Zero-click search: when Google and AI answer instead of websites, what should businesses do?

Zero-click search is when users get their answer directly on Google's results page — from the AI summary at the top or a quick answer box — or ask ChatGPT or Gemini directly, without clicking through to any website. According to SparkToro research based on Similarweb data, 68% of Google searches in the US in the first four months of 2026 ended without a single click. When Google shows an AI summary, users click even less. For businesses, this means appearing in the answer matters as much as ranking, and website traffic is no longer the only measure that counts.

What is zero-click search?

Zero-click search is a search in which the user does not click any result that leads to another website. They may have read the answer on the results page, clicked into another Google service such as Maps or YouTube, typed a different query, or simply left. This is not new: quick answer boxes, business profile panels, exchange rates, weather and football scores have kept users on Google's page for years.

What is new is that artificial intelligence has turned zero-click from something that affects short questions into something that affects almost every question. In the past, to find out "which accounting software should a 20-person company choose", users had to open several comparison articles. Now Google AI Overviews, AI Mode, ChatGPT or Gemini can put together a long answer with bullet points and a few small source links — and most users stop there.

Two things are often lumped together and need to be kept apart. Zero-click rate is the share of searches that lead to no click out to a website. CTR (click-through rate) is the share of people who see a specific result and then click it. AI can raise the first and lower the second at the same time, but studies measure them differently, so the figures cannot be directly added to or subtracted from each other.

What percentage of searches get no click?

The latest figure with a public methodology is about two thirds. SparkToro, Rand Fishkin's research company, used Similarweb's user panel data and concluded: "In the first four months of 2026, a whopping 68.01% of Google searches ended without a click.". The data comes from browser users on desktop and mobile in the US, January–April 2026, and according to the author it does not include the Google mobile app. In other words, the figure holds for the US market, not globally.

The same research team compared six markets over the same period: the UK was highest at 69.5%, followed by the US at 68.0%, France 65.3%, Canada 63.8%, Italy 63.4%, and Germany lowest at 62.1%. The gaps between countries are small, which suggests this is a general trend in how Google displays results rather than something specific to one country. The study has no data for Vietnam or Asia.

Compared with two years earlier, the zero-click rate has risen. The SparkToro study published in July 2024 found that 58.5% of searches in the US and 59.7% in the EU got no click, using Datos data. The 2026 article recalculates the 2024 US figure as 60.45%, and the author notes that comparing the two years is "a bit of apples and oranges" because the datasets differ. Whichever baseline you pick, the direction is clear: the share of users leaving Google for a website is gradually falling.

One notable detail in the 2026 study: only 0.34% of searches moved to AI Mode in the first four months of the year. So most zero-click activity still happens on the familiar results page, with the AI summary at the top, rather than because users have switched entirely to chat mode.

How much do AI summaries reduce clicks?

Every independent study reaches the same conclusion: when Google shows an AI summary, users click through to websites much less. They differ only in how much, depending on how it is measured.

The most reliable study of actual user behaviour comes from Pew Research Center. Pew tracked the browsing history of 900 US adults in March 2025 and found that people who saw an AI summary clicked a traditional search result in only 8% of visits, while those who did not clicked almost twice as often, at 15%. More strikingly, users clicked a source link inside the AI summary in just 1% of visits. And 26% of visits with an AI summary ended the browsing session altogether, compared with 16% on pages with only traditional results.

From the website owner's side, Ahrefs analysed 300,000 keywords in Google Search Console and concluded that, as of December 2025, AI Overviews reduced the click-through rate of the top result by 58%. Ahrefs' own previous measurement in April 2025 was 34.5%, meaning the drop is deepening over time. Seer Interactive tracked 3,119 keywords across 42 clients and found that organic click-through rate for informational queries with AI Overviews fell 61%, paid ads fell 65%, and even queries without AI Overviews fell 25–41%.

A randomised experiment by two researchers, Agarwal and Sen, with 1,065 desktop Chrome users in the US found that AI Overviews cut clicks out to websites by 38% on queries that showed a summary, and the zero-click rate rose from 54% to 72%. This is a draft not yet peer-reviewed; we read the results through a report by Search Engine Journal and could not open the original paper.

Besides the figures above, many articles are circulating the claim that "over 80% of searches with AI Overviews get no click". We have not found a readable original source for this figure — the pages that cite it attribute it to different companies — so we do not use it. The closest verified figure is 72%, from the Agarwal and Sen experiment.

Why is zero-click growing so fast?

The simple reason is scale: billions of people are now used to getting answers from AI. At the I/O event on 19 May 2026, Google CEO Sundar Pichai announced that AI Overviews have more than 2.5 billion monthly users, and AI Mode passed 1 billion monthly users after just one year. The Gemini app passed 1 billion monthly users in August 2026. On the OpenAI side, CEO Sam Altman said ChatGPT has more than 800 million weekly users, according to TechCrunch (October 2025) — the latest figure we could verify.

Interestingly, the number of searches has not fallen. In 2024, Gartner forecast that search volume on traditional engines would drop 25% by 2026 — we could not open Gartner's original press release and only read it through a report on Slashdot. Meanwhile, Google says queries on Search just reached an all-time high in the most recent quarter. People are searching as much as before, or even more; what has changed is that they click through to websites less.

Google also says AI still sends traffic to websites. In its Q2 2026 earnings report, Mr Pichai said Google is sending "billions of clicks to websites every week through AI features in Search". Both can be true at once: total clicks remain large, but each website's share is shrinking, and clicks are concentrating on the sources that AI chooses to cite.

What about Vietnam?

Vietnam is already covered by both of Google's biggest AI features. The official help pages for AI Overviews and AI Mode both list Vietnam among the supported countries and Vietnamese among the supported languages. AI Overviews began rolling out in Vietnam in late October 2024, according to Báo Quốc tế (The World & Vietnam Report).

There is currently no public study measuring the zero-click rate specifically for Vietnam with a clear methodology like the studies above. So we do not give a figure for the Vietnamese market. But the factors that drive zero-click — AI summaries in Vietnamese, and ChatGPT and Gemini working in Vietnamese — are all already in place. Vietnamese businesses should prepare as if this trend will follow the same path as in the US and Europe, perhaps just a few steps behind.

The most affected group is websites that rely on informational queries: news sites, how-to blogs, product comparison sites and glossary pages. Less affected are pages with a concrete action that AI cannot do for the user: ordering, signing up, booking, downloading documents, or looking up a customer's own data.

What should businesses do about zero-click?

The first step is to change the goal: from "how much traffic can we pull in" to "in how many answers are we mentioned and trusted". According to Google's official guidance for site owners, there are no extra requirements or special optimisation techniques for appearing in AI Overviews or AI Mode: a page only needs to be indexed by Google and eligible to appear with a snippet. Google also says impressions in AI features are counted in the Search Console Performance report, under "Web". In other words, the foundation is still basic SEO done right, plus a writing style that makes it easy for AI to cite.

In practice, there are five things businesses can do right away:

• Answer directly in the first sentence. Open each major section with a complete answer that still stands on its own when cut out of the article. This is the passage AI cites most often.
• Put data and sources right in the paragraph. A specific figure with a link to the original source is more credible than a vague "faster, cheaper".
• Use tables for comparisons. Price tables, option comparisons and timelines are easy for both people and machines to read correctly.
• Mark up structured data. Declaring business information, products and FAQs with schema helps search engines understand the entities correctly.
• Keep pages fresh. Show the update date clearly and correct outdated figures; stale content is less likely to be chosen as a source.

The second step is to measure differently. If you look only at traffic, you will see the numbers fall and assume your content has got worse, when in fact your brand may be mentioned more often. The table below is Namtech's proposed measurement framework, not research data.

The third step is to build channels that do not depend on clicks from Google: email lists, customer communities, Zalo OA, video channels, and above all a brand strong enough that customers type the company name directly. When AI summarises in place of websites, the name mentioned in the answer is what customers remember when they need to buy.

Businesses can also choose not to be cited. Google allows the nosnippet, data-nosnippet, max-snippet or noindex tags to limit what is shown in AI features. This suits paid content or proprietary data, but the trade-off is losing the chance to appear in answers at all.

Namtech's view: zero-click is also happening inside businesses

Zero-click is not only about Google. The same habit — ask AI, get an answer, never open the original document — is spreading into how employees work every day. Instead of opening the process folder or searching old emails, employees ask ChatGPT or Gemini directly, and often paste contracts, price lists and customer data into it to get a quick answer.

If the answer comes from a public AI service, the business faces two risks: internal information leaves the organisation, and the answer is not based on the company's actual documents. Namtech's approach is to bring the "ask and get an answer" experience inside: an AI assistant running on the company's own server, answering from internal documents and citing the exact original document so employees can check it. Employees still get the speed of zero-click, while the company's data and knowledge stay in the company's hands.

The lesson is the same on both sides: when users no longer click through to the source, the quality and reliability of the source matter even more. To be cited outside, content must be clear and well-grounded. For employees to use AI safely inside, internal documents must be organised so AI reads them correctly.

Zero-click search does not remove the need to search; it shifts value from clicks to citations — the businesses that answer clearly, with data and sources, will be the names AI mentions.

FAQ

Note: This article compiles public sources checked on 3 November 2026. Zero-click studies currently measure mainly the US and European markets; there is no public data specific to Vietnam yet. The Agarwal and Sen results and the Gartner forecast are cited via secondary sources; we could not read the originals. Table 3 is Namtech's proposed framework, not research data.