Changes in AI search in the last month

Given the speed of development in this field, it doesn’t make sense to look at developments from the last quarter or last six months. Instead, we focus on what happened in March.

AI is now a primary search behaviour

A study released this month suggests that AI tools now account for roughly 56% of global search engine volume—not search traffic, but search behaviour. The sessions people used to spend on Google are increasingly happening inside ChatGPT, Gemini, Perplexity, and Claude.

The number, from Graphite.io CEO Ethan Smith, has been contested. Measuring AI "search" against traditional search is inherently difficult to do cleanly. But even if the true figure is half that, the direction is clear: AI has become a primary discovery surface for a large and rapidly growing audience.

The shift is happening faster than brands are adapting

Google still dominates. But its share of search-related activity reportedly fell from 89% in 2023 to 71% in Q4 2025. That is not a gradual decline—it is a structural shift unfolding over two years.

The change is also uneven across markets. In the US, usage of AI assistants grew roughly 300% year-on-year by December 2025, while global growth has begun to plateau. Adoption is concentrated in high-value, commercially active audiences.

From choice to answer

This shift matters not just because of where people search, but how they search.

A Google query returns a list of options. A ChatGPT query returns a single, synthesised answer—often without clear next steps or links.

The brand mentioned in that answer occupies a fundamentally different position than a brand ranked fourth in a traditional search result.

Being found is no longer enough. Being recalled is what counts.

From discoverability to recall

Harvard Business Review put it succinctly this month: companies must move from optimising for clicks to engineering recall within AI systems—by publishing original data, naming proprietary frameworks, and associating insights with credible experts.

Traditional SEO rewarded discoverability. AI rewards memorability.

To be surfaced in AI-generated answers, brands must generate enough signal—across authoritative sources—that they become part of what the model “reaches for” when synthesising responses.

Authority over keywords

This is a different kind of work.

Less about keywords, more about authority.
Less about your own website, more about the ecosystem around you.

Research consistently shows that most LLM citations come not from brand-owned sites, but from third-party editorial content, social platforms, and community forums.

What this means in practice

The brands currently winning in AI-mediated search tend to share a few characteristics:

• A clear, consistent point of view repeated across contexts

• Original data and proprietary thinking cited by others

• A structured, ongoing approach to AI visibility (not a one-off audit)

The bottom line

The methodological debate around the 56% figure will continue. But the reality it points to is already here.

If your visibility strategy is still centred on ranking in a list of blue links, you are optimising for a shrinking surface—while the one that is growing remains largely unmeasured.

Sources

• Graphite.io / Ethan Smith, AI search volume study (March 2026)

• Harvard Business Review, LLMs Are Overtaking Search (March 2026)

• Search Engine Land reporting on Rand Fishkin / SparkToro rebuttal (March 2026)