How People Search for Information Now: Google, AI, Social

For two decades, "searching" meant one thing: type a few words into Google, scan ten blue links, click one. That habit is being replaced. More and more people now type a full question into a chatbot and read one answer.

This article sets out what the 2026 data says about that shift, where the numbers come from, and what it means for anyone who wants to find reliable information quickly.

The Short Version

AI Chatbots Are Replacing Google for Questions

The clearest sign is in how often people still search. Clickstream data from Datos and SparkToro showed Google desktop searches per user in the United States falling nearly 20% in a year, even though the number of people using Google held steady.^[21] People have not left Google. They are bringing it fewer questions.

Those questions are going to chatbots. In a Pew survey in February 2026, 44% of US adults said they use ChatGPT, up from 34% a year earlier and 18% in 2023.^[4] OpenAI said the same month that ChatGPT had reached 900 million weekly active users.^[5]

The technical reason is retrieval-augmented generation, usually shortened to RAG. A modern chatbot does not answer only from what it learned in training. It runs a web search in the background, reads the pages it finds, and writes one answer with its sources attached. The reader no longer has to open five results and piece the answer together. One question replaces several searches and several clicks, which is why it feels faster.

Google has responded by becoming a chatbot itself. AI-written answers now appear on more than 20% of its searches, and Google said in 2026 that its chat-style AI Mode had passed 1 billion monthly users.^[2] Whichever product wins, the list of ten blue links is the thing being replaced.

What Google Still Holds

Replacement is not complete, and by raw volume Google is still far ahead. In research published in March 2026, SparkToro and Datos counted searches on 41 major websites, including search engines, shops, social networks, and AI tools. Google alone accounted for 73.7% of all desktop searches across those sites in the United States. Traditional search engines together took about 80%, commerce sites such as Amazon about 10%, social networks about 5.5%, and AI tools 3.2%.^[1]

Two things explain the gap between that figure and everyday experience. The count covers desktop browsers only, so it misses the phone apps where much chatbot use happens. And habits differ sharply by age: 61% of Americans aged 18 to 29 use ChatGPT, against 19% of those 65 and older.^[4] Navigation, local lookups, and quick repeat searches still go to Google; questions that need an explanation increasingly do not.

Bing, Yahoo, Yandex and the Rest

Google is not the only search engine, and in some countries it is not the first choice. Statcounter, which measures which search engine sends visitors to a large sample of websites, gives this picture for October 2026.^[13]

The worldwide average hides large national differences:

• Bing is the clear second engine in the United States, with close to one search referral in ten, about double its worldwide share.
• Yandex leads in Russia with 70.15% against Google's 28.43%, and holds 23.74% in Turkey, where Google has 74.79%.
• Baidu leads in China with 49.08%, with Bing second at 21.66%. Google accounts for under 7% there.
• Yahoo and DuckDuckGo are small everywhere but steady, and each is larger in the United States than worldwide.

These figures count visits that search engines send to websites, not searches typed, so they say more about which engine people click through from than about how often they search. They still make one point clearly: for most of the world "search engine" means Google, and where it does not, the alternative is a local leader such as Yandex or Baidu, not a newcomer.

Most Searches Now End on the Results Page

A search used to be a doorway to another website. Increasingly it is the destination. Using Similarweb panel data for January to April 2026, SparkToro found that 68.01% of Google searches in the United States ended without a click to any result, up from 60.45% in 2024. A decade ago the figure was about 45%.^[2]

The same study points to AI-written summaries as the main cause of the recent jump. Pew Research Center measured their effect directly by analyzing the browsing of 900 US adults across 68,879 Google searches in March 2025. Users clicked a traditional result on 8% of visits when an AI summary appeared, compared with 15% when it did not, and clicked a link inside the summary itself on just 1% of visits.^[3]

The reader gets an answer faster. The cost is that the answer arrives with less visible context: who said it, when, and on what evidence. That context used to come free with the click.

What People Ask Chatbots

What people do with a chatbot is closer to consulting than to looking something up. A study by OpenAI researchers and Harvard economist David Deming, based on 1.5 million conversations, found that around 70% of consumer use was not work-related and that about 49% of messages were "asking": seeking information or advice, as opposed to having the tool produce something.^[6]

The questions are the everyday ones: what is this, how do I do that, where can I find it, which of these is better. In the same study, the largest single category of use was what the authors call practical guidance, at roughly 29% of all usage. Within it, general how-to advice made up 8.5% of all messages and tutoring or teaching another 10.2%. Plain fact-finding, which the authors call seeking information, grew from 14% of usage in July 2024 to 24% a year later.^[18]

For news specifically, the habit is growing from a small base. The Reuters Institute's Digital News Report 2026, covering 48 markets, found that 10% of people use AI chatbots for news, up from 7% the year before, and 16% of those under 35.^[7] For a comparison of the main assistants, see ChatGPT vs Claude vs Gemini.

Shopping: Asking Instead of Browsing

Shopping questions are following the same path. Product queries are common enough that the ChatGPT study tracks them as their own category.^[18] Retailers see the result on their side: Adobe reported that visits to US retail sites arriving from generative AI tools grew 393% in the first quarter of 2026 compared with a year earlier.^[20] The pattern is to ask an assistant to find sellers, compare options, or narrow a choice, and then click through to buy.

Social Video: Entertainment First, Information by Accident

The first public sign that younger people search differently came from Google itself. In July 2022, senior vice president Prabhakar Raghavan said that in the company's own studies, almost 40% of young people looking for a place for lunch did not go to Google Maps or Search: they went to TikTok or Instagram.^[8]

That does not make these apps search tools in the way a search engine or a chatbot is. People open them to be entertained. Pew found that 95% of adults who use TikTok go there for entertainment, and 81% call it a major reason; 62% say they also use it to look at product reviews or recommendations.^[19] Information reaches people there mostly as a by-product of scrolling, not as the answer to a question they typed.

News is the clearest case. In a Pew survey of 5,166 US adults in August 2026, one in five said they regularly get news on TikTok. Among adults aged 18 to 29 the share was 39%, compared with 25% of those aged 30 to 49 and 4% of those 65 and older.^[9]

Worldwide the pattern is the same. The Reuters Institute reported that in 2026, for the first time, social media and video networks were the most widely used way of getting news online across its 48 markets, used by 54% of respondents.^[7]

Forums: Popular, but Not a Source

Another habit is the search for an unpolished human opinion. Many people now add "reddit" to a query to skip pages written to rank and find a thread where someone describes what actually happened when they bought the product or tried the method.

Reddit's own figures show the scale. In the third quarter of 2026 the company reported 130.3 million daily active users, up 12% on the year, and 515.2 million weekly users, up 16%.^[10] Popularity is not reliability.

Posts are anonymous, no organization stands behind them, and nobody checks them before or after they are published. The most upvoted answer is the most popular one, not the most accurate. A thread can show what some people say they experienced; it cannot serve as a source for a fact.

The record shows how this goes wrong:

• Wrongly naming a suspect (2013). After the Boston Marathon bombing, Reddit users trying to identify the bombers named a missing student, Sunil Tripathi, who had nothing to do with it. Reddit's general manager publicly apologized for what the company itself called an online witch hunt.^[14]
• Health misinformation (2021). Reddit banned one large Covid-denial community and restricted 54 others, but only after more than 100 other communities shut their doors in protest at the company's inaction.^[15]
• A joke that became an answer (2024). Google's AI Overviews told users to add glue to pizza sauce. Reporters traced the advice to a joke comment posted on Reddit about eleven years earlier, which shows how forum content travels into AI answers.^[16]
• Fake people (2025). Researchers from the University of Zurich secretly ran AI accounts in a Reddit debate forum for four months. The accounts posted 1,783 comments posing as real people, including a trauma counselor, before anyone found out. Reddit's chief legal officer called the experiment improper and highly unethical.^[17]

The lesson is not that every post is false. It is that a reader has no way to tell from the post itself.

The Accuracy Problem

Convenience has moved faster than reliability, and the public has noticed. In the Reuters Institute's 2026 survey, only 20% of people said they trust answers from AI chatbots, compared with 37% who trust news overall.^[7]

Testing supports the caution. In March 2025, researchers at Columbia University's Tow Center for Digital Journalism tested eight AI search tools with 1,600 queries, asking each to identify the source of an excerpt from a real news article. The tools gave incorrect answers more than 60% of the time. The best performer was wrong on 37% of queries and the worst on 94%, and the tools rarely signalled any doubt about a wrong answer.^[11]

That test was deliberately narrow, and it does not mean six in ten chatbot answers on any topic are wrong. It does show that a confident tone and a visible citation are not evidence that the citation says what the answer claims.

How Easy AI Search Is to Manipulate

Speed has a second cost. A system that reads a few web pages and repeats what they say can be steered by whoever writes those pages.

• One blog post, twenty minutes. In 2026 a BBC technology journalist published a single made-up post on his own website claiming he was a champion competitive hot dog eater. Within about a day, ChatGPT, Google's Gemini, and Google's AI Overviews were presenting the claim as fact. The work took about 20 minutes and no technical skill.^[22]
• Hidden instructions. Academic researchers have shown that text planted in a web page, invisible to a human visitor, can make an AI search tool favour that page's product over its competitors.^[23]
• Legal responsibility. In June 2026 the Regional Court of Munich ruled that Google's AI Overviews are Google's own statements, after they generated false claims that two publishers had committed fraud. The court barred Google from repeating them.^[24]
• A fine for search conduct. In 2026 the European Union fined Google 460 million euros over its conduct in search, in a decision that did not yet cover its AI features.^[25]

The hot dog test worked partly because nobody else had written about the subject, so there was nothing to contradict it. That is exactly the situation with a small company, a new product, or a little-known person, which is where a reader most needs the answer to be right.

When the Chatbot Sends You to a Scam

The most direct harm comes when an assistant recommends a website, because a fake shop can look entirely legitimate.

• Fake stores in shopping answers. A study by the brand-protection firm Podqi, reported in November 2026, put shopping questions about 151 brands to ChatGPT, Gemini, and Perplexity and collected 6,795 answers. About 9% of them exposed the shopper to a risky site: 15% of Perplexity's answers, 8% of ChatGPT's, and 4% of Gemini's. Of the sites flagged, 351 were outright fakes.^[26]
• Wrong login addresses. The security firm Netcraft asked a GPT-4.1 model where to log in to 50 well-known brands. Of the 131 web addresses it gave, 34% did not belong to the brand at all. Most of those were unregistered or inactive, which means a scammer could register them and wait.^[27]

The study of shopping answers measured exposure to risky sites, not how many people lost money.

How to Search Well Now

The useful question is no longer "which search tool is best" but "which tool fits this question."

Four habits carry across all of them:

• Open the source. When an AI summary or chatbot cites a page, open it and find the sentence. This takes under a minute and catches most errors.
• Read laterally. Before trusting an unfamiliar site, leave it and see what others say about it. The method is covered in How to Spot Misinformation.
• Check the address before you pay. If an assistant recommends a shop or a login page, confirm the web address on the company's official site or type it yourself. Fake shops are built to look real.
• Ask a better question. A chatbot answers what was asked. Giving it the situation, the constraint, and what you already know produces a narrower and more checkable answer.

What It Means for Anyone Who Publishes

For writers, businesses, and site owners, the numbers point one way. SparkToro and Similarweb reported in November 2026 that visits to the top million websites were 12.2% below their 2022 peak, even though the number of people online had grown by about 24%, a fall of 29.2% in visits per user.^[12]

Being the source of an answer now matters more than being the destination of a click. Summaries and assistants draw on pages that state things plainly, attribute them, and keep them current. The mechanics of how pages are found and ranked are covered in How Search Engines Work; the difference today is that the reader may meet the page's content without ever seeing the page.

The Bottom Line

The question has moved from the search box to the chatbot, and Google is following it there. Every place people now ask fails in its own way: the AI answer is fast but easy to steer, the results page hides its sources behind a summary, the video is built to entertain, and the forum is anonymous and unchecked. Knowing which tool fits which question, and taking one minute to open the source, is now the core of searching well.

Sources & Further Reading

1. Fishkin, R. (2026). "New Research: Search Happens Everywhere; an Analysis of 41 Websites with Significant Search Activity." SparkToro. View source
2. SparkToro. (2026). "In 2026, Less than One Third of Google Searches Still Send a Click." View source
3. Pew Research Center. (2025). "Google users are less likely to click on links when an AI summary appears in the results." View source
4. Pew Research Center. (2026). "Americans and AI 2026: Chatbots, Smart Devices and Views on Impact." View source
5. TechCrunch. (2026). "ChatGPT reaches 900M weekly active users." View source
6. OpenAI. (2025). "How people are using ChatGPT." View source
7. Reuters Institute for the Study of Journalism. (2026). Digital News Report 2026: Overview and key findings. View source
8. NBC News. (2022). Report on remarks by Google senior vice president Prabhakar Raghavan at Fortune Brainstorm Tech. View source
9. Pew Research Center. (2026). "1 in 5 Americans Regularly Get News on TikTok." View source
10. Reddit, Inc. (2026). "Reddit Reports Third Quarter 2026 Results." View source
11. Jazwinska, K., and Chandrasekar, A. (2025). "AI Search Has A Citation Problem." Columbia Journalism Review, Tow Center for Digital Journalism. View source
12. SparkToro. (2026). "New Research: Traffic to the Open Web Has Fallen 12% in 5 years; Visits/web User Are Down 30%." View source
13. Statcounter Global Stats. (2026). Search Engine Market Share, October 2026: Worldwide, United States, Russian Federation, Turkey and China. View source
14. NBC News. (2013). "Reddit publicly apologizes 'for the pain' caused to family of falsely accused student." View source
15. Engadget. (2021). Report on Reddit banning and quarantining COVID-19 misinformation communities. View source
16. Fortune. (2024). Report on Google AI Overviews advising users to add glue to pizza. View source
17. 404 Media. (2025). "Researchers Secretly Ran a Massive, Unauthorized AI Persuasion Experiment on Reddit Users." View source
18. Chatterji, A., Cunningham, T., Deming, D. J., et al. (2025). "How People Use ChatGPT." NBER Working Paper 34255. View source
19. Pew Research Center. (2024). "How TikTok users view, experience the platform." View source
20. Yahoo Finance. (2026). "AI Traffic to US Retailers Jumps 393% in Q1 as Agentic Shoppers Outspend Humans." View source
21. Search Engine Land. (2026). "Google searches per U.S. user fell nearly 20% YoY: Report." View source
22. Digital Trends. (2026). "AI chatbots are lying to you, and it was embarrassingly easy to make them do it." Report on a BBC investigation by Thomas Germain. View source
23. Nestaas, F., Debenedetti, E., and Tramèr, F. (2024). "Adversarial Search Engine Optimization for Large Language Models." arXiv:2406.18382. View source
24. Silicon Republic. (2026). "German ruling holds Google liable for AI Overview results." View source
25. PPC Land. (2026). Report on the European Union's 460 million euro search fine against Google. View source
26. Startup Fortune. (2026). "Podqi study finds ChatGPT, Gemini and Perplexity send shoppers to fake stores." View source
27. Netcraft. (2025). "LLMs Are Recommending Phishing Sites: Here's Why That's Dangerous." View source