Zero-click search: how to win when nobody clicks

The clicks are gone and they are not coming back. That's the premise. Everything useful starts after you accept it.

Pew Research tracked real browsing behavior and found that when an AI summary appears, 8% of users click a search result. Without one, 15% do. That's roughly half your click-through, gone, on any query that triggers a summary. The number that should genuinely alarm you is the other one: just 1% of users click a link inside the AI summary itself. Getting cited in the summary does not send you traffic. It essentially never does.

Ahrefs' 300,000-keyword Search Console study found the #1 organic result loses 58% of its CTR when an AI Overview sits above it. You can hold your rankings perfectly and lose more than half your traffic.

So the strategy is not "how do I get the click back." It's "what am I optimizing for instead."

What are you actually optimizing for now?

The mention. Not the click, not the session, not the ranking. Whether the model says your name when a buyer describes their problem.

This is a genuinely different goal and it changes what you build. Think about what happens after the model recommends you. The user doesn't click the citation (1% do). They read the answer, they now know your name, and later — hours or days later — they search your brand directly, or type your domain, or ask a colleague. Your analytics will record that as direct traffic or branded organic. The AI answer that caused it is invisible in every report you have.

Which means: your existing measurement stack is structurally incapable of seeing the thing that's now driving your demand. That's the real crisis, not the traffic loss.

How do you get named in the answer?

By being present in the sources the answer is built from, and by writing passages a model can lift cleanly. Two separate jobs.

Presence in the source set. Ask the engine your buyer's question and look at what it cites. It'll be a review site, two roundup posts, a forum thread, maybe a vendor comparison page. That handful of URLs is the answer's raw material. If you're not in them, no amount of on-site optimization puts you in the answer. This is the corroboration problem, and it's covered end to end in why your competitor shows up in ChatGPT and you don't.

Liftable passages. When a model does read your page, it extracts a span of text. Make the span exist:

• Answer the heading's question in the first sentence beneath it. Not the third paragraph.
• Put facts in plain text. Prices, limits, specs. Not in images, not injected by JS.
• Use tables for anything comparative. They get cited disproportionately, because the structure survives extraction.
• Write self-contained sentences. "Spottlo tracks four engines on every plan" survives being lifted out of context. "It includes all of them" does not.

What do you measure when the click doesn't happen?

Three layers, and you need all three because none of them is sufficient alone.

Layer 1: mention rate and share of voice

This is the direct measurement, and it's the one that didn't exist two years ago. Take the 25 questions your buyers actually ask. Run them through ChatGPT, Perplexity, Gemini, and Google AI Overviews. Record: were you mentioned, where in the list, who else was mentioned, what did the engine cite.

Do it weekly, because model outputs are non-deterministic and a single run is noise. The same prompt three times can produce three different vendor lists. What's stable is the rate across many runs, which is why this has to be a sampling exercise and not a spot check.

Share of voice — your mentions as a fraction of all vendor mentions across the prompt set — is the number that actually maps to competitive position. More detail in AI share of voice explained.

Layer 2: brand lift

The lagging indicator, and the one that will convince your CFO.

If AI answers are naming you more often, branded search volume goes up. Direct traffic goes up. "How did you hear about us" starts collecting weirder answers. None of these are attributable to a specific AI answer, and that's fine — you're looking for a correlation between your mention-rate line and your branded-demand line, over months.

Watch: branded query volume in Search Console, direct session volume, and the self-reported attribution field on your demo form. If mention rate rises for two quarters and branded demand doesn't, something is wrong with the quality of how you're being mentioned, which is worth knowing.

Layer 3: the AI referral segment

Small, and disproportionately valuable.

Some clicks do happen. When OpenAI added clickable brand links inside answers, Similarweb measured ChatGPT referral traffic rising 157.7% week-over-week. The channel is small but growing and the engines keep making it more clickable.

Segment it in GA4 by referrer (chatgpt.com, perplexity.ai, gemini.google.com, and the utm_source=chatgpt.com parameters OpenAI now appends) and then look at what those users do. See measuring AI search traffic in GA4 for the setup.

Why the traffic you lose is worth less than the traffic you keep

Here's the compensating dynamic, and it reframes the whole loss.

Semrush found an AI-search visitor is 4.4x as valuable as an organic search visitor, measured by conversion rate. Adobe Analytics found AI-referred traffic to US retail sites converts 54% better than non-AI traffic, while that traffic grew 138% year over year.

The mechanism is obvious once you see it. The old organic click was cheap and unqualified: someone typed a vague query, skimmed a SERP, clicked a blue link, and 70% of them bounced because your page wasn't what they wanted. The AI click is expensive and pre-qualified: the user described their problem in a full sentence, the model matched you to it, told them why you fit, and only then did they click. They arrive knowing what you do and believing you might be right. Of course they convert.

So the honest accounting isn't "we lost 58% of clicks." It's "we lost most of the low-intent clicks and kept a smaller stream of high-intent ones, and we need many fewer of them." Whether that trade nets out positive depends entirely on whether you're the brand getting named. If you are, the economics improve. If you aren't, you get the traffic loss with none of the compensation, and that's the failure mode most companies are currently walking into.

What changes in your content strategy?

Stop writing for the click. Some specific reversals:

Kill the curiosity gap. Headlines and intros engineered to make someone click to find out the answer are now actively harmful. The model reads the page, doesn't find the answer, and cites someone who stated it plainly.

Give away the answer. The top-of-page summary you were afraid to write because it would cannibalize the read — write it. It's the passage most likely to be lifted, and being lifted with your name attached is now the goal.

Write for extraction, not engagement. Time on page is not the metric anymore. A page that gets read by a model, extracted, and attributed has done its job whether or not a human ever loads it.

Stop measuring content by sessions. A post that drove 200 sessions and got you named in 6 of your 25 tracked prompts is outperforming a post that drove 2,000 sessions and got you named in none.

That last shift is organizationally hard, because someone's bonus is tied to the traffic number. Fix that first or none of the rest happens.

The thing nobody wants to say out loud

Zero-click means your website is no longer your primary distribution surface. It's the reference material behind an answer that someone else's model composes.

You don't own the interface anymore. You own the facts that the interface uses. That's a smaller, weirder job, and the companies that adapt to it fastest are going to take the consideration-set slots from the ones still running a 2019 content playbook and wondering why traffic is down 40% with rankings intact.

What to do next

• Pull your top 20 informational queries and check which ones trigger an AI Overview or a chat answer. That's the share of your funnel that's already zero-click. Model your traffic against the Pew and Ahrefs numbers, and take the forecast to your leadership before they notice on their own.
• Change the KPI on your content team's dashboard from sessions to mention rate. This is the hardest change in this post and the one that unblocks everything else.
• Segment AI referrals in GA4 today and look at their conversion rate against organic. If the 4.4x pattern holds for you, you have your business case for funding GEO work.
• Rewrite your three most important pages so the answer sits in the first sentence under each heading, with the facts in a table. Then check whether a model can correctly summarize the page from its text alone.
• Baseline where you actually stand across ChatGPT, Perplexity, Gemini and Google AI Overviews. Free report, no signup, at spottlo.com — or see how the scans work first.