AI isn’t just changing where we search. It’s changing how we search.

A new report from Datos and SparkToro shows that Google hasn’t lost a significant amount of search traffic to its upstart competitors just yet — but generative AI chatbots are already affecting the tone of traditional searches. The rise of contextual answer-gatherers has led Google users to phrase their queries with more detail, and that strategy seems to be producing more accurate results.

In terms of the report’s topline findings, Google can breathe a sigh of relief. During Q4 2025, traditional search activity accounted for about 10% of U.S. desktop computer use. That figure was flat year-over-year, suggesting that Google search customers aren’t abandoning the tried-and-true platform.

The more interesting year-over-year trend concerns the number of searches per Google user. That figure fell by 20% among U.S. users, even as it stayed comparatively flat in Europe.

Could it be that U.S. Google users now find relevant results more efficiently than they once did? And if so, who is responsible? Google’s AI Overviews have added contextual (and controversial) insights to the top of results pages, but Datos’ click-tracking data show that Google AI Mode only accounts for a tiny percentage of desktop usage.

According to the report, user behavior is one main factor making searches more efficient. Queries between six and nine words are growing at the fastest rate of any length group measured by Datos and SparkToro. Some users even experimented with very long search queries consisting of more than 15 words.

The rise of TikTok as a search engine has shown us that younger generations prefer to add more context to their searches, and AI bots like ChatGPT saw their fortunes rise as they provided that more conversational form of inquiry. According to Datos and SparkToro, AI tools only account for .77% of U.S. desktop activity, but those programs are influencing more traditional rivals.

I’m not just talking about Google adding the contextual touches it thinks its search customers want — the Q4 2025 State of Search report uncovers major shifts in the language of search engines, and it’s hard to ignore the influence ChatGPT and its ilk have had in that area. We are teaching AI chatbots to think like humans, but the chatbots are also teaching us to think like machines. It’s a surprising result that reminds us that AI training is actually a two-way street.