Instant answers make everyday searches easier, but new research raises questions about who benefits when the websites supplying information receive fewer visitors.
Searching for a medication interaction, a recipe, or an unfamiliar technology once meant opening several websites and comparing explanations. Increasingly, an artificial intelligence summary appears before the links. The answer arrives immediately; the underlying sources may never receive a visit.
For users, that can mean less searching. For publishers, independent researchers, and small businesses, it threatens the exchange that helped finance the open web: websites publish useful information, search engines help people discover it, and visits support advertising, subscriptions, or sales.
The question is not whether faster answers are useful. It is whether the businesses producing original information can survive if discovering their work no longer means visiting it.
People Are Clicking Less When AI Answers Appear
A Pew Research Center analysis tracked browsing activity from 900 American adults and examined 68,879 unique Google searches. When an AI summary appeared, users clicked traditional search results in 8% of visits, compared with 15% when no summary appeared. Just 1% of visits containing a summary produced a click on its cited sources.
People also ended browsing sessions more frequently after encountering AI summaries: 26%, versus 16% on pages without them. Longer, question-shaped searches were especially likely to generate AI answers.
These are observed associations, not proof that summaries caused every lost click. The kinds of questions triggering an overview may differ from searches that generate conventional results. Nevertheless, the findings identify where publisher visibility may fail to convert into readership.
Convenience Is Outpacing Trust
The appeal is understandable. A summary can combine definitions, explanations, and comparisons without forcing someone to sort through multiple tabs. A reader looking up a complex policy can ask for a simpler account before exploring details.
Yet convenience should not be mistaken for confidence. In a separate Pew survey of 5,153 American adults, 65% said they at least sometimes encountered AI summaries. Among those who had seen them, only 20% considered the information extremely or very useful, and just 6% trusted it a lot.
The Reuters Institute’s 2026 Digital News Report found that weekly use of standalone AI chatbots for news rose from 7% to 10% across its surveyed markets. Only 1% regarded AI as their main news source. Among chatbot news users, 42% asked follow-up questions, showing that interactive explanation—not simply replacing headlines—is part of the attraction.
The Economics Behind a Zero-Click Answer
Generative search often combines retrieval with language generation: a system locates material, then synthesizes a response instead of presenting only a list of destinations. That architecture changes the business relationship between a search platform and the websites it consults.
Cloudflare’s 2025 crawl-to-referral research estimated that Google’s bots retrieved about 14 pages per referred visitor in June, compared with roughly 1,700 for OpenAI and 73,000 for Anthropic. These measures reflect crawler requests and attributed visits, not revenue lost, and referrals from some applications can be difficult to identify.
Still, the imbalance matters. A publisher pays for reporting, expertise, fact-checking, and hosting even when an AI service answers a reader without sending traffic back. Fewer visits can make it harder to finance information that future search systems depend upon.
In a Reuters Institute survey of news leaders, publishers expected search referrals to decline 43% over three years. That is a forecast from industry respondents, not an observed outcome caused solely by AI.
An Answer Can Be Convenient and Incorrect
There is also an information-quality problem. Generative systems predict and assemble text; even when linked to retrieved documents, they can misidentify sources or present unsupported statements confidently.
The Columbia Journalism Review’s Tow Center tested eight AI search tools against news-article identification questions and found incorrect answers on more than 60% of its queries. The test was specific to citation accuracy, not a measurement of all answers across every search task.
For an everyday user, a misleading citation makes verification harder. For publishers, attribution errors mean original reporting may be summarized without directing readers to the organization responsible for producing it.
Search Companies and Publishers See Different Opportunities
Google disputes the idea that AI features necessarily diminish the web’s value. In its official explanation, search chief Liz Reid said overall organic clicks remained relatively stable year over year and that visitors arriving from AI-enhanced searches were more engaged. Google did not provide comparable underlying data in that statement.
The difference is important: fewer clicks on a particular results page can coexist with more searches overall or higher-value visits for some sites. Effects will vary by query, industry, and revenue model.
Google’s guidance for website owners continues to emphasize original reporting, distinctive expertise, and useful content over tactics designed only to influence AI rankings. For publishers, that suggests investing in reporting, communities, newsletters, and direct audience relationships that summaries cannot easily replace.
Publishers face a distribution dilemma. The Tow Center’s 2025 investigation of publisher controls found that 32% of the 50 US news websites blocked OpenAI’s search crawler. Blocking access may protect material from automated collection, but it can reduce opportunities for discovery through emerging interfaces. Allowing access, meanwhile, does not guarantee accurate citations or traffic. Neither choice offers a straightforward replacement for the old search bargain.
The Web Still Needs a Reason to Publish
AI search has improved the experience of finding quick explanations. Its longer-term value, however, depends on a less visible requirement: someone must investigate claims, maintain websites, update information, and produce discoveries worth summarizing.
For technology founders, search developers, and publishers, the central challenge is designing discovery systems that reward those contributions. Better attribution, measurable referrals, and sustainable licensing may matter as much as faster answers.
An internet that answers every question but undermines the people producing trustworthy information risks becoming less useful over time.
