Voice Search Statistics 2026: The Numbers That Are Real and the Ones That Are Not

Voice search is the strangest topic in SEO. Everyone has an opinion, almost nobody has data, and the single most quoted figure in the field was never a research finding at all.

That matters more in 2026 than it did five years ago, because voice assistants have genuinely changed. Alexa, Siri and Google Assistant have all been rebuilt or replaced on top of large language models in the last eighteen months, and the thing sitting in 39% of American living rooms is no longer the keyword-matching machine that SEO advice was written for.

So this article does two jobs. It gives you the numbers that are real and verifiable, and it is blunt about the ones that are not. There is less solid voice data now than there was in 2019, and pretending otherwise is how the industry ended up citing a 2014 offhand remark as a 2020 statistic.

Key voice search statistics for 2026

• 39% of Americans aged 12 and over own a smart speaker, an estimated 112 million people, per Edison Research's Infinite Dial 2026 [1] [12]
• US ownership was flat for four years before that: 35% in 2022, 36% in 2023, 34% in 2024, 35% in 2025, per Edison Research [1]
• 45% of UK adults aged 16 and over own at least one smart speaker, an estimated 22 million people, up from 25% in 2021, per Edison Research's Infinite Dial UK 2025 [2]
• In the UK, 37% own an Alexa device, 16% a Google Home and 5% an Apple HomePod, per Edison Research [2]
• There are "over 600 million Alexa devices out there today", per Amazon [3]
• Over 350 million customers used Alexa for Shopping in the 12 months to June 2026, interactions up more than 5x year on year, per Amazon CEO Andy Jassy [4]
• Apple's next generation foundation models are "based on Google's Gemini models and cloud technology", per a joint statement from Google and Apple, January 2026 [5]
• Gemini replaced Google Assistant in "over 250 million cars" running Android Auto, in 45 languages, per Google, November 2025 [8]
• 41% of US adults who drove or rode in a car in the past month have CarPlay or Android Auto, and 34% actively use it, per Edison Research [1]
• The last public share-of-search figure Google gave was "1 in 5 searches on mobile app in USA are voice searches", from Google I/O 2016, cited in Kleiner Perkins' Internet Trends 2016 [9]
• The "50% of searches will be voice by 2020" line traces to Baidu's Andrew Ng in September 2014, who said at least 50% of searches would be "either through images or speech", per the same Kleiner Perkins deck [9]
• Ofcom's Media Nations 2026 UK report contains no smart speaker penetration figure at all [11]

How many people own a smart speaker in 2026?

Just under four in ten Americans and just under half of British adults. Edison Research puts US ownership at 39% of the population aged 12 and over in 2026, an estimated 112 million people, and UK ownership at 45% of adults aged 16 and over in its most recent UK wave [1] [2].

The trend line is the interesting part. US ownership climbed from 7% in 2017 to 33% in 2021, then stopped: 35%, 36%, 34%, 35% across 2022 to 2025, before moving to 39% in 2026 [1].

Source: Edison Research, The Infinite Dial 2026 [1]

That four-year plateau is the most honest thing in this article. Smart speakers hit saturation among the people who wanted one, sat still through the entire period when voice search was supposedly taking over, and only moved again in 2026 once the assistants inside them got interesting.

Note also that the 2024 reading is lower than 2023. Survey noise, almost certainly, given the sample of 2,050. The honest resolution of this data is roughly plus or minus two points, so do not build a strategy on a one-year wobble.

Do more people own smart speakers in the UK or the US?

The UK, and by a wide margin. Edison Research's directly comparable 2025 figures put UK ownership at 45% of adults aged 16 and over against 35% in the US on the same age base [2].

That gap surprises people who assume American tech adoption leads. It mostly reflects Amazon's pricing aggression in the UK and the country's unusually strong radio habit, which gave smart speakers an obvious everyday job.

Source: Edison Research, The Infinite Dial UK 2025 [2]

Within the UK, Amazon dominates. Alexa devices are in 37% of homes, Google Home in 16% and Apple HomePod in 5%, with each brand roughly doubling since 2021 [2].

For a UK business, Alexa is the only voice platform with meaningful reach, and Alexa is now an Amazon shopping surface first and a general assistant second. That shapes what "optimising for voice" can realistically mean here.

What share of searches are done by voice in 2026?

Nobody credible knows. No search engine publishes it, no panel provider measures it reliably, and every percentage circulating in 2026 traces back to survey self-report or to numbers from the mid-2010s.

The last figure Google itself put into the public record was at Google I/O in May 2016: one in five searches on its mobile app in the USA were voice searches. That appears on slide 125 of Kleiner Perkins' Internet Trends 2016 alongside Bing reporting 25% of Windows 10 taskbar searches as voice, Baidu at one in ten queries, and Siri handling more than a billion requests per week as of June 2015 [9].

Ten years on, none of those companies have updated the number. That silence is data of a kind. Firms publish metrics that flatter them.

The measurement problem is also genuinely hard now. When you speak to Alexa+ or Gemini and get a synthesised answer with no SERP behind it, it is not obvious that a "search" happened at all, and there is no query log a third party can sample.

So when you see a precise voice-share percentage in a 2026 listicle, check the trail. It almost always ends at a survey question about whether someone has ever used a voice assistant.

Where did "50% of searches will be voice by 2020" come from?

From a conference remark by Andrew Ng, then chief scientist at Baidu, in September 2014. He said that in five years' time at least 50% of all searches would be "either through images or speech" [9].

That quote sits on slide 125 of Mary Meeker's Internet Trends 2016 deck, in a panel titled "Voice = Gaining Search Share". The slide's source line lists nine sources for six separate claims, including Baidu World 2014, Google I/O 2016, ComScore, Recode and Fast Company [9].

Three things then happened to the quote. "Images or speech" became "voice". "Searches" became "Google searches". And ComScore, one of nine sources on a crowded slide, got the attribution.

It was never a study. It was a prediction about a different set of technologies, in a different market, made by someone at a company that mainly served Mandarin speakers typing on small keyboards. Ng's five years ran to 2019, and neither voice nor image search came close to half of queries in any Western market on any published measure.

The useful lesson is not that voice does not matter. It is that a single unsourced number, repeated often enough, redirected a decade of content strategy budget. Ask for the primary source before you plan around a figure.

How have voice assistants changed now they run on LLMs?

Completely, and in under two years. All three major assistants have been rebuilt on or replaced by large language models, which changes what a voice query is and what it returns.

Amazon launched Alexa+ in February 2025, and by July 2026 it was available to all customers in the US and Canada, with early access in the UK, Brazil, Mexico, Italy, Spain, Germany, Austria and France. It costs $19.99 per month, or nothing for Prime members [3].

Google began upgrading mobile users from Google Assistant to Gemini in March 2025, extending the swap to tablets, cars, headphones, watches, speakers, displays and televisions [7].

Apple took the longest route. In January 2026, Google and Apple issued a joint statement confirming that "the next generation of Apple Foundation Models will be based on Google's Gemini models and cloud technology" [5]. When Apple announced Siri AI at WWDC in June 2026, the press release described Apple Foundation Models running on device and on Private Cloud Compute, and did not mention Gemini once [6].

The gap between those two announcements is worth sitting with. Siri is now substantially powered by Apple's largest search competitor, and neither party is keen to say so loudly.

For SEO, the consequence is concrete. Voice answers are no longer read out from a featured snippet you can win with schema and a 29-word paragraph. They are generated, which puts voice inside the same visibility problem as ChatGPT and AI Overviews rather than beside it.

Is anyone actually buying things by voice?

Amazon reports very large engagement numbers and almost no voice-specific ones. That distinction is the whole story of voice commerce.

On the Q2 2026 earnings call, Andy Jassy said over 350 million customers had used Alexa for Shopping in the last 12 months, that active users nearly doubled in the quarter, and that interactions were up over 5x year on year [4].

Read that carefully. Alexa for Shopping is the merged product that absorbed Rufus, Amazon's text-based shopping chatbot, so the vast majority of those 350 million interactions are typed, not spoken. It will be quoted as a voice statistic anyway. It is not one.

Amazon does not disclose voice-initiated sales, and never has. Neither does Google, Apple or any major retailer. There is no published, methodologically sound figure for voice commerce value in the UK or the US in 2026, which is why every number you find comes from a market-sizing firm extrapolating from assumptions.

The historical record is unkind. In February 2018, OC&C Strategy Consultants forecast that UK voice shopping would grow from £0.2bn in 2017 to £3.5bn by 2022, based on a survey of 1,500 smart speaker owners, and predicted that almost half of UK households would own a smart speaker by 2022, up from 10% [10].

The device forecast was close but slow: UK ownership reached 45% in 2025, three years late [2]. The £3.5bn commerce forecast has never been shown to have landed, and no credible source has published a UK voice commerce figure anywhere near it.

Is in-car the real voice use case?

It is the most defensible one, because the hands-free constraint is real rather than a preference. It is also where the platform shift is furthest along.

Edison Research found that 41% of US adults who drove or rode in a car in the past month have CarPlay or Android Auto in their primary vehicle in 2026, up from 32% in 2024, and that 34% actively use it. Among those who have it, 83% use it [1].

Sources: Edison Research, The Infinite Dial 2026 [1] and The Infinite Dial UK 2025 [2]. Base: adults 18+ who drove or rode in a car in the past month.

Google confirmed in November 2025 that Gemini was replacing Assistant across Android Auto, which it said was "available in over 250 million cars on the road", rolling out globally in 45 languages [8].

The in-car number that matters for marketers is not ownership, it is what people ask for. Navigation, music and messaging dominate. A local business can reasonably care about being findable by voice in a car. A B2B SaaS company cannot.

Why has the data got worse, not better?

Because the organisations that used to measure voice stopped, and nothing replaced them. This is the least discussed fact in the category.

NPR and Edison Research ran the Smart Audio Report, the best public voice tracker anywhere, and the last edition was published in 2022. Ofcom's Media Nations 2026 UK report mentions voice assistants exactly once, in a passing note about smart TV navigation, and gives no penetration figure at all [11].

When a regulator drops a metric, that usually means it stopped changing enough to be interesting. It is not evidence that voice died, but that voice settled into a boring, stable utility, which is precisely the sort of thing researchers stop tracking.

The practical effect is that the industry's voice statistics have calcified. Listicles recycle 2019 figures because there are no 2026 figures to replace them with, and each recycling drops a caveat.

How to read these numbers

Edison Research's Infinite Dial is a survey, not server-side measurement. The 2026 US wave was 2,050 people aged 12 and over, fielded in January 2026, half via the probability-based SSRS Opinion Panel and half non-probability, weighted to national population figures [1]. The UK 2025 wave was 1,313 people aged 16 and over, fielded in March 2025 [2].

At those sample sizes, year-on-year moves of two or three points are within noise.

Ownership is not usage. Owning a smart speaker tells you nothing about whether someone asks it questions.

Company-reported figures are marketing. "Over 600 million Alexa devices" is a cumulative device count, not active users, and Amazon has not published an Alexa active user number in years [3]. The 350 million Alexa for Shopping customers figure covers a product that is mostly text [4].

Anything described as market size for voice commerce is a model, not a measurement. If a report cannot tell you what it observed and how, it observed nothing.

Finally, be suspicious of any voice statistic without a year attached. If the underlying research is from 2019, the assistant it describes no longer exists in that form.

What this means going into 2027

Voice is now a delivery format for AI answers, not a separate search channel. Whatever you do to become citable inside ChatGPT, Gemini and AI Overviews is the same work that makes you retrievable by an LLM-powered assistant. There is no separate voice SEO discipline left to build.

Expect the share-of-search question to become permanently unanswerable. When Alexa+ or Siri AI synthesises a response from multiple sources, there is no query and no click for anyone to count.

Watch the Alexa+ rollout in the UK. It is still early access here as of July 2026, and full availability would put an LLM assistant into the 37% of British homes with Alexa hardware [2] [3]. That is the largest potential shift in UK voice behaviour on the table.

Assume commerce claims stay unverifiable. Amazon merged its voice and text shopping assistants into one reported number, so voice-attributable sales are now structurally invisible.

And treat the next big voice statistic the way you should have treated the last one. Ask who ran the study, what they asked, how many people they asked, and whether the sentence you are reading matches the sentence in the source. In this category, it very often does not. If you want help working out how findable your business really is inside AI answers and voice assistants, you can book a call with me.