AI at Home: Has the Revolution Reached the Living Room?

About a third of American adults say they have a smart speaker in their home. Six in ten American adults have read a summary written by an AI at the top of a search results page. And roughly half of American adults now say they use an AI chatbot.

Then you get to the last number, and the picture changes. Among Americans aged 65 and older, 23% say they have ever used a chatbot. Among adults 18 to 29, it is 66%.

So has the revolution reached the living room? Yes, in the sense that the machines are in the room. No, in the sense that a large share of the people sitting in it have decided not to talk to them.

The gap between those two facts is the story, and much of it turns on a question that rarely gets asked out loud: what counts as using AI?

One organisation, three answers

The clearest illustration comes from AARP, which has surveyed Americans aged 50 and older repeatedly and published the results under different definitions.

In a data story published in January 2026, AARP reported that “eight in ten adults 50-plus report having used some form of AI,” while nearly one in four (23%) said they had never used AI.

In its 2026 Tech Trends survey, fielded September 9 to October 6, 2025, AARP reported that AI usage “[has] nearly doubled, rising from 18% in 2024 to 30% in 2025.” A note on that sample, because it does not resolve cleanly: AARP's own methodology line describes an online survey of 3,838 U.S. adults, with no age floor stated, while the report itself is titled for and reports on the 50-plus population. We therefore describe it as AARP's 50-plus survey and flag the ambiguity rather than quietly tightening a sample description the source does not make.

In the same report, asked specifically about “AI platforms such as ChatGPT or Copilot,” 19% said they used such tools, 21% were interested, and 59% expressed no interest.

Eight in ten, three in ten, two in ten. All AARP, all described as adults 50-plus, published within months of each other, though only one of the three publishes a method line and the data story publishes none at all.

The “eight in ten” figure needs its own warning attached. The data story carrying it publishes no sample size, no fieldwork dates, no mode and no question wording anywhere in its text, so its base population cannot be checked. Its own components do not even reconcile: 42% beginners, 28% intermediate, 7% advanced, 1% expert and 23% who have never used AI sum to 101%, and 100 minus 23 gives 77%, not 80%. It is a fair illustration of how broadly “AI use” can be drawn. It is not a number to put in a headline.

They are not contradictions. They are different questions. “Some form of AI” includes the voice assistant, the recommendation engine, the fraud filter, the smart doorbell. “AI tools” means something you pick up and use. “AI platforms such as ChatGPT” means a chatbot you deliberately open.

Most public argument about whether ordinary people use AI is really an argument about which of those three questions was asked.

The staircase

Pew Research Center's survey of 5,119 U.S. adults, conducted February 17–23, 2026, gives the cleanest age comparison available, because it asks the same question of everyone.

Ever use AI chatbots: 66% of adults 18–29, 61% of those 30–49, 42% of those 50–64, and 23% of those 65 and older.

The gap runs through everything Pew measured. Twenty percent of adults 65 and older have heard “a lot” about chatbots, against 66% of those under 30. Six percent of the oldest group describe themselves as extremely or very confident using them, against 31% of the youngest.

Yet the older group is not uninterested in what AI could do. Asked by AARP which applications interested them, in the same data story whose base is unpublished, adults 50-plus put translation services first at 74%, followed by home and public safety monitoring at 71%, learning and skill development at 67%, and fraud detection and financial planning at 66%. Health monitoring and smart home controls each drew 63%.

Where interest falls away is where the machine is asked to behave like a person. Emotional support drew 38%. Self-driving assistance drew 36%. Among Americans 65 and older, 38% have ever read an AI summary in a search result, against 72% of those under 30.

The pattern is less “cannot” than “will not, for this.” Older adults want the machine to translate a menu or watch a door. They are markedly less keen on it holding a conversation.

Why people say no

Pew asked people who never use chatbots why not, offered them five possible reasons, and allowed more than one.

Eighty-three per cent said they were simply not interested. Seventy-nine per cent were concerned about how their personal information would be used. Seventy-six per cent did not trust the tools to be accurate. Fifty-five per cent said they did not know how to use them.

That ordering matters, because it complicates the easiest version of this story. A knowledge gap is real — a majority of non-users say so — but it is not the leading reason. The leading reason is that they do not want to. Pew's own summary is blunt: “these non-users don't feel like they're missing out.” The sentence does not stop there: Pew continues that most are not interested “and have no plans to use chatbots soon.”

The distinction deserves care in both directions. It would be wrong to characterise older adults as incapable; it would be equally wrong to pretend the 55% who say they do not know how to use these tools are expressing a pure preference. Pew's phrasing captures both: “some opt out by choice while others aren't sure how to keep up.”

North of the border

Canada's most direct household measure comes from the Communications Security Establishment's Get Cyber Safe tracking survey, an online survey of 2,330 Canadians aged 18 and over conducted January 9–29, 2026.

Reported use of AI tools such as ChatGPT, Copilot or DALL-E rose from 32% in 2024 to 62% in 2026. Asked where they used them, 32% said both at work and at home, 24% said at home only, and 6% said at work only.

Three cautions before that 62% is repeated anywhere. The sample is online Canadians, not all Canadians, and it contains deliberate oversamples of 846 parents and 300 small-business managers. The survey's subject is cybersecurity awareness rather than technology adoption. And its AI questions concern AI tools and AI-generated content, not the whole landscape of embedded features. It measures a real thing; it is not a national adoption rate.

Its age finding is nonetheless specific and slightly counterintuitive: at-home-only use is highest among Canadians aged 65 and older. Use in both settings is more common among those under 45 and among households earning over $150,000.

Statistics Canada's numbers describe work, not home, and are included here because they show how sharply AI use tracks education. In data collected between September 2024 and July 2025, 37% of workers with a bachelor's degree or higher had used generative AI at work in the previous year, against 7% of those with a high school diploma or less. Separately, 35.9% of Canadian workers reported using generative AI tools at work over the twelve months to March 2026.

Leger's AI Index, an online survey of 1,511 Canadians aged 16 and over fielded in late May and early June 2026, put AI tool use at 79% and found 18% turning to AI first when searching for information online — a share Leger says has tripled since it began tracking in 2024.

Three Canadian numbers: 62%, 79%, and 35.9%. Same country, same year, three different questions.

Elsewhere

International comparisons are thinner and softest. A Google-commissioned Ipsos survey across 21 countries, fielded in late 2025, found 66% reporting use of an AI tool or application in the previous twelve months, and over half doing so in 20 of the 21 countries. Ipsos describes that as “up 18% from 2024” — a phrase that does not distinguish percentage points from relative percent, so it is quoted here rather than computed from.

The Ipsos AI Monitor, across 32 countries, reports that 54% say products and services using AI have “profoundly changed their daily lives” in the past three to five years, and that 24% of Americans now use AI tools “often,” up from 17% the previous autumn.

That last figure travels with a correction printed in the same section: “frequent usage of AI chatbot-like tools is flat in the U.S. What has shifted is the regularity.” The same passage notes that the share who never use AI tools fell from 26% to 17%. More people are touching these tools. They are not necessarily touching them any more often.

Two limits on the Ipsos and Google figures. The 2024 wave of the 21-country survey covered 17 countries, not 21, so the year-on-year comparison is not like for like. And the 66% is an average across the 21 country samples, not a share of the world's population.

What people say it does for them

The benefits in this data are almost entirely self-reported, which is worth stating plainly before any of them are repeated.

Among all American adults, Pew found 30% say the tools help their productivity and 5% say they hurt it — about three in five of the half of adults who use chatbots. Twenty-eight per cent say chatbots help them stay informed, against 5% who say they hurt. On creativity, 21% say help and 11% say hurt.

On the two measures where AI's effect is most contested, the largest group says there is no effect at all. On happiness, 36% say chatbots neither help nor hurt. On relationships, 35% say the same — with 6% reporting help and 7% reporting harm.

Companionship use exists and is small: 7% among adults under 30, 1% among those 65 and older.

One finding complicates the comfortable framing of AI as a convenience. In an Ipsos poll, one in three Americans said they had talked to a chatbot instead of a doctor about a medical question — four in ten among those under 55. In Pew's data, 11% of adults 65 and older say they use chatbots for medical advice.

The trust ledger

None of this has made Americans comfortable. Seventy-one per cent say increased AI use will make their personal information less secure. Sixty-three per cent say AI is advancing too quickly. Only 51% of adults 50-plus, in AARP's data, believe the benefits outweigh the risks.

And the oldest group is not the most hostile. It is the most undecided. Pew found 35% of adults 65 and older expect AI to be negative for society, against 48% of those under 30. But 21% of the older group said they were unsure, against 8% of the youngest. On AI's effect on them personally, 29% of those 65 and older were unsure.

That is a distinctive position. Younger adults have made up their minds and are more negative. Older adults have not. Three in five adults 50-plus, in AARP's survey, say technology is not designed with their age in mind — which is a statement about the product, not about the customer.

The gap did not close

Pew's own chart carries both years, and the answer turns out to be sharper than a simple yes or no.

In 2024, 55% of adults aged 18 to 29 and 12% of those 65 and older used chatbots. That is a gap of 43 points. In 2026 the figures are 66% and 23%. The gap is 43 points.

Every age group rose. The youngest and the oldest rose by exactly the same 11 points. The middle rose faster: adults 30 to 49 went from 41% to 61%, and those 50 to 64 from 25% to 42%.

So adoption spread upward through the population without changing the distance between the youngest and the oldest. Older adults are adopting at roughly the same absolute rate as young adults, from a much lower base, which raises their share quickly and leaves the gap exactly where it was.

Two cautions apply to that arithmetic. The 2024 question was asked only of people who had heard at least a little about chatbots, so the starting figures are probably somewhat lower than a like-for-like measure would produce; Pew marks the series with a dotted line for precisely this reason. And the 2026 question widened from ChatGPT alone to chatbots generally. The direction is sound. The size of each individual jump is not.

For comparison, AARP's separate measure of AI usage among adults 50-plus rose from 18% in 2024 to 30% in 2025 — a different question, a different population and a different result.

What this article cannot tell you

It cannot tell you whether household AI use is good for anyone. No figure here measures an outcome; the benefits are what users report about themselves.

It cannot tell you the American income or education split, because that breakdown was not retrieved.

It cannot tell you how often Canadians use AI, only where they use it.

And it cannot tell you that a generation has been left behind. What it can tell you is that non-users, by a wide margin, say they are simply not interested, and that Pew does not break that reason out by age in the material retrieved, so it cannot be attached to older adults specifically.

What it can tell you is narrower and more useful: the machines are already in the house. Whether they are part of the household depends entirely on which question you ask. And the group most often described as unable to answer is, on this evidence, better described as opting out than as being excluded, though the survey cannot tell us which older adults are which.