Google is no longer a search engine

The days when Google was a search engine are, in many ways, over. According to some analyses , the correct term to describe the world's most visited website is now "answer machine." This isn't simply a lexical issue, but a profound transformation of the very nature of Google and the function it plays within the internet. This transformation took another important step forward with the introduction, in May 2024, of AI Overviews—the artificial intelligence system within Google Search, which generates content in response to our searches—but which had already begun several years ago.

It was 2018 when the Mountain View giant introduced the latest version of "featured snippets." Through these boxes, displayed at the top of the page, Google no longer simply displays traditional links, but provides the most relevant excerpt from a web page for our search. A few lines that in many cases—game scores, song lyrics, dates of historical events, and other brief information—are sufficient to satisfy users' requests, who therefore no longer need to click on the links provided.

For example, if you search for "how fast does a cheetah run," you won't need to click the first link that appears (which, in my case, takes you to the newspaper Focus.it) to get the information you want: simply read the snippet, which reports that the feline's top speed is between 80 and 130 kilometers per hour. Through snippets, Google has begun to directly provide answers and, in a sense, cannibalize the web. The introduction of generative artificial intelligence, however, has accelerated this transformation.

The advent of ChatGPT in 2022, and the ever-increasing percentage of people using it to search for information, has posed an existential risk to Google and other search engines, which have since suffered a steady decline in visits. As reported by Search Engine Journal, between May 2025 and the same period last year, Google lost 2% of visits, Bing 18%, Yahoo 11%, and Baidu 12%.

One of the reasons why Google has managed to stem the decline in visits compared to its smaller siblings is linked precisely to the introduction of AI Overviews, which allowed Big G – always a leader in the artificial intelligence sector – to respond quickly to the threat of ChatGPT.

The integration of AI Overviews into the search engine, however, has dramatically increased so-called “zero-click” searches, in which Google is not the starting point for browsing the web, but rather the only site we need to visit, within which we can find all the answers to our questions and requests.

A recent Pew Research study , which analyzed the behavior of 900 Google users, gives us an idea of exactly how much the world of online searches is changing. First, 58% of them conducted at least one search in March 2025 that resulted in an AI Overviews summary. Predictably, when this happens, people are much less likely to click on links Google displays anyway. Specifically, those who don't encounter an Overviews summary click on the links displayed twice as often (15% versus 8%). This data alone tells us that AI Overviews could potentially halve the traffic Google has historically sent to the rest of the web.

What's more, as the research states, "Google users are more likely to completely complete their browsing session after visiting a page with an AI-generated summary than on pages without one. This occurred in 26% of cases where an AI-generated summary was present, compared to 16% on pages with only traditional search results."

Overall, in March 2025, 18% of Google searches returned AI Overviews content. This percentage is sure to rise, raising several questions: what are the risks for users? And what impact will all this have on the open web, which has traditionally received a significant share of traffic—and therefore advertising revenue—from Google?

For users, the main risk is the persistent— and perhaps unsolvable —problem of hallucinations, when a large language model presents information as if it were facts that are actually incorrect or completely fabricated. As a general rule, all AI-generated content should be checked for authenticity, but how many users actually do this?

The Pew Research study provides an answer, according to which only 1% of users click on the links accompanying the AI Overviews summary, while 8%, as mentioned, open the classic links displayed on the rest of the Google page. This means that only 9% of users have the opportunity to verify the accuracy of the information provided.

Another risk is that AI-generated summaries offer a single answer, whereas a traditional Google search offers a range of results, allowing us to choose from sources we deem most reliable or closest to our sensibilities. This is especially important when our searches involve sensitive topics, such as the climate crisis, the war in Ukraine and the Middle East, or other geopolitical issues (for example, are we confident that AI Overviews will provide us with an accurate account of Google's controversial programs in collaboration with the US and Israeli militaries?).

The other key aspect concerns the impact that Google's transformation (and the spread of similar systems, starting with ChatGPT) will have on the health of the open web. According to the Wall Street Journal, "in the media world, Google searches generate about 40% of all traffic received by the most important websites." The shift in searches in favor of generative artificial intelligence systems and the adoption of this technology by Google and all the others has led publishing groups to "estimate a loss of around 20-40% of the traffic generated by Google," the WSJ writes.

In a world perpetually in financial difficulty like the publishing and journalism world, Google's transformation from a search engine to an answer machine—and the resulting decline in visits—could therefore represent yet another very severe blow.

Finding a solution (including the financial agreements that many major companies, including Gedi , have entered into with OpenAI, Google, and others) is in everyone's interest. On the other hand, if newspapers and other websites stopped providing information due to the spread of Overviews and other similar systems, how would the AI giants train their large language models with up-to-date news and services?