Until recently, someone planning a trip to Japan during cherry blossom season would most likely turn to travel websites, read articles about bloom dates, and compare recommendations. Today, that same question can often be answered directly in search: artificial intelligence pulls information from multiple sources, assembles its own broad recommendation, and saves the user from having to browse through page after page.
That is convenient for the user, but the consequences for website owners can be severe. A travel publication spends money producing destination guides in the hope of attracting readers through search and earning advertising revenue, while an online retailer publishes reviews and buying advice to draw potential customers to its site. If a search engine can now answer the very questions those pages were created to address, both businesses risk losing part of their audience-and with it, potential customers.
Large publishers are already seeing traffic from search decline and are being forced to look for other sources of revenue. For online retailers, the picture is less straightforward: AI may reduce website visits, but it can also introduce shoppers to products and brands they might never have found otherwise. The economics of the internet are beginning to shift.
The Answer Is There, but the Click Never Comes
In 2025, the Pew Research Center examined how people use Google Search and how often they click through to websites when an AI-generated answer appears. Researchers analyzed the search activity of 900 U.S. adults in March of that year and found a clear difference: when an AI summary appeared, users clicked on traditional search links in 8% of visits, compared with 15% when no AI summary was shown. Links cited directly inside the AI-generated answer were clicked in only 1% of visits.
Publishers themselves are also reporting declining search traffic. According to a January 2026 Reuters Institute report, visits from traditional Google search to more than 2,500 news websites fell by 33% between November 2024 and November 2025, with the decline reaching 38% in the United States. It is still impossible to determine exactly how much of that loss was caused by AI: search-algorithm changes, shifting audience behavior, and the fact that AI-generated answers do not appear for every query all affect the results.
Google, for its part, says the overall volume of clicks from traditional search to websites remained relatively stable over the previous year. According to the company, users are asking more complex questions, while those who do visit websites after seeing AI-generated answers often show stronger interest in the content they find. Google does acknowledge, however, that traffic is being redistributed: some websites are losing visitors while others are gaining new audiences.

What Happens to Revenue When Fewer Readers Arrive
In November 2025, U.S. publishing group People Inc., which owns titles including People and Food & Wine, presented investors with data showing how traffic to its digital properties had changed. In the third quarter of 2023, Google Search accounted for 54% of visits to the group's main websites; two years later, that share had fallen to 24%. The publisher continued producing content and maintaining a large audience, but increasingly had to reach that audience through other channels.
For an advertising-supported media business, declining traffic has a direct financial effect. Every pageview creates an opportunity to display an ad, while a page that is never opened generates no revenue from that impression. People Inc. has already encountered this problem: in the third quarter of 2025, digital advertising revenue fell by 3%, with the company attributing part of the decline in programmatic advertising revenue to the spread of Google's AI-generated answers.
At the same time, total digital revenue for the publisher rose 9% during the same period, reaching $269 million. Growth came from other parts of the business, including affiliate commerce and content licensing. In other words, People Inc. was able to offset part of the decline in advertising revenue with income streams that did not depend entirely on pageviews.
Analytics company Piano found a similar pattern after examining hundreds of publishing websites in 2025. Search traffic across the sample fell by 36%, while revenue from new subscriptions attributed to search declined by 16%. The analysis suggested that a large share of the lost audience consisted of visitors who opened a single page and rarely returned. They contributed significant traffic volume, but relatively little subscription revenue.
For a publisher that relies mainly on advertising impressions, losing even casual visitors can hurt. But when a large share of revenue comes from subscribers, the behavior of the loyal audience matters more. Two websites can therefore experience the same decline in traffic and face very different financial consequences.
Publishers are gradually rethinking how they reach readers. People Inc. is expanding its presence on third-party platforms and developing businesses that generate revenue independently of Google referrals. For other media companies, newsletters are becoming one way to maintain a direct relationship with the audience: subscribers receive new stories directly from the publication, so another change to search results does not cut off access to them. That strategy takes time, however, because readers still have to be persuaded to return to a particular source even when the answer they want is already available inside search.
And while publishers are currently feeling the effects of AI mainly through lost visits, in e-commerce the same technology is beginning to reshape how shoppers discover products.

A Shopper Can Discover a Brand Without Ever Opening Its Website
Consider a small cosmetics brand that sells products through its own online store. In the past, a potential customer might discover it through Google, open a product page, and study the ingredients and specifications. Today, she might ask an AI system to recommend a skincare product that meets a specific set of needs and receive several suggestions with explanations. The introduction to the brand now happens before the shopper ever visits the store, and the AI may recommend a product from a company she had never heard of before.
In September 2026, Salesforce published research on changes in online shopping based in part on data from more than 1.5 billion shoppers. The company found that traffic to retail websites from AI chats rose by 150% to 428% year over year across different quarters, while overall traffic grew much more slowly. This is still a relatively new acquisition channel, so the very high growth rates are partly explained by a small starting base.
For businesses, this creates a new way to reach customers without relying on traditional search advertising, while also making competition harder to predict. If someone asks an AI assistant to recommend a cosmetic product, they will not necessarily see items from the brands that rank highest on Google. The AI may draw on product specifications published by retailers, independent reviews, and other available sources. The manufacturer, meanwhile, may have little idea why its product appeared in the recommendations-or why it was omitted.
Measuring the effectiveness of marketing becomes more complicated as well. Suppose a shopper first hears about a cosmetics brand from an AI recommendation, then reads reviews and visits the online store directly several days later. In sales analytics, that order may appear as a direct visit even though the initial interest came from another service. The company can see that a purchase occurred without always being able to identify what originally brought the customer in.
That means online retailers have to rethink some of their familiar metrics. Google referrals still matter, but they no longer tell the whole story: a shopper may first encounter a product in an AI chat and complete the purchase later on the seller's site. Sales analysis increasingly needs to consider actual purchases and returning customers rather than judging marketing performance from traffic alone.

The way companies manage product information is changing too. If an AI system recommends a product using outdated specifications or an incomplete description, the shopper may come away with the wrong understanding of what it does. Businesses therefore have reason to keep product information up to date not only on their own websites but also on the platforms where their products are listed. This does not guarantee that a product will appear in AI-generated recommendations, but it reduces the chance that a potential customer will encounter inaccurate information.
Traditional online stores are not disappearing. People still need to check availability and complete purchases, while fully autonomous shopping through AI agents remains at an early stage. But product discovery is increasingly beginning outside the seller's website, and companies are already having to account for that path when planning how they sell.
Who Will Pay for the Information AI Retells?
For publishers, one possible way to offset the loss of search traffic is to license content directly to AI companies. In November 2025, People Inc. announced an agreement with Microsoft to participate in the creation of the Publisher Content Marketplace. The platform is intended to allow publishers to make their content available to AI developers for a fee, with Microsoft Copilot expected to be the first buyer.
Google is also expanding its relationships with media companies. In October 2026, the company said that more than 400 publishers worldwide, including CNN, the Financial Times, and Axios, were taking part in its News AI pilot program. Under the initiative, Google pays for access to additional content for new news features, including products within Google News. These are still individual commercial agreements, however, rather than a system that compensates every website owner whose reporting is used in AI-generated search responses.
Large publishers have recognizable brands and extensive archives that are valuable to AI developers, giving them some leverage to negotiate commercial use of their content. Smaller specialist websites may also hold highly valuable information, but not all of them have the scale or resources needed to secure similar agreements.
At the same time, publishers are reconsidering what they produce in the first place. According to the Reuters Institute's January survey, many media executives said they planned to increase investment in original investigations and first-hand reporting while reducing the volume of standard informational content.
That decision reflects the nature of the information itself. A search engine can compile general facts from many sources that already exist, while an original interview with an entrepreneur only becomes possible after a journalist speaks directly with that person and obtains new information. Once the interview is published, however, an AI system can summarize that too, which means originality alone does not guarantee that readers will return to the source.
Search engines can increasingly provide answers without sending users to the websites that produced the underlying information, but those answers still depend on someone else doing the work first. Whether the people and companies creating that information can continue to be paid for it under the new model may ultimately determine how sustainable AI-powered search becomes.