Liz Reid, Google's Head of Search, gave the marketing industry a clear instruction this week. She just framed it diplomatically.
In a Search Engine Journal interview that ran Monday, Reid talked openly about the fragmentation of search behavior under AI. Users are submitting longer queries, more specific queries, and follow-up queries that depend on context the user did not have to supply in classic keyword search. The implication is that the keyword consolidation strategies most SEO programs have run on for a decade are now structurally misaligned with how people actually search.
This is not a small adjustment. The dominant model in SEO since 2014 has been to identify high-volume head terms, build pillar pages around them, and build supporting content underneath. The model worked because user search behavior was concentrated. The top 1,000 keywords for any given category typically captured 60 to 70% of the addressable search volume.
That concentration is dispersing. AI search rewards specificity, which means user behavior is shifting toward queries the keyword tools never measured because the queries did not exist a year ago.
What Fragmentation Means for Content Architecture
The structural shift is from query optimization to intent optimization.
The classic SEO page was built to rank for a specific keyword cluster. Title, headers, internal links, anchor text all optimized around a controlled vocabulary. The page that ranked for “best CRM for small business” was built differently from the page that ranked for “CRM software comparison.”
In a fragmented intent environment, the same underlying need shows up across hundreds of variations the keyword tools cannot enumerate in advance. A user asks “CRM for a 10-person agency that does retainer billing in Stripe.” The classic page architecture cannot match that query because the query was never identified. The AI engine, however, can pull a relevant chunk from a well-structured comparison page and surface it inside an AI-generated answer.
The architectural shift this forces is real. Instead of building 100 pages targeting 100 keywords, the more efficient structure is 20 deeply structured pages built around answer-shaped intent, with strong internal organization and clear chunk boundaries. The 20 pages cover 1,000 query variations through retrieval rather than direct keyword targeting.
What to Change in Your SEO Operations
The operational reset is not a full rebuild. It is a reweighting of effort.
First, audit the production pipeline. Most content teams are still operating on a keyword brief model where the writer receives a target keyword, a search volume number, and a competitor analysis. That brief format will not produce content that retrieves well in fragmented intent search. The new brief should specify the underlying user need, the operational context the user is likely in, and the specific decisions the content has to support. Keyword data still informs prioritization, but it stops driving the page structure.
Second, change how you measure success. Position tracking on head terms is becoming a lagging and increasingly noisy indicator. The leading indicators that matter now are citation rate in AI Overviews, ChatGPT, and Perplexity for the queries you care about. Run sample audits monthly. The brands that show up consistently are the ones doing the architectural work. The ones still chasing position one for high-volume keywords are watching their organic conversion rates drift down regardless of where they rank.
Third, shift the internal linking strategy. The classic hub-and-spoke model assumed users entered through a category page and traversed downward. In a fragmented intent environment, users are increasingly arriving directly at deep pages through AI-generated answers. The internal linking has to support both lateral discovery (related answers, related contexts) and upward navigation (broader category pages) because the entry point is no longer predictable.
Reid's framing was diplomatic. The substance was direct. The keyword era is dissolving into something that looks more like answer matching than search ranking, and the brands that restructure first will compound the visibility gains for years before the rest catch up.
The keyword tools are still useful. They are also no longer enough.
FAQ
What does keyword fragmentation in AI search actually mean?
Users are submitting longer, more specific queries and using context that the previous keyword model never captured. A single underlying need now manifests as hundreds of variations rather than a small set of high-volume head terms. The keyword consolidation strategies most SEO programs have run on for a decade are misaligned with how AI engines retrieve and answer these fragmented queries, because the engines work at the chunk and intent level rather than at the keyword level.
How should content architecture change to handle keyword fragmentation?
Move from building one page per keyword to building deeply structured pages around answer-shaped intent. Twenty well-built pages can cover 1,000 query variations through chunk-level retrieval, while 100 keyword-specific pages cover only the keywords they were targeting and miss most of the new fragmented variations. The brief that drives content creation should specify user need and operational context, not target keyword and search volume.
What metrics should replace keyword position as the primary SEO success measure?
Citation rate in AI Overviews, ChatGPT, and Perplexity for the queries that matter to your business is now the leading indicator. Position tracking on head terms is becoming a lagging and increasingly noisy signal. Run monthly citation audits across a representative sample of commercial queries to see which sources show up consistently, then reverse engineer what those sources are doing structurally and apply it to your own content.