Study: AI shifts search demand, non-branded hit hardest

A new study of more than one million high-volume keywords finds that 29% of tracked search demand is in measurable decline, but the volume is being redistributed rather than destroyed, and non-branded informational queries are taking the hit. The research, published on 2 July 2026 by research agency Fractl with Search Engine Land, is one of the largest public datasets yet on how AI is changing search behaviour.

What happened

The study analysed 1,010,848 keywords with at least 10,000 monthly searches, drawn from Semrush data across 379 brands in eight verticals, alongside a survey of 1,004 US consumers. According to the authors, the declining keywords represent roughly 10.29 billion monthly searches while growing keywords represent roughly 10.31 billion, an almost exact offset. Information-heavy categories fell hardest, with FinTech down 37.7%, while transactional categories such as SaaS, Lifestyle, and Insurance held steady or grew. The consumer survey points the same way: 70% of respondents say they use AI more than a year ago, yet only 17% say they use traditional search less.

59% of consumers surveyed say they will visit a brand’s website after an AI chatbot mentions it, and 18% have bought something based on an unverified AI recommendation.

Fractl and Search Engine Land, “What 1 million keywords reveal about AI’s impact on search”

Why it matters

Non-branded queries are the most vulnerable to AI replacement, and they make up 90% of the volume the study tracked. Those are exactly the discovery searches, the category and “near me” style queries, that fill the top of the funnel for location-based businesses. As AI Overviews, AI Mode, and chat assistants absorb informational intent, the question shifts from how a brand ranks on a results page to whether AI systems mention it at all. The study’s consumer data suggests that when they do, the mention converts: a majority of consumers follow up on brands an assistant names.

What this means for multi-location brands

For an enterprise brand with hundreds of locations, demand redistribution means the discovery layer is moving to systems that answer instead of listing ten links. AI assistants ground their local answers in structured data about businesses, so the estate-wide accuracy of names, addresses, hours, and services across Google Business Profile, Apple Maps, and Bing becomes the input that decides whether your locations are quotable. Central teams should manage business listings at scale as the data foundation, measure how AI surfaces each location with Places AI, and treat ranking in AI search results as a discipline with its own metrics rather than a by-product of classic SEO.

The bottom line

Search demand is not disappearing, it is moving, and the losses are concentrated in the non-branded queries that used to feed local discovery. Brands that make every location’s data accurate and machine-readable now will be the ones AI answers cite when those queries stop reaching the results page.