citelity.Join waitlist →
May 10, 2026·11 min read·zero-click-search

What is zero-click search? Why most Google searches no longer send traffic

Zero-click search is when a user gets their answer on the results page itself — from an AI Overview, featured snippet, knowledge panel or People Also Ask box — and never visits a source. It's the majority of Google searches, it has been rising for a decade, and it isn't reversing. Here's what actually causes it, which pages are most exposed, and the strategic response that works.

Zero-click search is when a user gets their answer on the search results page itself and never clicks through to a source. The answer comes from an AI Overview, a featured snippet, a knowledge panel, a People Also Ask box, or a direct answer. SparkToro's clickstream analysis has tracked the majority of Google searches ending this way for years — a trend that started with featured snippets around 2014 and accelerated sharply once AI Overviews rolled out. It is not a temporary algorithm phase, and there is no version of the last decade in which it reverses. The strategic response isn't to wait it out. It's to become the source cited inside the answer, which is what AEO makes possible.

If you've watched your rankings hold while your clicks fell, you've experienced this without necessarily having a name for it. Impressions up, position stable, traffic down. That pattern is the signature of zero-click, and it's why "just rank better" stopped being useful advice.

What counts as zero-click

A zero-click search is one where the user's need is met inside the results page. That can happen through:

  • AI Overviews — a generated answer above the results, citing several sources but usually complete enough on its own
  • Featured snippets — the "position zero" box lifting a paragraph or list from a single page
  • Knowledge panels — entity information for people, companies and places
  • People Also Ask — expandable Q&A pairs that load in place
  • Direct answers — conversions, time zones, calculations, scores, weather
  • Local packs — a map and business listings that answer most location queries outright
  • Image and video results — when the intent was to see rather than read

In every case the click that used to happen doesn't. For a publisher, each one is a query that once sent traffic and no longer does.

How it got here

The trajectory has been visible for over a decade, and each phase reduced clicks further without any phase reversing an earlier one.

Worth being precise about what's measurable here: SparkToro's clickstream work is the source usually cited for the headline share, and it measures searches ending without a click to the open web. Breakdowns by query type, device or country circulate widely but rarely come with published methodology — treat those with more caution than the headline figure.

Why "rank better" stopped working

For years the mainstream response to featured snippets was to treat them as another thing to win. The implicit assumption — that being featured would preserve or grow clicks — turned out to be wrong in a specific way.

Winning position zero captures more of a smaller pool. When a snippet exists, total clicks across all results drop, because the answer is already on screen. Being the source captures a larger share of what remains, but what remains has shrunk.

AI Overviews compound that. An Overview synthesizes several sources into an answer that often resolves the query completely. Cited sources get attribution and brand exposure; the click rate is lower than the pre-Overview equivalent.

Intent framing was too coarse. Segmenting queries as informational, transactional or navigational implicitly assumed informational intent still required a visit. It increasingly doesn't.

The conclusion most SEO writing was slow to state: a higher ranking on a zero-click query produces very little additional traffic. The lever has moved.

AI
Free tool · No signup
Free Schema Validator
Paste any URL → find the structured data your page is missing, with ready-to-paste JSON-LD fixes.
Check your schema

Which pages are most exposed

Some query types have moved almost entirely to zero-click. Others still send traffic reliably. The pattern below is a directional ranking based on how these query types behave, not a set of measured rates — nobody publishes zero-click percentages broken out by page type, and any article that gives you eight decimals of precision here has invented them.

Page typeExposureWhy
Quick factualHighestGoogle answers these natively. There is no source to click.
Glossary / definitionalHighAn Overview synthesizes a definition from several sources, and the result is usually fuller than any one page.
"Best of" comparisonHighOverviews summarize the options well enough that many readers stop there.
Local / locationHighThe map pack resolves most location intent in place.
Detailed tutorialsModerateMulti-step work with screenshots is harder to compress into an answer.
TransactionalLowThe user wants to act, not to learn. The click is the point.
Long-tail technicalLowestToo specific for confident synthesis; engines defer to specialist sources.

The strategic read: shift new content investment toward the bottom of that table, and apply AEO work to protect what's at the top.

The shift: optimize for citation, not clicks

The mental adjustment is the whole thing. Instead of competing for a shrinking pool of clicks, compete to be inside the answer that replaced them.

Being cited is worth more than it looks on a traffic chart. Your name appears in the answer a reader trusts. Some fraction click through, and those who do arrive further along than a cold search visitor — they've already seen you positioned as a source. The rest register the brand for next time.

What to measure, since clicks alone now understate your position:

  • Click-through traffic — still the baseline, still worth watching
  • Citation rate per representative query, sampled manually across engines
  • Brand impressions in cited contexts, even without a click
  • Conversion quality from cited traffic compared to general organic

The optimization work itself is covered in our AEO content checklist, and the underlying mechanics in what is AEO.

Can you opt out?

Partly, with a trade-off that's usually not worth it.

Robots.txt directives can block specific AI crawlers. But the crawlers divide into classes that do very different things, and blocking the wrong one is the most common self-inflicted wound in this space:

  • Training crawlers — GPTBot, ClaudeBot, CCBot, Google-Extended. Blocking these keeps your content out of training corpora. It does not affect whether you get cited.
  • Retrieval crawlers — OAI-SearchBot, PerplexityBot, Claude-SearchBot, Bingbot. Blocking these removes you from that engine's answers entirely.

Blocking Googlebot itself removes you from Google Search altogether, losing both the Overview citation and the ranking beneath it.

For nearly every site, the realistic choice isn't opting out of zero-click. It's deciding whether you appear inside the answer or not at all.

AI
Free tool · No signup
Free AEO Content Score
Paste content or a URL → a 0-100 score across ten AEO signals, with the fixes worth doing first.
Score your content

What adapted sites have in common

Patterns worth copying, observed across sites that have handled the shift well:

They publish less and deeper. Fifty substantive pieces outperform two hundred thin ones. Depth survives synthesis; thin content gets absorbed into it.

They name their authors. Anonymous "editorial team" content appears to be filtered harder, especially where trust matters.

They publish original data. A number you measured yourself can't be synthesized from elsewhere. It gets cited as a primary source, and other people's articles cite it too.

They build things you have to use. Calculators, validators, configurators. An interactive tool cannot be zero-clicked — the user has to arrive to use it.

They diversify. Newsletter, community, social, direct. Not as a growth hack but as insurance against a channel whose rules they don't set.

They track citations, not just rankings. Weekly manual checks on a fixed query set. It's the only way to know whether the work is landing.

FAQ

What share of Google searches are zero-click?
SparkToro's clickstream analysis has found the majority of Google searches ending without a click to the open web, and the share has risen steadily since featured snippets appeared around 2014. Precise figures vary by study, methodology and period, so treat any single number as an estimate rather than a constant. Breakdowns by query type or device circulate widely but usually without published methodology behind them.
Why has zero-click search grown so much?
A decade of incremental Google product changes: featured snippets from 2014, People Also Ask expansion from 2018, rich results and direct answers through the early 2020s, and AI Overviews from 2024. Each moved more of the answer onto the results page. None has been rolled back. The direction serves Google, which keeps users in its own surfaces, and serves users, who get faster answers — while reducing traffic to the sites the answers are built from.
Can I stop AI from using my content in zero-click answers?
Partly, and the trade-off is usually bad. Robots.txt can block AI crawlers, but they fall into two classes with opposite consequences. Blocking training crawlers like GPTBot, ClaudeBot or Google-Extended keeps your content out of training data and does not affect citation. Blocking retrieval crawlers like OAI-SearchBot, PerplexityBot or Bingbot removes you from those engines' answers entirely. Blocking Googlebot removes you from Google Search altogether. For most sites the realistic choice is whether to appear inside the answer or not at all.
Which content still reliably drives clicks?
Four types hold up. Long-tail technical content, where queries are too specific for confident synthesis. Tutorials with original screenshots or video, which an answer can't reproduce. Interactive tools and calculators, which require the user to arrive. And original research, which gets cited as a primary source precisely because it can't be synthesized from anywhere else.
Should I stop publishing informational content?
No — the goal changes rather than the format. Informational content still earns traffic when it becomes the cited source inside the answer, and citation carries brand exposure that a traffic chart doesn't show. What changes is the standard: a direct-answer opening, real schema, named authorship and citation-ready structure are now the difference between being the source and being one of the pages the source was assembled from.

Closing

Zero-click is the baseline now, not a phase. The searches that stopped sending traffic are not coming back on any trajectory the last decade supports.

What's left is a different game with a different scoreboard. Be inside the answer rather than below it. Build the kind of thing an answer can't replace — original data, real tools, genuine depth. Track citation rate the way you used to track rankings. And spend some of your effort somewhere Google doesn't set the rules.

If you're starting: audit your top twenty pages, sort them by the exposure table above, apply AEO work to the most exposed, and point new content at the least. Defensive on what exists, offensive on what's next.


Sources cited in this piece

  • SparkToro — clickstream analysis of zero-click search behaviour on Google; the source for the headline share of searches ending without a click.
  • Aggarwal et al., "GEO: Generative Engine Optimization" — arXiv 2311.09735, KDD 2024. Background on which content structures affect citation in generative engines. arxiv.org/pdf/2311.09735
  • Crawler classification (training vs retrieval) reflects the operators' own published bot documentation.

Last updated 31 July 2026: removed per-page-type zero-click percentages that had no published source, replaced them with a directional exposure ranking, and added the crawler-class distinction.

Written by
Ed Grows
Building citelity — the AEO stack for indie founders, content sites, and SaaS products. Sharing what works in real-world testing.
← Back to all posts