Google is quietly turning Search into an AI assistant

For more than two decades, Google's homepage has represented a simple contract with the web: enter a query, receive a list of results, choose a source. Google is now testing something quite different. SFGATE reported this week that, for some desktop users, the familiar Google Search button has been replaced by three AI-oriented actions: Create images, Ask about files and Brainstorm. Google characterised the change as a small desktop experiment and said it does not alter core search functionality. Traditional search remains available by entering a query. But the buttons are not the main story. What they represent is.

<h2>Search is becoming an interface, not a results page</h2>

Traditional search followed a comparatively straightforward sequence: query, results, website. Google organised information, the user selected a result, and the website supplied the answer. Generative AI changes that sequence: question or task, AI interpretation, synthesised answer or action, possibly a website. The word possibly matters. Google has already moved in this direction through AI Overviews and AI Mode. The homepage experiment goes one step further by placing AI actions directly where the conventional search interaction used to sit. The boundary between searching Google and asking Google's AI is becoming increasingly difficult to define.

<h2>The click is losing its central role</h2>

The behavioural consequences are already visible. A recent Pew Research Center study, based on browsing data from a representative panel of 900 US adults, found that users clicked a source cited within an AI Overview on only about 1% of visits where an AI Overview appeared. The researchers also found that traditional result clicks occurred on 8% of visits with an AI Overview, compared with 15% of visits without one, while browsing sessions ended more frequently on pages with an overview. The study is observational and does not establish causation. Still, it illustrates the changing role of the website: its information can contribute to a consumer's decision even when the consumer never visits the source directly.

<h2>The SEO versus AEO debate may eventually look artificial</h2>

There has been considerable debate over terminology: AEO, GEO, AI SEO, LLMO, or simply SEO evolving. Google's product evolution suggests that the distinction may ultimately matter less than the operational reality. If traditional search, AI answers and AI assistants converge into the same interface, organisations will not be optimising for completely separate environments. They will be optimising for an AI-mediated discovery system that sometimes produces links, sometimes produces answers and increasingly may perform actions.

SEO remains essential because search infrastructure, websites, authority and discoverability still matter. But optimisation now has another audience: the system interpreting that information on the user's behalf. The questions change accordingly. Can Google find and rank the brand? Does the AI understand it? Which claims does it associate with it? Which sources does it trust when describing it? When does it recommend the brand rather than a competitor? What does it say when the customer never reaches the website? Those are not merely ranking questions. They are questions about representation inside AI-mediated decisions.

<h2>This is bigger than three buttons</h2>

Google may abandon this particular homepage experiment. Technology companies test interfaces constantly, and a limited test is not a product commitment. But the direction of travel is clear. Search is becoming more conversational, generative, multimodal and, increasingly, agentic. The user is not necessarily asking Google to find a webpage anymore. They may be asking it to do something. For the past 25 years, the internet economy was built around winning the click. The emerging AI-search economy may be built around winning something earlier: the answer, the recommendation and eventually the action.

<h2>AEO Updates Takeaway</h2>

Google's three experimental buttons are not the story. The story is that the world's dominant search engine is gradually becoming an AI assistant with a search engine underneath it. Brands should treat that shift as a design constraint for their visibility strategy. The asset that matters is no longer only the page that wins a click. It is also the credible, retrievable, well-qualified information that helps an AI interpret the brand correctly before a user has any reason to visit the page at all.

<h3>References</h3>

[1] SFGATE, <a href='https://www.sfgate.com/bayarea/article/google-search-changes-22381960.php' target='_blank' rel='noopener noreferrer'>Google is quietly testing out new ways to crush the human spirit</a>, August 12, 2026.

[2] Pew Research Center authors, <a href='https://arxiv.org/pdf/2608.04831' target='_blank' rel='noopener noreferrer'>Investigating Click Behaviors On Google Search Result Pages That Produce an AI Overview</a>, August 2026.