Flexible Search Content Has a Role: 44% Say Keywords or Questions Depends on the Topic

Search behavior is becoming more flexible as people move between keywords, full questions, AI summaries, and traditional results. The strongest pattern in the survey is choice: users select the format that fits the task.

A September 2026 ContentEngine survey of 100 U.S. respondents shows that search habits are changing in several directions at once. Many respondents read an AI summary before scanning links, use chatbots during complex research, and write longer queries than they did 18 months earlier. They still want access to detailed websites, human-written articles, and sources they can check.

That combination gives content teams a practical signal. Search optimization needs to support short keyword searches and conversational questions. It also needs to account for the two-step path that is becoming common: an AI-generated summary helps someone decide what matters, then a website provides the evidence, context, instructions, or expertise required to act.

The ContentEngine survey found that 65 percent of respondents always or frequently read Google’s AI-generated summary before looking at traditional links. Twenty-six percent said the summary was the first thing they looked at, while 39 percent used it to decide whether they needed to scroll further. A stratified estimate placed the figure at 65.9 percent.

That is a significant change in the search interface. A webpage may receive attention before a searcher ever sees the title, description, or position of a traditional result. The summary acts as a first filter. It can help someone decide whether a subject deserves more time, whether a result appears relevant, and whether a source seems worth opening.

The survey also shows why summary visibility should not be treated as the same thing as website replacement. Respondents described several different behaviors:

The largest group uses the summary as a screening tool. That group may click later, after deciding that the subject requires more detail. A page that answers the basic question quickly and then supports deeper investigation has a useful role in that process.

The finding needs to be read alongside behavioral research from the Pew Research Center. Pew found that Google users were less likely to click result links when an AI summary appeared on the page. A separate study using browsing data from 900 U.S. adults found that clicks to sources cited in AI Overviews were rare, occurring in about 1 percent of visits to pages with an AI Overview. Google has reported a different measure, saying that average click quality has increased and that it sends slightly more quality clicks to websites.

These findings measure different parts of the journey. Self-reported survey answers, browsing behavior, source-link clicks, and Google’s quality-click metric cannot be treated as interchangeable. They do point to a search environment where exposure, click volume, click quality, and source selection may move in different directions.

Search behavior is not moving uniformly from keywords to conversational questions or from websites to AI tools. The ContentEngine survey suggests that people choose the format that fits the task: short keywords for clear needs, longer questions for complex decisions, AI summaries for orientation, and websites for verification and depth. For content teams, the strongest strategy is flexible rather than exclusive. Pages should address both keyword and question-based intent, provide a direct answer near the top, and give readers credible evidence or useful detail that earns further attention.

The survey found that 42 percent of respondents believe their typical queries have become longer over the past 18 months. Nineteen percent report significantly longer queries, such as full sentences or questions. Twenty-three percent use slightly longer searches, often three or four words instead of one or two.

The weighted estimate raises the longer-query figure to 49 percent. That is close to half of the sample, although the result still should be described as a substantial minority rather than a universal shift. Forty-five percent say their query length has stayed the same, and 13 percent say their queries have become shorter.

Longer queries appear especially useful to people who have adopted them. Among respondents who report longer queries, 81 percent express positive satisfaction with searches containing five or more words. Among respondents whose query length stayed the same or became shorter, 58.6 percent express positive satisfaction. The difference is meaningful within this sample, although the survey cannot establish that longer queries caused the higher satisfaction.

When asked directly about more specific searches, 68 percent say they are either much more satisfied or somewhat satisfied. Thirty-four percent say they find exactly what they need faster. Another 34 percent see relevant results but sometimes encounter clutter. Only 6 percent report dissatisfaction because additional words seem to confuse the search engine.

Longer searches help when they add useful constraints. A question such as the following gives a search engine more direction than a single broad term:

how long does a cast iron skillet take to season in an oven

A business search may benefit from the same approach:

accounting software for a small construction company with job costing

The extra language identifies the problem, audience, and desired outcome. It can also reveal the searcher’s stage of research. Someone searching for a definition may need a short explanation. Someone asking for a comparison, a process, or a recommendation may need a page with evidence and practical detail.

An AI summary compresses information before a searcher chooses a result. That changes the job of a source page. The page must provide material that can be understood in a short synthesis, while also giving readers a reason to continue.

A useful source page usually has several visible qualities:

Structured data can help search systems interpret a page, although markup does not guarantee inclusion in an AI response. The content itself still needs to be accurate, readable, and supported by a credible source.

The 44 percent who read an AI summary and then click for details are especially important. They are looking for a second layer of information. A page that repeats the summary in longer form may give them little reason to stay. A stronger destination might show the method behind a recommendation, explain an exception, include a calculator, compare alternatives, or provide first-hand evidence.

That distinction also changes how content performance should be evaluated. A page may appear in an AI summary and receive fewer direct clicks while gaining brand exposure. Another page may attract fewer visits but generate longer sessions, more qualified leads, or more return visits. Search reports need to separate visibility from visits and visits from meaningful engagement.

The most useful finding is not that AI summaries or longer queries are replacing traditional search; it is that search is becoming layered. A person may use a summary to decide what matters, a chatbot to organize research, and a human-written page to verify a consequential decision. That makes rigid keyword-only or AI-first strategies too narrow. Content teams should build pages that are easy to summarize, difficult to replace, and transparent enough to trust, while measuring qualified engagement rather than treating every click as the only sign of success.

Trust in AI-generated search responses is conditional across the survey sample. Forty percent trust AI responses about the same as a standard website. Thirty-five percent are skeptical or express no trust. Twenty-five percent trust AI responses more than standard websites or trust them completely.

The relationship between trust and summary use is strong. Among respondents who express higher or complete trust, 96 percent always or frequently read AI summaries first. Among skeptical respondents, 31.4 percent still read summaries regularly.

That second figure matters. People can find a tool useful while remaining cautious about its accuracy. A summary may save time on a simple fact, help organize a research question, or point toward a useful source. The same person may still verify a medical, financial, legal, technical, or purchasing decision on a trusted website.

The future preference question reinforces this behavior. Half of respondents prefer a hybrid search experience with a summary that still prioritizes links to human-written articles. The weighted estimate reaches 55.6 percent. Another 26 percent prefer a simple website directory without AI interference. Together, those groups represent 76 percent of the unweighted sample choosing an experience that preserves a central role for website links.

Frequent AI-summary readers show the same preference. Of the 65 respondents who always or frequently read summaries first, 64.6 percent prefer the hybrid model with links to human-written articles. Reading an AI summary does not remove the desire for source access.

Research from Gartner adds a related signal. Its 2025 survey found that 53 percent of 377 U.S. consumers distrusted AI-powered search results, and 61 percent wanted an option to turn AI summaries on or off. User control remains part of the trust equation. Searchers may accept assistance more readily when they can inspect the source, compare answers, or choose a conventional result page.

The survey found that 69 percent of respondents use AI chatbots such as ChatGPT, Gemini, or Perplexity for at least some complex research. The weighted estimate rises to 72.1 percent.

Usage intensity tells a more precise story:

The figures describe a mixed workflow. Chatbots help with brainstorming, follow-up questions, summaries, comparisons, and research planning. Traditional search remains useful for finding specific websites, checking current information, locating original documents, and evaluating sources.

For content teams, this creates two discovery requirements. A page needs to be easy for search engines to identify and useful as a source for systems that synthesize answers. That requires more than repeating target phrases. It requires stable facts, clear explanations, recognizable expertise, and enough substance for a system or reader to distinguish the page from a thin summary.

Chatbot visibility also has a measurement problem. A referral may be absent even when a brand influenced the research process. A reader may encounter a company through a chatbot, search for the company directly, and visit the website later. Standard last-click reporting will miss some of those paths.

The ContentEngine findings describe a small sample rather than the behavior of all search users. The sample includes 100 U.S. respondents between ages 18 and 62. Fifty-three percent are from the South, and the available material does not document the recruitment process, sampling frame, or weighting method in enough detail to support broad population estimates.

The results are also self-reported. The survey does not include search logs, clickstream data, historical traffic records, or objective assessments of answer quality. It cannot establish that longer queries caused greater satisfaction, that AI summaries caused fewer website visits, or that chatbot use displaced a specific amount of traditional search.

Age patterns provide useful hypotheses. Respondents ages 18 to 34 are more likely to report longer queries, at 56.7 percent, compared with 34 percent among those ages 35 to 54 and 39.1 percent among those ages 55 to 62. The youngest group shows the highest preference for the hybrid search future, while the oldest group shows the highest preference for a website directory without AI. Chatbot use also does not decline steadily with age, since 73.9 percent of the oldest group report at least occasional chatbot use for complex research.

Those differences deserve further study with larger and more representative samples. They should not be turned into simple assumptions about which age group uses AI or which group prefers traditional search.

The survey supports a practical view of search behavior. AI summaries handle quick orientation. Chatbots help with parts of complex research. Traditional results and human-written pages remain important when a person needs proof, detail, a current source, or a decision they can defend.

That environment rewards content that works at more than one level. A page can answer a short query directly, reflect the language of a longer question, provide enough evidence for an AI system to interpret, and give a reader a reason to continue past the summary.

The most durable search strategy therefore has room for both formats. Keywords help establish the subject and capture direct demand. Questions reveal context and intent. The topic determines which one should lead, and the page’s quality determines whether the searcher keeps going once the first answer appears.