Written by Julien Ricciarelli-Bonnal
18 September 2026
The essentials
People appear to be moving away from very short searches and towards longer, more specific queries. A recent analysis of Google Ads data shows that three- to four-word searches are gaining significantly in impressions and conversions, while one- and two-word queries are declining. Google has also said that queries submitted through AI Mode are, on average, three times longer than traditional searches. The change goes beyond the search engine itself: after years of learning how to speak in keywords, users are beginning to explain more naturally what they actually want to find.
For nearly twenty-five years, Google taught us to simplify the way we searched. A full question became two or three words, stripped of articles, context and sometimes even grammar. A user would not ask, “Which laptop should I choose if I work regularly on trains and need strong battery life?” They would type “best laptop battery life”. It was not a natural way of expressing a need, but it fitted the logic of the search engine perfectly.
Conversational interfaces are starting to change that habit. When users interact with ChatGPT, Gemini, Claude or Perplexity, they can describe a problem, add a constraint, specify a budget or ask for a comparison. Search engines are now adopting the same logic. Google is expanding AI Overviews and AI Mode, while making search increasingly capable of interpreting complex questions rather than simply matching a few terms with an indexed page.
The first available data suggests that this change is not confined to AI interfaces. It is beginning to influence traditional Google search as well.
Three- and four-word queries are taking a much larger share
An analysis published in September by Search Engine Land, based on Google Ads search term data, provides a useful view of how this behaviour is evolving over almost two years. Between January 2025 and August 2026, the share of impressions generated by one- and two-word queries reportedly fell from 42% to 24%. Over the same period, searches containing three or four words increased from 33% to 48% of impressions.
This is not an official Google study covering all global searches, so the figures should not be treated as universal. But the trend is significant enough to matter, particularly because it extends a movement already observed on the same dataset a year earlier.
Google itself provides another signal. Earlier this year, the company said that queries submitted through AI Mode were, on average, three times longer than those entered into traditional search. Users are therefore not simply searching for the same things through a different interface. They are expressing more of their intent.
That may seem like a small change when reduced to a few extra words in a search box. In practice, it changes the amount of information a user gives the engine before seeing the first result.
After learning how to speak to Google, users are starting to speak normally again
Traditional search created a strange but widespread skill: learning how to remove information from a question in order to increase the chances of getting a useful answer. For years, adding too much context could feel like making the search unnecessarily difficult. Users therefore learned to isolate the two or three terms they assumed mattered most.
Generative models reversed that logic. The more contextualised the request, the more tailored the answer can usually become. Saying that you are looking for a laptop for travel, with a budget of €1,200 and with battery life taking priority over graphics performance is no longer noise. It is useful information.
As people become accustomed to using these interfaces, they may naturally carry the same behaviour elsewhere. Someone who regularly describes a need in detail to an AI assistant has less reason to return automatically to telegraphic syntax when opening Google.
Google is also encouraging the shift itself. With AI Mode, the company explicitly wants users to ask longer, more complex and more specific questions, while allowing follow-up requests that preserve the context of the initial query. The search engine that once trained people to compress their intentions is now encouraging them to expand them.
Longer searches may also sit closer to a decision
The change becomes particularly interesting for marketers when conversions are considered. In the Google Ads analysis, short queries still account for a significant share of conversions, but their weight is falling. Three- and four-word searches, by contrast, are gaining substantially.
The reason is straightforward. Longer queries often contain more clues about the user’s actual intent. “Car insurance” opens an enormous field. “Car insurance young driver no excess” already describes a situation, a profile and an expectation much more precisely. “CRM software” could correspond to a very broad exploration stage, whereas “CRM for 20-person SME Outlook compatible” dramatically narrows the search.
A long query is not automatically commercial. Some highly detailed informational searches may never lead to a purchase. But when commercial intent does exist, additional context can help the search engine, the advertiser and the content provider understand much more clearly what the person is looking for.
This development is also a reminder that a marketing strategy can no longer be limited to a handful of historic keywords. If people are expressing their needs with greater precision, understanding the situations, problems and criteria surrounding a search becomes as important as knowing the most frequently typed phrases.
SEO and paid search will have to understand richer intent
For years, part of search marketing was organised around volumes attached to relatively stable query formulations. Identify a keyword, measure its potential, create a page capable of answering that search and improve its visibility. The reality was always more complicated than that, but the underlying logic remained easy to understand.
The growth of longer queries fragments demand further. Ten users looking for the same category of product may now describe their needs in ten different ways, adding constraints, use cases or specific situations. Value therefore shifts gradually away from matching one exact formulation and towards understanding and covering the underlying intent.
Paid search is going through the same transformation. Google has spent years pushing advertisers towards broader matching systems and campaign types in which the algorithm interprets more of the search context. An increase in query length and variety reinforces that logic. Predicting every relevant formulation manually becomes increasingly unrealistic.
This does not mean keywords disappear. They remain useful for understanding a market, structuring campaigns and analysing demand. But they are becoming less capable, on their own, of describing how a person frames a problem.
Businesses will need to answer questions, not simply occupy queries
The same transformation has direct consequences for content. When a user formulates a more precise search, a vague page built around a broad term may answer the need less effectively than content capable of addressing the criteria behind that search.
Companies therefore have good reason to look beyond the most obvious search volumes. Sales conversations, customer service questions, objections raised before purchase and the criteria people use when comparing alternatives are becoming even more valuable inputs. They increasingly resemble the type of detail users can now include directly in a search.
The answer is not to create one page for every possible sentence. That would simply replace an old obsession with keywords with a new obsession with long-tail phrasing. The objective is to understand intent well enough to create content capable of answering several formulations of the same underlying problem.
Search is therefore becoming more technological and more natural at the same time. Engines are relying on increasingly sophisticated models to interpret requests, while users have less need to adapt their language to the machine.
For years, using Google meant learning how to compress a problem into a handful of words. Generative interfaces are gradually removing that constraint. If the trend continues, businesses will need to think about more than where they appear on a results page. They will need to understand the full question that comes before that appearance.
Google is not simply adding AI-generated answers to search. It may also be changing a habit that has existed for decades: the way people formulate what they are looking for.
If changes in search behaviour are forcing you to rethink visibility, content or acquisition, we can help you identify the adjustments your marketing strategy now requires to remain aligned with how people actually search.
Written by Julien Ricciarelli-Bonnal
18 September 2026
