10% of People Already Start Complex Research With Chatbots According to Recent Survey

A September 2026 survey suggests that chatbots have become part of mainstream research habits, although only about one in ten respondents starts complex research with an AI chatbot.

The headline rounds the survey’s 9 percent figure to 10 percent. That distinction matters because the same survey found that 69 percent of respondents use chatbots at least occasionally for complex research. Most of those people still keep Google and other traditional search engines in the workflow. The emerging habit looks like a two-step process: AI helps people get oriented quickly, and websites provide detail, evidence, and confidence before a decision.

The ContentEngine survey collected responses from 100 people in the United States on September 14, 2026. Respondents were asked whether they had shifted complex research toward tools such as ChatGPT, Gemini, or Perplexity.

The answers broke down into four groups:

The first three groups together account for 69 percent of the sample using a chatbot at least sometimes. The 9 percent chatbot-first group shows the clearest change in research behavior, while the 40 percent who split their use evenly represent the larger shift in practical terms. AI has become part of the research path for many people without becoming the sole entry point.

That pattern gives marketers a more useful planning signal than the raw adoption figure. A brand that optimizes only for chatbot visibility may miss the majority of users who begin with a search engine, compare several sources, or return to a website after asking an AI tool for help.

A September 2026 ContentEngine survey of 100 U.S. respondents suggests that AI has entered mainstream research habits without replacing traditional search. Sixty-nine percent use chatbots at least occasionally for complex research, but only 9% make a chatbot their primary starting point. Most users follow a hybrid path: they use AI summaries or chatbots to understand a topic quickly, then consult websites for evidence, context, and confidence. The findings point to a tougher environment for casual search traffic and a stronger need for content that is clear enough to summarize, authoritative enough to trust, and useful enough to reward a visit.

The survey found similar behavior around Google’s AI-generated summaries. Sixty-five percent of respondents said they always or frequently read an AI summary before reviewing traditional search links. Twenty-six percent said the summary was the first thing they looked at. Another 39 percent used it to decide whether they needed to scroll down.

The summary often works as a filter. It helps a person decide whether a search result deserves more time, whether the topic is simple enough to resolve immediately, or whether additional evidence is needed.

When respondents were asked how often an AI Overview removes the need to visit a website, the largest group chose a middle position:

The 44 percent figure is especially important for publishers and marketers. These users are still potential website visitors. They arrive after seeing a compressed version of the topic, so the page has to offer something that the summary cannot provide in a few lines. That could include original data, a complete explanation, product specifications, expert interpretation, examples, a calculator, a process guide, or a clear answer to a follow-up question.

A page that repeats the summary word for word gives the visitor little reason to stay. A page that supplies the supporting evidence and practical detail has a stronger chance of earning attention after the initial AI exposure.

Self-reported behavior from the ContentEngine survey fits the broader picture in some areas and leaves important questions open. Pew Research Center analyzed browsing data from 900 U.S. adults and found that users were less likely to click result links when an AI summary appeared on a Google results page.

A separate study based on the same type of browsing environment reported that clicks to sources cited in AI Overviews occurred in about 1 percent of visits to pages containing an AI Overview. That result points to a difficult reality for publishers: being cited inside an AI response does not guarantee a visit.

Google has presented a more favorable view of the traffic picture. The company says average click quality has increased and that it sends slightly more quality clicks to websites than a year earlier. Google defines a quality click as one where a visitor does not quickly return to the results page. It has also said that AI Overviews can expose users to more links on a search page.

These findings measure different things:

Together, they suggest that search traffic is becoming more selective. A website may receive fewer casual visits while attracting visitors with a clearer reason to continue reading. That shift affects how publishers interpret traffic reports. Page views alone may offer less insight than engaged sessions, return visits, newsletter signups, downloads, conversions, or other actions that show genuine interest.

The most important finding is not that 9% of respondents start complex research with a chatbot; it is that search behavior is becoming layered. AI can win the first moment of attention, but websites still control much of the information people need before acting. Publishers and brands should therefore stop treating AI visibility and organic traffic as interchangeable goals. The durable advantage will belong to pages that provide a clear answer quickly, then add evidence, specificity, and practical value that a compressed summary cannot reproduce.

AI chatbot adoption does not equal automatic confidence. In the ContentEngine sample, 40 percent said they trusted AI-generated search responses about the same as a standard website. Thirty-five percent expressed skepticism or no trust, while 25 percent trusted AI responses more than standard websites or trusted them completely.

The relationship between trust and usage was uneven. Among the 25 respondents who expressed higher or complete trust, 24 said they always or frequently read AI summaries first. Among the 35 skeptical respondents, 11 still read summaries first regularly.

That second figure reveals an everyday behavior that marketers should take seriously. People can use a tool for speed while holding reservations about its accuracy. A person may ask a chatbot for an overview, then visit a known website to check dates, sources, safety details, prices, legal language, or product limitations.

Research from Gartner adds another signal. A 2025 survey of 377 U.S. consumers found that 53 percent distrusted AI-powered search results, and 61 percent wanted the option to turn AI summaries on or off. The Reuters Institute’s 2026 research indicates that trust in several dedicated AI systems has increased in some markets, although trust remains connected to the specific tool, use case, and source context.

For a website, credibility has several visible components:

A search summary can introduce a brand. The website still has to support the claims that the summary presents.

The survey found variation across age groups, although the sample is too small for broad population claims. Among respondents aged 18 to 34, 56.7 percent reported longer queries over the previous 18 months. The figure was 34 percent among those aged 35 to 54 and 39.1 percent among those aged 55 to 62.

The youngest group showed the highest preference for the hybrid future at 63.3 percent. The oldest group showed the highest preference for a website directory without AI at 39.1 percent.

Chatbot use did not follow a simple age pattern. Among respondents aged 55 to 62, 73.9 percent reported at least occasional chatbot use for complex research. The figure was 57.4 percent for the middle age group.

For marketers, these results argue against assigning AI behavior to a single age segment. Different audiences may use the same tools for different reasons. A younger person may use a chatbot to explore a new topic, while an older person may use one to summarize technical information before checking a familiar source. Audience research remains more reliable than assumptions based on age alone.

The survey asked respondents to choose their preferred future search experience. Half selected a hybrid model with an AI summary that still prioritizes links to human-written articles. Twenty-six percent preferred a simple website directory without AI interference. Fifteen percent wanted a personal assistant that could take actions such as booking a flight. Nine percent preferred a computer that provided one perfect answer.

The weighted estimate placed the hybrid preference at 55.6 percent. When the hybrid and website-directory options are combined, 76 percent of respondents chose an experience that preserved a central role for website links. The weighted estimate for that combined group was 77.9 percent.

Frequent AI-summary readers showed strong support for the hybrid model. Among the 65 respondents who always or frequently read AI summaries first, 42, or 64.6 percent, preferred a summary paired with links to human-written articles. Among the 35 people who rarely or never read summaries first, 8, or 22.9 percent, preferred that hybrid option.

The preference suggests that regular AI users still value source access. They may want a quick explanation at the top of the search experience, followed by the ability to inspect the article, compare viewpoints, and understand the evidence.

That expectation affects how content should be built. A source page has to work in two settings:

A page can support the first setting with clear headings, concise definitions, direct answers, structured data, and well-labeled sections. It can support the second with original reporting, useful tools, detailed examples, and information that requires context.

The ContentEngine survey offers a useful snapshot, with several limits that shape how the results should be used. It included 100 U.S. respondents, ages 18 to 62. Fifty-three percent of the sample came from the South. The available information does not document the recruitment process, sampling frame, or weighting methodology in enough detail to support population-level conclusions.

The results also rely on self-reported behavior at one point in time. Respondents may remember their search habits imperfectly, and stated behavior may differ from browsing logs. The survey does not measure actual traffic changes, objective answer quality, or the causal effect of AI features on clicks.

Some response options also combine several ideas. For example, the hybrid future option includes an AI summary, links, and human authorship in one choice. Those features may appeal to respondents for different reasons. The results show a direction of preference, while further research would need to isolate each factor.

The survey is best read alongside behavioral research from Pew, platform reporting from Google, and broader trust research from Gartner and the Reuters Institute. Those sources measure different populations and behaviors, so their figures should not be treated as directly interchangeable.

The 9 percent chatbot-first figure matters because it shows where the market is heading. The 69 percent overall chatbot-use figure matters because it shows how widely the tool has already entered research. The difference between those numbers describes the current search environment more accurately than either number alone.

Most respondents in this sample still move between systems. They may ask a chatbot to define a topic, use Google to find a specialist source, scan an AI summary, and then open an article to check the details. Their path can include several short decisions instead of one direct click from a results page to a website.

That behavior gives websites several jobs:

The strongest content will account for all five. It will be easy to summarize and worthwhile to read in full. It will use clear language without flattening important qualifications. It will target recognizable search terms while addressing the longer questions that appear in modern research sessions.

Search is becoming a layered habit. AI handles part of the orientation process, while human-written pages continue to carry much of the burden for detail, verification, and judgment. The survey’s headline number captures an early group of chatbot-first researchers. The larger story is the growth of a mixed workflow in which users expect speed and source access from the same search experience.