Is AI Killing Search and SEO?

In August 2025, OpenAI announced a huge milestone: ChatGPT hit 700 million active weekly users, a number that has quadrupled since last year. That’s great for OpenAI’s bottom line, but it may spell disaster for any company that relies on traditional search engines.

That’s because users increasingly rely on ChatGPT and other generative AI tools for many queries they would typically input into a search engine like Google or Bing. According to Similarweb, there are more than 250 million daily visits to websites from generative AI tools as of August 2025. That’s up from just over 200 million daily visits just seven months earlier. More than 84% of that traffic currently comes from ChatGPT.

The more that generative AI tools fulfill queries for users that used to rely on search, the fewer clicks go to websites, and the fewer eyeballs end up on search engine ads. This dynamic caused Google’s search market share to drop below 90% for the first time since 2015, according to Search Engine Land.

The implications for any business or service that relies on search engine optimization (SEO) are vast. If the number of users of generative AI tools keeps growing at its current blistering rate, many worry that traffic and clicks to websites from search engines could fall off a cliff. That would torpedo existing business and monetization models across many corners of the Internet.

That’s leading many to ask: Is SEO dead? And, if so, what comes next?

The Death of Search?

It’s not just ChatGPT eating into Google’s market share that’s causing problems. The rise of generative AI itself has Google starting to cannibalize its own traditional search experience.

The company has launched AI Overviews, which summarize the answers to certain search queries above the typical list of blue links that users click on to visit websites. This makes search results more resemble the conversational natural language provided by a tool like ChatGPT.

“What seems to be true is that AI Overviews reduce user traffic to websites, since more users are not clicking on the result links anymore after getting the answer directly,” said Janek Bevendorff, a language technology and information retrieval researcher at Germany’s Bauhaus University Weimar.

That’s backed up by at least one recent study. In Spring 2025, Pew Research Center analyzed online browsing activity by 900 U.S. adults. They found that users who encounter AI summaries in Google search results are far less likely to click on links to other websites. They only clicked on links in 8% of visits versus 15% who clicked on links when they didn’t encounter an AI summary. With Pew finding that one-in-five Google searches currently produce AI summaries, that’s a big hit to website owners who rely on clicks from search.

While that’s bad news for content providers, it’s probably insufficient to conclude that SEO as a whole is dead, said Bevendorff. SEO involves far more than optimizing rank position in search results. To appear in AI Overviews, a page still needs to be relevant, discoverable, structured, and machine-readable by crawlers and content extractors.

“SEO will almost certainly adapt to these new conditions,” Bevendorff said. That seems born out by the data, at least for now. Whether it’s through traditional or AI-generated search, Google still receives 373 times the amount of searches that ChatGPT does, according to SparkToro. But will it last?

From SEO to GEO

Times are changing and content providers see the need to adapt to AI-generated search results. One way many are pursuing is generative engine optimization (GEO), or the process of structuring and publishing content so it’s more likely to be included in AI-generated results.

The idea is: the large language models (LLMs) that power tools like ChatGPT and Google’s AI Overviews are periodically trained on a vast amount of data from the Internet, then use that memorized training, plus a range of weights, rules, and considerations, to deliver a natural language answer that best addresses a user’s query. GEO aims to influence how LLMs deliver these results.

Some strategies that GEO proponents recommend include things like structuring website content to directly answer common questions, since AI engines prefer concise, self-contained information they can easily extract. They also advocate producing content and commentary across a wide variety of channels and mediums like podcasts and media coverage, not just on a website or blog. That provides an even greater chance of a brand, business, or content getting ingested and used in an LLM’s training data.

GEO doesn’t necessarily mean a wholesale change in strategy or approach, either. Many GEO strategies also end up focusing on the same issues as SEO, said Bevendorff.

Bevendorff noted that traditional SEO techniques can be applied to information used during retrieval augmented generation (RAG), which some AI systems rely on to retrieve information to answer user queries. Tools like ChatGPT will also increasingly initiate conventional Web searches to supplement the information they’ve been trained on with up-to-date, real-time information. All of that comes from a Web that’s optimized for standard SEO. That means plenty of sites and companies are still getting benefits from SEO, even in an AI-first search environment.

“The major new challenge here is to appear relevant not only to the retrieval system, but also to the result selection process for the following generation step,” Bevendorff said. Without any type of source retrieval, placing a link to a brand or product into a fully AI-generated response is much, much harder. In fact, it is probably unattainable for most content producers, because it borders on data poisoning for large language model training data.

“As such, the turnaround times would also be bound to model training cycles and the effectiveness evaluation of respective GEO measures would be difficult at best,” said Bevendorff.

Translation? What’s old is new again. In the past, website owners would speculate endlessly about what actions they could take to make Google happy, resulting in a mix of data-backed SEO advice and less-scientific “hope and pray” tactics. Now, the same exact thing is happening as everyone scrambles to guess what might make LLMs happy.

Logan Kugler is a technology writer specializing in artificial intelligence based in Tampa, FL. He has been a regular contributor to CACM for 15 years and has written for nearly 100 major publications.

You Just Read

Is AI Killing Search and SEO?