A question, a summary, and sometimes no reason left to click. With answers generated by artificial intelligence, search engines no longer simply point the way: they explain, compare and recommend. For an advertising-funded media outlet or a store that relies on Google to bring in customers, this shift changes the equation. Visibility no longer guarantees a visit. Looking ahead to September 2026, this transformation calls for a rethink of a dependency built over two decades. The outlook presented here extrapolates from established developments in 2024; it is not a verified assessment of 2026.
From search engines that rank to engines that answer
The trend did not begin with ChatGPT. Featured snippets, how-to guides, sports results and local information were already keeping users on the search results page. Generative AI, however, vastly expands the scope: it can draw on multiple sources to answer a complex question, rather than display a single piece of information.
Microsoft integrated conversational features into Bing as early as 2023. Google launched its AI Overviews in the United States in May 2024 before expanding their availability. Perplexity popularized an interface combining written answers with references. In July 2024, OpenAI unveiled SearchGPT, a search prototype. These initiatives embody the same ambition: to become the place where a query is resolved.
The benefits are tangible. Planning a trip, understanding an administrative procedure or comparing two technologies can require less back-and-forth. But the errors observed when AI Overviews launched also exposed the limits of the approach. A fluent summary can draw on a dubious source or mistake a joke for advice. Convenience is no guarantee of accuracy.
Clicks become a less predictable resource
The web’s implicit contract was simple: publishers allowed search engines to crawl their pages; search engines sent them visitors. This exchange was never perfectly balanced, but it supported a clear business model. Synthesized answers change what publishers receive in return: content can feed the service without attracting visits proportional to its usefulness.
Imagine a website that publishes a guide to maintaining a coffee machine. If the search engine presents the essential steps, readers may leave satisfied without opening the article. The site then loses an opportunity to display an advertisement, offer a newsletter subscription or recommend an accessory. This scenario illustrates a risk; it does not measure an already established, across-the-board decline.
The impact is likely to vary by search type. A definition or a short procedure lends itself well to a self-contained answer. An investigation, an in-depth review or an exclusive local news story gives users more reasons to visit. Some clicks from a summary could also bring more qualified visitors. It is therefore important to distinguish between visit volume, commercial value and audience retention, rather than reduce everything to a traffic chart.
Media outlets face the source paradox
For newsrooms, the paradox is stark: original reporting becomes valuable to AI systems, while its distribution may become less profitable. Sending a journalist into the field costs money. Summarizing their work costs far less. If the interface captures the reader’s attention and relationship, those who fund the reporting may find themselves pushed into the background.
Two strategies have already emerged. Some groups are negotiating: the Associated Press reached an agreement with OpenAI in 2023, followed notably by Axel Springer that same year and the Financial Times in 2024. Others are challenging the use of their content, including The New York Times, which filed a lawsuit against OpenAI and Microsoft in late 2023. These approaches raise distinct questions about licensing, training and the presentation of content in answers.
An agreement does not solve everything, however. Its scope depends on the rights granted, attribution, compensation and oversight mechanisms. Major brands have bargaining power that is difficult for an independent outlet to match. The potential risk is a two-tier market: a handful of paid partners and a multitude of sources whose contributions remain difficult to monetize.
For retailers, being recommended is not enough
Commerce faces a variation of the same problem. When an assistant answers “which vacuum cleaner should I choose for a small apartment with a cat?”, it reframes the criteria, selects products and may narrow the comparison to a handful of models. The battle is no longer just about a page’s ranking, but about making it into that selection.
Retailers therefore have an incentive to focus on very practical details: reliable specifications, availability, up-to-date prices, delivery terms and return policies. Structured data and product feeds can make this information easier to use, without guaranteeing a recommendation. A product page padded with marketing language is no substitute for verifiable information, especially when several sellers offer the same item.
Looking ahead to 2026, one possibility deserves attention: customer journeys in which the assistant guides almost the entire decision before handing over a customer ready to buy. This traffic could convert at a higher rate while making the seller more dependent on the intermediary. The strategic question then becomes: who controls the relationship, the comparison and the terms of access to the customer?
Measure differently, diversify meaningfully
The first response is to stop relying solely on session counts to guide decisions. Publishers need to track subscriptions, direct return visits and sign-ups generated by search. Retailers need to examine margins, conversions and repeat purchases. Search queries must also be segmented: general information, brand searches and purchase intent do not necessarily experience the same effects.
The second response is to make a visit useful beyond the answer itself. For a media outlet, that might mean original documents, databases, expert analysis with a named author or exchanges with the newsroom. For a store, it could mean specialist advice, demonstrations, a product configurator and after-sales service. The aim is not to hide information artificially, but to offer something a summary cannot replace.
The third response is to rebuild direct channels. Newsletters, apps, communities and loyalty programs reduce exposure to a single channel, though none is a cost-free solution. Search engine optimization remains necessary, but it becomes one component of a distribution portfolio. As for formulas promising guaranteed inclusion in AI answers, caution is warranted: these systems evolve, and their selection processes remain partly opaque.
Negotiating a new distribution of value
Technical decisions alone will not be enough. Distinguishing between crawling for search, collecting data for training and using content in an answer is becoming essential. Blocking a bot can prevent certain uses while reducing discoverability, depending on the service involved. Discussions must therefore focus on understandable permissions, visible attribution and metrics that allow the value actually exchanged to be assessed.
What now? The most plausible scenario is not the disappearance of websites, but greater competition to justify every visit. If synthesized answers continue to gain ground, media outlets and retailers will need to invest as much in what makes them distinctive as in their visibility. The collective challenge will be to preserve an economy in which producing reliable information remains profitable: without solid sources, even the best answer engine eventually runs on empty.
