AI Overviews are eating your traffic. Here's the fight back.

In 2026, most Google searches end without a single click. When an AI Overview sits at the top, it's the overwhelming majority. Your organic traffic chart didn't break. The rules underneath it did.

You've seen the shape of it in your own analytics. Impressions holding up, or even climbing. Clicks sliding. Rankings that look fine on paper while the sessions they used to send quietly drain away. Nobody shipped a bad quarter of content. Google simply started answering the question on the page instead of sending the visitor to yours.

This piece is the honest version of what's happening and what still works. Not the panic version, and not the "SEO is dead" version, because it isn't. It's changed job, and the teams reading it correctly are pulling ahead of the ones still optimising for a results page that no longer exists.

Most searches no longer send a click anywhere

Here's the uncomfortable baseline. Across 2026, roughly two thirds of Google searches resolve without the user clicking through to any website at all. The answer appears in the results, the searcher reads it, and they're done. When an AI Overview is present, that zero-click rate climbs higher still, into the low-to-mid eighties.

The knock-on for organic traffic is real. Plenty of sites have watched search visits fall by twenty to forty per cent inside a year without dropping a single ranking position, purely because the click that used to follow the ranking stopped happening. Position one under an AI Overview is worth a fraction of what position one was worth two years ago.

If your reporting still treats "we rank for it" as "we get the traffic", the gap between those two things is now the size of your problem.

"Just rank number one" was always a proxy

The entire discipline was built on a simple chain: rank higher, get more clicks, get more pipeline. That chain held for twenty years. Google has now inserted itself into the middle of it, reading your page, summarising it, and keeping the reader on the results page. You did the work. Google kept the visit.

So the goal quietly moves. It's no longer only "be the top blue link". It's "be the source the answer is built from, and be worth the click that's left". Those are different briefs, and most content programmes are still writing for the old one.

Visibility without the click is still worth having

Here's the part teams miss while they're mourning the traffic. Being named inside an AI Overview keeps your brand in front of the buyer even when the click never happens. And when a brand is cited in the answer, the clicks it does still get tend to arrive warmer, from people who've already seen you framed as the credible source.

So the traffic gets smaller and higher intent at the same time. The tourist clicks vanish. The "already half-sold" clicks remain. That's not a disaster to be survived. For a lot of B2B teams it's a cleaner funnel than the one they had, provided the measurement catches up to it.

Stop treating the page as a destination to be clicked. Start treating it as a source to be quoted.

1. Build pages a machine can actually read

The Overviews are assembled by systems that reward pages they can parse cleanly. That means the answer stated near the top, a sane heading hierarchy, real structured data in JSON-LD, and semantic HTML rather than a soup of nested divs. Answer the question in the first two sentences, then earn the deeper read underneath.

This is the same discipline I laid out for the other side of the house in getting recommended by ChatGPT, Google AI and Perplexity. Whether the reader is a homeowner or a head of demand gen, the machine in the middle wants the same thing: clarity it can lift without guessing.

2. Own something the answer has to cite you for

Summarisers are ruthless with commodity content. If your page says what forty others say, the Overview blends you into a generic answer and credits no one. Give it something only you have: original data, a benchmark, a survey of your customers, a number nobody else can produce. Proprietary facts are the one thing an AI can't paraphrase away, because it has to point at the source.

3. Grow the channels the Overview can't intercept

Some traffic is now structurally out of reach of Google's answer box, and that's exactly where to invest. A real email list. Branded search, where people look for you by name. A community or audience that comes to you directly. None of these sit between a Google query and its answer, so none of them get skimmed off. The more of your demand that runs through channels you own, the less the AI Overview can quietly tax you.

4. Ship things an answer physically can't replace

A text Overview can summarise a text page. It can't be your interactive tool, your calculator, your configurator, or your live product experience. Those are reasons to click that a paragraph of AI summary simply cannot satisfy. It's why we build tools alongside content at Zenlio; a landing page teardown you actually run is worth more per visit than a listicle an Overview can eat in one bite. Interactive beats informational when informational is being commoditised by the sentence.

5. Be fast, because slow pages get read last

Crawlability and speed quietly underpin all of it. A page that loads instantly, on clean HTML, with a low time to first byte, is easier for these systems to fetch, parse and trust, and it converts the surviving clicks better too. I've put the revenue case for this in why your slowest page is costing you the deal. In an extraction economy, a bloated page is penalised twice: once by the reader who bounces, once by the machine that gives up parsing it.

Your CMS might be the thing holding you back

Here's the bit nobody in the demo wants to hear. The pages that AI systems find hardest to read are very often the ones a heavy CMS produces: div soup, render-blocking scripts, a tag manager nobody has audited in two years, three seconds of load before a word appears. The exact conditions that lose you the click also lose you the citation.

You don't need to rebuild the whole estate to fix this. You need your highest-intent pages built lean, clean and fast, on markup a machine can trust, and pointed at from the rest of the site. That's the work I do most weeks. If you want to size the drag your current platform is putting on you, the CMS cost calculator puts a number on it, and escaping CMS lock-in covers how a phased move actually works without a big-bang migration.

The click you used to win is gone. The question is whether you're the source it went to.

The teams that win the next two years aren't the ones with the most content. They're the ones whose pages are clean enough to be quoted, distinctive enough to be credited, and fast enough to earn the clicks that are left. That's a build problem as much as a content one. It's a good problem to have, because most of your competitors are still refreshing meta descriptions and wondering where the traffic went.