Because a top ranking no longer guarantees a visit. Google now answers many questions inside the results page itself. Your page can hold position two and still lose clicks to an AI summary sitting above it. The ranking did not fail. The click did.
We see this pattern in almost every audit we run now. A founder shows us a Google Search Console chart where impressions climb month after month while clicks flatten or fall. Nothing broke. The search result page changed shape underneath them.
This is zero-click search, and it is now the normal state of Google rather than an edge case. Once you accept that, the job changes. You stop chasing raw traffic and start competing for the answer itself.
Zero-click search is any search where the person gets what they need without visiting a website. They read an AI summary, a featured snippet, a map pack, or a knowledge panel and close the tab. The query was satisfied. No site got a visit.
SparkToro published new research in 2026 using Similarweb clickstream data covering January through April 2026. It found that 68.01% of United States Google searches ended without a click. In 2024 that figure was 60.45%, and about a decade earlier it sat near 45%.
Read that trend line carefully. Zero-click behavior did not appear with AI. It has been growing for ten years as Google added weather boxes, calculators, sports scores, and definitions. AI Overviews accelerated something that was already happening.
Two forces stacked on top of each other. Google kept adding features that answer questions in place, and generative AI made those answers cover far more query types than a weather box ever could. The same SparkToro research recorded a 9.51 percentage point drop in clicks between 2024 and 2026.
That drop is not spread evenly. Simple factual questions lost the most, because an AI summary handles them perfectly. If your page ranked well for "what is a 301 redirect" or "webflow pricing", you probably felt it first.
Commercial and comparison queries held up better. When someone is about to spend money or make a decision they cannot undo, they still want to see the source. That is the part of search worth defending.
They roughly halve it. Pew Research Center studied the browsing behavior of 900 United States adults across 68,879 unique Google searches in March 2025. When an AI summary appeared, people clicked a traditional search result in 8% of visits. When no summary appeared, they clicked 15% of the time.
Those two numbers explain most of the panic we hear from clients. An AI Overview does not remove your listing. It moves your listing below a block of text that already answered the question, and it cuts your odds of a click by nearly half.
The strategic response is not to fight the summary. It is to be inside it. Google cites sources in AI Overviews, and a citation carries your brand name to a reader who is actively researching. We wrote about how to earn those placements in our guide to optimizing for Google AI Overviews, which covers the structure these systems reward.
Content that cannot be summarized without loss. Original data, real numbers from real projects, tools, calculators, comparison tables people need to scan, and specific opinions from a named source. If a paragraph can be compressed into one sentence without losing value, it will be.
We think about this as the compression test. Before we write anything for a client blog, we ask whether an AI model could answer the question completely from public knowledge. If it can, we are writing a page that will get summarized and skipped.
The pages that survive are the ones carrying something only that business has. A studio can publish its own build results. A software company can publish its own usage patterns. A local firm can publish real pricing for its market. None of that exists in a model's training data.
This is also why we push clients toward proof over volume. A case study built on your own numbers is much harder to replace with a summary than a generic service page, because nobody else can copy those numbers.
Answer the question in the first 40 to 60 words under each heading, use headings that match how people actually ask, and keep the answer in clean text that a crawler can read without running JavaScript. Simple structure beats clever structure every time.
The technical floor matters more than people expect. If your article body only appears after a JavaScript bundle runs, some crawlers will never see it. We build content sites to render their text in the static HTML for exactly this reason, and it is the single most common fix we make on inherited projects.
Schema markup helps the machine understand what a page is. Google's own structured data documentation lists supported types including Article, Breadcrumb, Organization, Product, Review snippet, and Video. That list changes over time, so we check Google's documentation directly rather than trusting an old blog post about it.
Plain writing helps too. Short sentences with one idea each are easier for a model to lift cleanly. Long, hedged sentences with three clauses get skipped in favor of a competitor who said it simply.
Yes, and this is the hardest change for most teams. Clicks alone no longer describe your visibility. You need to track impressions, citation appearances in AI tools, branded search volume, and direct traffic, because influence now happens before anyone reaches your site.
Google Search Console still gives you impressions and average position, which tell you whether you are present in the result even when the click does not come. Watching impressions rise while clicks fall is evidence of zero-click exposure, not evidence of failure.
We also tell clients to search their own core questions in ChatGPT, Perplexity, and Google AI Mode by hand, once a month. It is manual and unglamorous, but it shows you whether you are being named. Semrush now sells an AI Visibility Toolkit that tracks brand mentions across ChatGPT, Gemini, and Perplexity, and tools like that scale better than manual checks. We still trust our own eyes on the handful of queries that matter most.
Branded search is the honest lagging indicator. When people start typing your company name into Google, something upstream is working, and that upstream is often an AI answer that mentioned you.
A bigger role, not a smaller one. When clicks are scarce, each one has to convert. A slow page wastes a visit you fought hard to earn, and speed also affects whether crawlers can fetch your content reliably at scale.
Our own project work keeps proving this. The average PageSpeed score across our builds is 98, and on the Offbeat Travel project we shipped a 0.9 second load with a 96 PageSpeed score alongside a 190% lift in organic traffic. Fast sites make the surviving traffic count.
There is a second reason speed matters here. AI crawlers request a lot of pages. A site that responds slowly or times out gets crawled less thoroughly, which means less of your content is available to be cited in the first place.
No, and we think that claim is lazy. Traditional search ranking is now the entry ticket to AI visibility rather than the finish line. Ranking in the top results is what makes you eligible to be cited, so the old fundamentals became a prerequisite instead of the whole game.
Here is our contrarian position. The businesses complaining loudest about AI eating their traffic were usually ranking for questions they had no special right to answer. They wrote the same "what is" article as everyone else and collected easy clicks for years. That model is over, and honestly it should be.
The businesses doing well are the ones with genuine expertise and the discipline to publish it clearly. That has always been the actual job. Zero-click search just removed the shortcut. If you want the mechanics of being quoted rather than summarized, our piece on getting cited by AI search engines goes deeper on the format side.
Start with one audit. Open Google Search Console, sort your pages by impressions, and find the ones where impressions are strong but clicks are weak. Those pages are already visible inside zero-click results. Rewrite them to answer faster and to carry something only you can say.
Then fix the boring technical layer. Make sure your content is in the static HTML, your headings match real questions, and your pages load fast. After that, shift some effort from volume to proof, because original numbers are the only thing a summary cannot replace. Our breakdown of how featured snippets work is a useful companion here, since the same answer-first structure feeds both snippets and AI summaries.
If you are staring at a traffic chart you cannot explain, we are happy to look at it with you. Send us the pages you care about and we will tell you honestly whether the problem is your content, your build, or just the shape of search in 2026. Let's talk, over at phoenix.studio.
Tell us where you want to go. We'll tell you how we'd get you there.