For years, customers learned to search the way our systems expected. They shortened questions, removed context and tried different combinations of keywords until something useful appeared.
Generative AI is changing that habit. Increasingly, people describe what they are trying to accomplish. They include a budget, an operating condition, a problem or a preference. They ask follow-up questions. They expect the experience to understand how those details fit together.
After nearly 20 years working across search strategy, analytics and digital discovery, I see a significant opportunity in that change. Customers are giving organizations more information about their needs at the moment they need help.
The challenge is turning that additional context into a more relevant response. A longer query has little value if the experience still ignores the requirements inside it.
The Evidence Is Moving Beyond the AI Search Box
Google reported in May 2026 that the average U.S. AI Mode search was three times the length of a traditional Search query. It also reported growth in both short and long queries within AI Mode. That is an important distinction: conversational search is expanding how people ask, while familiar search behaviors continue alongside it.
Evidence from paid search points in a similar direction. In a September 2026 analysis published by Search Engine Land, Jason Tabeling reported that one-to-two-word queries fell from 42% to 24% of impressions in the Google Ads dataset he analyzed between January 2025 and August 2026. Three-to-four-word queries increased from 33% to 48%.
These are advertising impressions within an analyzed dataset, rather than a count of every Google search. Even so, the movement matters. More specific queries are becoming an increasingly important part of the demand businesses need to understand.
Forrester provides another piece of the picture. Its February 2026 Consumer Pulse Survey found that 26% of U.S. online adults had used ChatGPT in the previous month to find products they were interested in buying, up from 19% in October 2025.
In B2B, Forrester reported in October 2025 that 95% of buyers planned to use generative AI in at least one part of a future purchase. More than half said it had helped them consider additional or different vendors while saving time.
Adoption figures do not measure query length. They show why the change deserves attention: AI-assisted discovery is becoming part of how both consumers and business buyers research decisions and expect answers.

Coveo Data Shows the Change Inside Enterprise Experiences
The evolution is also visible within Coveo-powered search experiences.
Across 65 million finance and insurance queries analyzed from July 2024 to July 2026, average query word count increased 17.1% following adoption of Coveo Relevance Generative Answering (CRGA). Average length rose from approximately 3.5 words in 2024 to 4.1 in 2026.
A separate consumer experience analysis found a 21% relative increase in the share of queries containing five or more words between 2023 and 2025. Two-to-four-word searches remained the dominant behavior.
That combination is particularly useful for search leaders. Customers are expressing more detailed needs, while short searches remain central to the experience. The opportunity is to improve understanding across that full range of the customer journey.
These observations are consistent with behavior evolving during the generative AI era. They do not isolate AI as the sole cause, or establish a universal growth rate across industries. Interface design, audience mix and the tasks people bring to search can also affect query length.
My interpretation is that organizations should prepare for a wider range of customer expression. Someone may enter a product name in one interaction, describe a complex requirement in the next and then refine it with a short follow-up or need a conversational search experience. Each interaction can carry useful intent.

B2C and B2B Customers Are Revealing Different Kinds of Context
Consider a consumer who searches for “walking shoes.” The category is clear, but much of the decision remains hidden.
Now consider “waterproof walking shoes under $150 for a week in rainy weather.” The customer has supplied a feature, a budget and a use case. That information can help an experience narrow the options and explain why a product is relevant.
It also gives the business something useful to learn. Repeated questions about fit, suitability, delivery or compatibility can expose gaps in product attributes and supporting content. If those details are missing, adding conversational AI will not make the underlying information more complete.
In B2B, the additional context often concerns whether a solution will work at all.
“Replacement pump” becomes “replacement pump compatible with our existing assembly for continuous operation at 80 degrees Celsius.” An evaluator asking about support search may specify multilingual content, permission-aware answers and links to source documentation.
These examples show that identifying the topic is only the first step. The experience must also distinguish between customer preferences and requirements that need to be verified. As B2C and B2B customers come to expect richer, more responsive interactions, organizations must evolve or risk losing traffic and customers to competing services. “We have always done it this way” is no longer a viable response.
Exact searches still matter. A part number, an error code or a product identifier may communicate more precise intent than a long question. Follow-up messages can also be short because the context is already established. Word count helps describe behavior, but it cannot explain the entire task.

Site Search, SEO, and GEO Need to Work From the Same Customer Questions
For Search teams, this evolution changes how we should interpret demand. A keyword remains useful, but it often represents only part of what a customer needs to know.
Someone researching an industrial component may need operating limits, compatibility, documentation and availability. Someone comparing a consumer product may need fit guidance, care instructions or evidence that it suits a particular use case. Content planning should account for those connected questions.
My recommendation is to organize around the decision. Give the customer a clear answer, explain the conditions that apply and make the supporting information easy to find. Use descriptive headings and consistent product terminology. Keep specifications and policies current. Build related content around genuine questions rather than producing a page for every possible wording variation.
That approach also supports generative engine optimization, or GEO. When an answer engine synthesizes information, the organization needs its facts, relationships and limitations to be represented accurately.
Google explains that its AI search features may issue multiple related searches across subtopics and sources to construct a response. My implication for content teams is practical: make each important section understandable and useful on its own. Identify the product, service or policy being discussed. Connect a recommendation to its evidence. State where the guidance applies and where it does not.
Google also makes clear that established SEO practices remain relevant to its AI features. There is no special schema or AI-specific file required for inclusion. Good technical SEO and trustworthy content remain foundations; they do not guarantee that a page will rank or appear in a generated answer.
The Customer Journey Continues After the External Answer
Customers do not experience Site Search, SEO, GEO, commerce search and knowledge retrieval as separate organizational functions. They experience a question and the effort required to resolve it.
Forrester has described the convergence of search marketing, commerce search and enterprise search as generative AI makes experiences more conversational and assistive. I see that convergence as an operating responsibility: the teams shaping external visibility and internal discovery need to work from consistent knowledge and customer intent.
For a Coveo-powered experience, that means evaluating query understanding, retrieval and the response together. Coveo documents its Semantic Encoder as retrieving indexed content through semantic similarity, helping support complex queries and retrieval for generative answering. Coveo Relevance Generative Answering (CRGA) retrieves enterprise content to ground responses, provides citations and enforces content permissions during retrieval.
Those capabilities create a foundation for responding to richer requests. Their effectiveness still depends on the information available and how the experience is configured.
For a B2C journey, test whether a budget, occasion or product preference leads to appropriate results. For B2B, test technical requirements, compatibility and access rules. Where the information cannot verify a requirement, the experience should make that limitation clear.
The query is an opportunity to improve knowledge as well as results. When customers repeatedly ask a question the organization cannot answer, the next improvement may belong with a product data owner, a content team or a knowledge specialist.

Measure Whether the Experience Uses the Additional Context
As an analyst, I want to know whether changing query behavior leads to better outcomes.
Start with the distribution of query lengths, alongside the average. Separate submitted searches from search-as-you-type events and system-generated searches. Distinguish the first question from later conversational turns. Compare consistent audiences and account for language, device and seasonal differences.
Then connect the behavior to the task. In B2C, examine conversion, add-to-cart rate, abandonment and revenue per search session. In B2B, include successful product identification, quote requests and order completion. In support, look for verified resolution, fewer repeat contacts and appropriate escalation.
Review reformulation as well. A customer adding detail may be refining a useful result, or compensating for an experience that failed to understand the first question. Examine the sequence before deciding what the longer query means.
Time on Search Results Page and Deep Time on Page can add context, provided they are interpreted against the task. More time can reflect productive evaluation or unnecessary effort. The useful question is whether the customer reached a successful outcome with confidence.
Start With the Questions Already Reaching Your Business
Organizations can act on this change now. Review the last 90 days of customer search activity. Identify the recurring questions, the requirements inside them and the points where the experience fails to respond. Bring those findings to the teams responsible for content, product data, relevance and measurement.
Use external discovery research to broaden that view. Compare the questions customers ask answer engines with those they bring to your website, commerce experience or support portal. Look for inconsistent facts, missing guidance and unnecessary repetition across the journey.
After nearly 20 years in search, I believe this is one of the most valuable opportunities the AI era has created: customers are giving us a clearer view of what they need. Organizations that listen carefully and respond with trusted, relevant information can turn that context into better decisions and measurable business value.
Sources
[1] Google. How AI Mode is changing the way people search in the U.S. May 19, 2026. https://blog.google/products-and-platforms/products/search/ai-mode-us-insights/
[2] Jason Tabeling, Search Engine Land. Google Ads data shows query length shift post-AI Mode. September 11, 2026. January 2025 to August 2026 comparison; paid-search impression shares within the analyzed dataset. https://searchengineland.com/google-ads-data-shows-query-length-shift-post-ai-mode-458162
[3] Forrester. US Consumers’ Adoption Of Answer Engines For Product Discovery Is Gaining Traction. April 7, 2026. Public report summary citing the February 2026 Consumer Pulse Survey and October 2025 comparison. https://secure.forrester.com/report/us-consumers-adoption-of-answer-engines-for-product-discovery-is-gaining-traction/RES193423
[4] Jessie Johnson, Forrester. From Keywords To Context: Impact And Opportunity For AI-Powered Search In B2B Marketing. October 2, 2025. https://www.forrester.com/blogs/from-keywords-to-context-impact-and-opportunity-for-ai-powered-search-in-b2b-marketing/
[5] Coveo finance and insurance query analysis. July 2024 to July 2026; 65 million queries; reported 17.1% increase in average query word count following generative answering adoption. Rounded annual averages approximately 3.5 words in 2024 and 4.1 in 2026.
[6] Coveo consumer experience query analysis. 2023 to 2025; reported 21% relative increase in the share of queries containing five or more words. Customer identity withheld. The Coveo analyses describe separate cohorts and periods; full sampling and control methodology was not available for independent review.
[7] Google Search Central. AI features and your website. https://developers.google.com/search/docs/appearance/ai-features
[8] Forrester. GenAI Converges Search Experiences And Disciplines. April 1, 2025. https://www.forrester.com/blogs/the-sudden-silobreaker-genai-converges-search-experiences-and-disciplines/
[9] Coveo documentation. About Coveo Machine Learning and AI. https://docs.coveo.com/en/1727
[10] Coveo. Your Entire Content Library Does Not Have to Be AI Ready to Get Started. https://www.coveo.com/blog/enterprise-search-generative-ai-content-readiness/
