For years, search intent was taught as four tidy buckets: informational, navigational, commercial and transactional. You matched a keyword to a bucket, wrote a page for that bucket, and moved on. That model is breaking down, and it's not just a theory — the data on how people actually search backs it up.

The old buckets are leaking

AI Overviews and chat-based search don't wait for a user to pick a bucket. Someone researching "best project management software" today might get a comparison, a recommendation, and a next-step suggestion in the same answer — informational, commercial and transactional intent, collapsed into one response. A query that used to sort cleanly into one funnel stage now often needs to serve three at once.

This isn't a guess. An Authoritas analysis of 10 million keywords across 14 industries found that mixed-intent queries — searches that don't fit neatly into a single traditional category — now account for roughly 73% of all search queries. A search for "best CRM software" could be someone comparing vendors before a purchase, or someone researching for a roundup article. The same phrase, two completely different intents, and increasingly, search engines are trying to satisfy both at once rather than picking one.

Where the searches are even happening has changed too

It's not just what people mean when they search — it's where they're searching at all. A Search Engine Land study from January 2026 found that 37% of consumers now start their searches with an AI tool rather than Google, a sharp increase driven largely by younger demographics. Traditional search hasn't gone away — the same research notes that roughly 90% of search volume and most commercial-intent, purchase-ready searches still happen on Google and similar platforms — but the starting point for research and comparison is splitting across two different paths.

Google's own product is adapting to this shift. A Semrush analysis of more than 600,000 keywords between November 2025 and April 2026 found that the share of commercial-intent search results pages showing an AI Overview grew 71% over that period. Interestingly, AI Overviews on purely transactional queries — the "I'm ready to buy right now" searches — actually fell 5% over the same window. Google appears to be leaning into AI summaries for the research and comparison stage of a purchase, while leaving the final "add to cart" moment closer to traditional results.

What this changes about planning content

The practical result is that content built to serve only one stage of the funnel is increasingly answering only part of the question a real person — or a real AI assistant, summarizing for a person — actually has. A handful of things follow from that:

  • Write pages that can answer the follow-up question, not just the headline one. If someone's asking "best X for Y," they're probably also wondering about price, setup effort, and alternatives — address those in the same piece rather than forcing three separate visits.
  • Put comparison and decision-support content earlier in the funnel than the old model would suggest, since research and commercial-investigation intent are now blending together well before someone's ready to buy. This is the same thinking behind how we approach content strategy generally.
  • Structure content so the parts that matter — the direct answer, the caveat, the recommendation — can be lifted cleanly by an AI system doing the summarizing, since a growing share of your audience may never see the full page at all. See our AI Search / GEO page for more on what that structuring actually involves.

None of this means throwing out intent-based planning entirely. It means the boundaries between intents matter less than whether a page can genuinely resolve what someone's trying to figure out, in one visit — whether that visit happens on a search results page or inside an AI assistant's answer. This is the thinking behind our own approach: understand the real question before building anything.

What this looks like on an actual page

Take a page currently built to rank for "best accounting software for freelancers." Under the old single-intent model, that page is pure commercial-investigation content: a list, some pros and cons, done. Under a multi-intent model, the same page also needs to quietly answer the informational question underneath it — what actually makes accounting software good for a freelancer specifically, as opposed to a small business — and increasingly, the transactional one too: pricing tiers, free trial availability, what switching from a spreadsheet actually involves. None of that has to bloat the page into an unreadable wall of text. It just has to be there, addressed directly enough that both a human skimmer and an AI system extracting an answer can find it.

Measuring whether it's actually working

The old proxy metrics — ranking position, raw sessions — still matter, but they don't tell the whole story anymore. Worth tracking alongside them: how often a page shows up in an AI Overview or gets cited by an AI assistant for its target query, and whether the traffic that does land on the page converts across more than one intent. Someone who arrives to compare options but also signs up for a trial on the same visit is a good sign the page is doing its multi-intent job properly. Neither of these numbers replaces rankings and sessions — they sit next to them.

Where this doesn't apply

Not every page needs to serve three intents. A pure navigational page — a login screen, a contact page — doesn't benefit from being stretched to cover research and comparison content that doesn't belong there. The shift matters most for the content that used to be the workhorse of a content strategy: comparison pages, "best X" roundups, and any page built around a decision rather than a fact.

The risk of overcorrecting

There's a failure mode on the other side of this too — pages so determined to cover every possible intent that they turn into unfocused, padded content nobody enjoys reading. The goal isn't maximum coverage, it's genuine usefulness across the two or three intents a real reader in that situation is likely to actually have. If a page needs a fourth or fifth intent bolted on to feel "complete," that's usually a sign it should split into two pages instead of one.