The everyday AI story is less about machines taking over chores than about people adding a new first stop to ordinary tasks. They ask a chatbot for an explanation, a starting draft or an image to edit. Sometimes they ask for advice in areas where a convincing answer is not enough. Across different stakes, the tool can return something usable before the person knows whether it is reliable.
A Pew Research Center survey published in June 2026 gives that change a useful scale. Of 5,119 U.S. adults surveyed February 17–23, 49% said they use AI chatbots, and 24% said they use them daily. Another 51% said they do not use them. The country remains divided on use. Pew’s 2024 figure was 33%, but the questions changed: earlier surveys asked about ChatGPT, while the 2026 survey asked about chatbots generally. The increase is suggestive, not a clean like-for-like measure of growth.
What matters more for understanding daily life is what people say they actually do. The largest use is not running a business or making art. It is looking for information.
The new first stop for a question
Pew found that 42% of U.S. adults use AI chatbots to search for information. Question-and-answer is the most common use in its list. Someone can ask for a quick explanation of an unfamiliar term, a summary of a long passage or a starting set of questions to investigate. Current public chatbots can assist with those tasks, without guaranteeing correct answers.
The distinction between a first answer and a verified answer is easy to miss because chatbots deliver both in the same conversational form. A neat explanation may be enough to orient a reader. A claim about a policy, a deadline or a product specification needs a source that can be checked. AI may ease the first pass while shifting effort to inspecting the response. Research has not disappeared; it increasingly begins in a conversation.
Pew also found that 13% use chatbots for news. That smaller share deserves a different habit. A news summary can help someone identify a topic to follow, but the underlying report, its date and its context matter. An answer that cannot lead the reader back to reporting should not become the final version of a developing story.
The products themselves are not interchangeable, and the survey does not measure which one gives the best answer. Pew reported that 44% of U.S. adults had used ChatGPT, 24% Gemini and 17% Copilot. Those are ever-used figures, not exclusive shares of a market; one person can have tried more than one. Across brands, the more revealing behavior is using conversation as a route into information.
Work is where the draft meets an owner
Among employed adults, 38% told Pew they use AI chatbots for work tasks. That is not 38% of all adults or a measure of how much work was automated. It says the tools have entered some working routines. Drafting an email, organizing notes or summarizing material can give an employee a version to react to. The finished work still depends on whether facts, tone, obligations and missing context have been checked.
The pattern is uneven by age. In a related March 2026 Pew short-read, 38% of employed adults ages 18–29 said they used ChatGPT at work, compared with 30% of those ages 30–49 and 18% of those 50 and older. That item concerns ChatGPT specifically. It shows an age gap in use, not effectiveness or future adoption.
Microsoft’s 2026 Work Trend Index offers a view from AI-using workers rather than the general public. In that vendor study of 20,000 AI-using workers across 10 countries, 86% said they treat AI output as a starting point. The most frequently named important human skills were quality control of AI output, at 50%, and critical thinking, at 46%. Those assessments are not independent proof of productivity gains. They suggest a division of work: let a tool prepare something inspectable, then keep a person responsible for what leaves the desk.
That responsibility becomes concrete in small details. A meeting summary should be checked against what was agreed. A customer reply should be checked against actual policy. A draft analysis should be checked against its underlying figures. If the output cannot be inspected without recreating the whole task, the apparent shortcut may offer less value than it first seems.
Making something, not just finding it
Information and office work do not exhaust the picture. Pew found that 24% of U.S. adults use chatbots to create or edit images or video, while 25% use them for fun or entertainment. The creative figure points to an accessible kind of production: trying a visual concept, changing an existing image or preparing material that a person can revise. It does not tell us whether the output was shared, whether it was any good, or whether it displaced a professional’s work.
“Create” can flatten very different acts. A playful image sent to a friend, a rough mockup for a presentation and a public-facing illustration have different standards. For the first, novelty may be enough. For the last, a person may need to check that the visual represents the subject honestly and that they have the right to use any underlying material. The value of a fast draft rises when it gives someone choices; it falls when speed becomes an excuse not to inspect the result.
A single verdict on “AI tools” is inadequate. The same person might find a chatbot helpful for a low-stakes idea and unhelpful for a factual answer. Adoption tells us a tool is being tried. It does not tell us that every task benefits equally.
When an answer starts to sound like advice
The more consequential uses are already part of the survey. One in five U.S. adults—20%—said they use AI chatbots for medical advice, and the same share use them for diet or fitness information. Ten percent use them for emotional support; 4% use them for companionship. These are reported uses, not endorsements of the tools as clinicians, counselors or substitutes for relationships. The survey does not show what questions people asked or what they did with answers.
There is a meaningful difference between asking what a medical term means and acting on a suggested treatment. A chatbot can help someone prepare questions for a clinician or understand an explanation in plainer language. It can also produce an answer that sounds assured without having the full history, examination or tests a professional would need. For a consequential health decision, the practical checkpoint is a qualified professional who can evaluate the person and the evidence. The same principle applies to legal or financial answers: use a response to clarify what to ask, then verify it with an appropriate professional or primary document before acting.
Emotional support raises a different boundary. A person may find it easier to put a thought into words when a system responds immediately. That possible use should not be confused with demonstrated therapeutic effectiveness. A fluent reply is not a professional relationship, and the sensitivity of what someone might disclose makes privacy a live concern. Pew’s data establish that people are turning to these tools; they do not establish that the tools are safe or effective for every intimate need.
The AI already sitting in the room
Not every everyday encounter begins by opening a chatbot. Pew reported that about a third of U.S. adults have a smart speaker, with smaller shares having an AI doorbell or thermostat. Device ownership is not chatbot use; it shows neither frequency of AI use nor a total adoption rate that can be added to the chatbot figure.
Still, it changes the physical setting of the question. AI can be encountered through a device at home as well as through a text box on a screen. That makes the category broader than one brand or one workflow, even though the survey’s clearest evidence of behavior remains the specific uses people reported for chatbots.
The expansion also helps explain the public’s mixed response. In the same Pew survey, more adults expected AI to have a negative than a positive impact both on them personally and on society. Majorities said AI is advancing too quickly and will put personal information at risk. Those views do not cancel the usage figures. They describe the conditions under which the usage is happening: people can find a tool useful for a narrow task while worrying about where the technology is going and what happens to their information.
That is the practical shape of mainstream, contested AI. It is neither a fringe curiosity nor a settled public benefit. For a reader deciding what to use, the strongest starting point is a task with a checkable result: ask for an explanation, a draft or an option set; inspect the source material and the output; and avoid sharing sensitive details when they are not needed. As the stakes rise—from an image idea to a work commitment to medical advice—raise the standard of review with them. The useful tool is the one that leaves you better informed and still able to make the decision yourself.
