparisons. Structure helps a model identify a self-contained, quotable passage.
They carry real evidence
Specific numbers, named sources, dates, and first-hand data get quoted more than general advice. Generic content is exactly what a model can already generate without you.
They are machine-readable
Valid structured data helps a system understand what your page is, who published it, and what entity it concerns. This is unglamorous work, and it is the part most sites skip.
The part nobody mentions: the traffic is better
There is a genuine upside buried in the gloom. AI search sends less traffic, but the traffic it sends converts better.
The reason is intuitive. Someone who reads a summary, decides they need more depth, and deliberately clicks through to a named source is further along in their thinking than someone who clicked the first blue link to see what it was. You are losing the casual browsers. You are keeping more of the buyers.
That has a practical consequence for how you report. If your dashboard tracks sessions as the headline number, the last two years look like a decline. If it tracks qualified leads per thousand sessions, the same period may look flat or better. Measure the thing you actually sell.
What to do this quarter
- Find out where you stand. Take the twenty queries that matter most to your business and check them manually. Does an AI Overview appear? Are you cited? This takes an afternoon and tells you more than any tool.
- Fix your highest-value pages first. Put the direct answer in the first hundred words. Add structure. Add specifics.
- Get your structured data right. Organisation, service and article markup, valid and consistent with your visible content.
- Change what you report. Add impressions and citation presence alongside sessions, so a fall in clicks does not get read as a fall in visibility.
- Do not abandon classic SEO. AI Overviews are largely assembled from pages that already rank. Traditional relevance and authority are the entry ticket, not an alternative strategy.
The uncomfortable summary
You are unlikely to get the old click volumes back. What is available instead is a smaller, better-qualified stream of visitors, and a citation slot that most of your competitors are not yet trying to win.
The businesses that adapt early to this will hold that ground for years, exactly as the businesses that took SEO seriously in 2010 did. The ones that keep optimising purely for a rank they can no longer convert will spend the next two years wondering where the traffic went.
If you want a clear read on how your site is currently performing in AI search, our SEO optimisation team can audit your key queries and show you exactly where you are being cited and where you are being skipped. Get in touch and we will start with the twenty queries that matter most to you.
