UK AI Usage in 2026: What It Means for SEO and Marketing

There are now enough studies about AI use in the UK to confuse almost anybody.

One says nearly 60% of people are using generative AI. Another says roughly a third have used AI for web searches. Then we have research showing that only 3% use AI as their main way to find a product to buy.

So, which number is right?

Probably all of them.

They are measuring different things. Using an AI tool once is not the same as using it every day. Reading an AI Overview in Google is not the same as deliberately opening ChatGPT. Preferring a platform does not mean using it for every search. And being satisfied with a tool certainly does not tell us how many buying decisions it influences.

This distinction gets lost surprisingly quickly.

Some people see ChatGPT usage growing and decide Google is finished. Others see Claude getting better satisfaction scores and suddenly we all need a Claude strategy. And because younger people use AI more often, every AI campaign apparently needs to target Generation Z.

The actual picture is more complicated. Annoying, I know.

AI is becoming part of how people research, compare and understand things. But it has not replaced search engines, websites, reviews, social media or all the other places people look before making a decision.

It has added another layer to an already messy process.

Earlier YouGov research already showed that AI search had not replaced search engines. The newer platform rankings, together with data from the CMA, Ofcom, DSIT and the ONS, give us a wider view of who is using AI and what they are actually using it for.

Why the numbers do not match

Several useful UK studies were published during 2025 and 2026. Their headline numbers look very different because the researchers asked different questions.

Research	What it measured	Main finding
Department for Science, Innovation and Technology	Use of generative AI during the previous three months	59% had used generative AI
Ofcom	Use of tools such as ChatGPT, Copilot and Gemini	54% said they used these tools
YouGov	Use of AI tools for web searches during the previous 30 days	31.7% had used AI for web searches
Competition and Markets Authority	The method used most often for particular search tasks	13% used AI most often for complex information and 3% for finding a product

The Department for Science, Innovation and Technology’s Public Engagement Survey 2025/26 is probably the broadest of these studies. It was based on responses from 30,698 people and found that 59% had used generative AI during the previous three months.

Among those users, 75% said they used it at least weekly and 35% daily. Saving time, generating ideas and summarising or explaining information were among the main reasons.

That is meaningful adoption. We are well beyond AI being something only a few enthusiasts and technology journalists use.

But it does not mean 59% of the population has replaced Google with ChatGPT. It does not even mean they are using AI for search. Someone using Copilot to improve an email is part of the 59% as well.

The Competition and Markets Authority’s consumer research into search behaviour asked a different and, from a marketing perspective, more useful question.

Which method do people use most often for a particular task?

Among UK smartphone users, AI was the most commonly used method for 13% of more complicated information searches. For simple information, the figure was 4%. For finding a particular website it was 5%, and for finding a product to buy just 3%.

AI usage is real. But the type of task makes a big difference.

Age matters, but it does not explain everything

The DSIT survey found generative AI use above 75% among people under 45. Usage then dropped among older groups, reaching 18% for people aged 75 to 84 and 8% for those over 85.

Nothing terribly surprising there.

The differences by education and location are at least as interesting. Seventy-five per cent of respondents with a degree had used generative AI. That compared with 48% of people with other qualifications and 19% of those without formal qualifications.

Reported usage was 69% in London but 51% in the North East.

So “AI users” are not one audience. Neither are “younger AI users”, for that matter.

Occupation, education, confidence with technology and the actual job someone is trying to do will all affect whether and how they use AI.

YouGov’s research into AI use by generation produced one finding I did not expect.

It found that 31.7% of British adults had used an AI tool for web searches during the previous 30 days. Millennials reported the highest use at 37.3%, followed by Generation X at 35.4%.

For Baby Boomers and older adults it was 26.4%. For Generation Z it was 24.6%.

In other words, the youngest group was not the one most likely to say they used AI for web searches.

Gen Z and Millennials did use AI across a wider range of activities, including generating ideas, summarising documents and helping with emails. But that is a rather different story from “young people no longer use Google”.

It is also worth remembering that this is cross-sectional research. It shows differences between groups. It does not prove that age itself caused those differences.

A preferred AI platform is not necessarily the most-used one

YouGov also looked at which AI brands people prefer.

In its overall UK AI brand rankings, ChatGPT had a preference score of 30.0. Alexa followed at 12.7, Microsoft Copilot at 11.6 and Google Gemini at 11.3.

ChatGPT led among Generation Z, Millennials and Generation X. Alexa led among Baby Boomers and older users.

Useful brand research, yes.

Search market share, no.

A preference score does not tell us how many searches take place on the platform. It does not tell us how frequently people use it or what influence it has on a purchase.

The study also puts tools such as ChatGPT and Claude in the same wider category as Alexa and Siri. That is understandable if the subject is consumer-facing AI brands. But asking Alexa for the weather is not quite the same as using Claude to analyse a contract or ChatGPT to compare software suppliers.

So we need to be careful with the category.

The results show which brands have managed to build familiarity and preference. They do not show which platform sends the most valuable customers or plays the biggest role in buying decisions.

Satisfaction is another thing again

YouGov’s research into the AI platforms with the most satisfied users adds another set of numbers.

Claude had the highest overall net satisfaction score at 56.2. Gemini followed at 46.5, with ChatGPT at 46.0.

Among current users, DeepSeek had the highest score at 77.2, followed by Claude at 74.5 and Perplexity at 70.5. ChatGPT recorded 66.6.

These are net satisfaction scores. They are not the percentage of users who are satisfied, which would be a fairly easy mistake to make when looking at them quickly.

They are not a simple quality league table either.

A smaller platform used by a fairly technical or enthusiastic audience may get a higher score than a mass-market product trying to serve everybody. The users are different and they probably expect different things.

Former users were generally less satisfied than current users. Again, not a great shock. But the figures do not prove dissatisfaction was the reason they stopped using the platform.

Maybe it was. Or perhaps they only needed it for one particular job. Perhaps another tool was included in their existing software. Perhaps their employer blocked it. We do not know.

Popularity and satisfaction are not the same thing. And neither tells us enough about how the platform affects a customer’s decision.

What are people actually using AI search for?

This is where the CMA research gets more interesting.

Its qualitative research into consumer search behaviour involved 58 existing generative AI search users. Participants completed search tasks, recorded their screens and discussed how and why they searched.

A study of 58 people cannot tell us how common a behaviour is across the UK. That is not what qualitative research is for. It can help us understand the behaviour itself.

People tended to use AI for jobs involving several steps, comparisons or a lot of information.

That included planning a holiday, comparing the advantages and disadvantages of products and services, or creating something such as a meal plan.

For straightforward searches, people still tended to use a traditional search engine. Find a website. Check a fact. Get from A to B without turning it into a conversation.

Even frequent AI users still used conventional search more often overall. They also returned to search engines and original sources when they wanted to check an AI answer. Particularly when the decision involved an expensive purchase, work or education.

And people were reluctant to let AI make the final purchase for them.

That sounds more like AI being added to the customer journey than AI replacing the customer journey.

A customer might use ChatGPT to understand the problem, Google to check some of the claims, Reddit to see what actual users complain about, YouTube to see the product in use and the company website to check specifications.

Analytics might only show the final branded search or direct visit.

I looked at this wider problem in The Messy Middle Didn’t Get Messier. It Got Harder to See. AI did not create fragmented customer journeys. It just made even more of the journey difficult to observe.

Giving all the credit to the last click was never particularly convincing. It is becoming less convincing by the month.

Google is still here. AI is also inside Google

Much of the discussion treats Google and AI platforms as completely separate alternatives.

They are not.

According to Ofcom’s Adults’ Media Use and Attitudes 2026 report, 75% of online adults read AI-generated summaries in search results at least sometimes. Forty-two per cent said they did so often or always.

Even among people who did not use standalone AI chatbots, 54% said they read AI summaries in search results at least occasionally.

This matters.

People do not need to make a conscious decision to use AI for an AI-generated answer to influence them. The answer is simply there in the search results.

For many businesses, this passive exposure may be more important than the number of people opening Claude or ChatGPT every morning.

At the same time, Ofcom’s Online Nation 2025 report found that Google Search still reached 82% of UK adults and handled approximately three billion UK searches each month.

The report also recorded 252 million UK web visits to ChatGPT in August 2025, up 156% year on year.

Web visits and searches are different measurements, so we should not put the two numbers next to each other and pretend they are market shares. But they show two things happening at the same time.

AI platforms are growing quickly. Conventional search is still enormous.

What does this mean for SEO?

No, SEO is not dead.

We have been through this before. Social media was going to kill search. Apps were going to kill websites. Voice search was apparently going to change everything. Some things did change. The old thing usually did not disappear.

What is changing is that rankings, clicks and organic sessions show us less of the full picture.

If someone uses AI to understand a subject or draw up an initial shortlist, a business might influence that decision without getting a website visit at that point.

The opposite is also true. Being mentioned in an AI answer is not automatically valuable.

If the platform describes the business inaccurately, recommends it for the wrong job or includes it without enough evidence to create confidence, the mention may be worthless. It could even be unhelpful.

Businesses need to look at whether their important information is clear and supported:

What does the business actually do?
Who is the service for?
Who is it not for?
Why would somebody choose it instead of the obvious alternatives?
What evidence supports the claims?
What are the limitations, costs and trade-offs?
Can important facts be checked somewhere other than the company’s own website?

This is not an argument for producing another 500 generic blog posts.

Publishing more content is not the same as becoming more useful. Repeating all the common questions in slightly different words will not create expertise either.

Often, the work is much more basic. Improve the service pages. Make product information clearer. Publish useful comparisons. Show evidence. Explain the difficult bits. Stop hiding everything meaningful behind vague marketing language.

SEO and AI-search data can also tell us more than whether we received another click or citation. It can show how customers describe their problems, which competitors they compare and what the market does not understand about the business.

If we only use the data to produce another visibility score, we are wasting most of its value.

What about marketing?

Marketers need to be careful with the easy audience segments.

Age matters, but it is not enough. Education, occupation, location and confidence with technology also affect AI use. More importantly, so does the task.

Someone asking AI to explain an unfamiliar subject is not in the same position as somebody comparing three suppliers. And neither is the same as someone looking for a price, a phone number or a checkout page.

Different task. Different stage. Different information needed.

Trust is part of it as well.

The Office for National Statistics’ June 2026 research into public attitudes found that 32% of adults would not trust AI with any of the tasks they were shown. Only 7% said they would trust it to make business decisions.

The ONS used a broad definition of AI, so this is not a clean measure of trust in chatbots. Still, it suggests people are not about to hand every important decision to a machine.

They may use AI to explore a decision without trusting it to make the decision.

That means businesses still need things people can check. Detailed product information. Named expertise. Reviews. Case studies. Independent coverage. Proper evidence, not just the company saying it is brilliant.

It also does not mean businesses need to appear on every AI platform.

Trying to cover Google, ChatGPT, Gemini, Claude, Perplexity, Reddit, TikTok, LinkedIn, YouTube and whatever launches next Tuesday is not a strategy. Particularly if the budget and team were already stretched before AI came along.

The useful question is where your customers look when making decisions that matter to the business.

Start there.

What should a business investigate?

Before buying another AI visibility tool or ordering lots of “AI-friendly content”, I would want answers to a few fairly basic questions:

Which customer questions actually affect consideration and purchase?
Where are customers asking those questions?
How is the business represented in the answers?
What evidence do the systems use?
Is the representation accurate and commercially useful?
What could reasonably be improved?

That means looking at conventional search, AI features within search and the standalone platforms relevant to the audience.

It also means separating three things which are often bundled together: being mentioned, being described correctly and being presented as a credible option.

They are not the same.

This is also how I would approach an AI Search Audit. Not by producing a supposedly universal AI visibility score, but by investigating how the business is understood and represented around questions its customers actually ask.

The answer might be better service-page copy. It might be clearer product data, stronger third-party evidence or useful comparison content. It might be inconsistent information about the company across different websites.

And sometimes the answer may be that AI search is not yet commercially important enough to deserve a large investment.

That is still a useful finding.

So, where does this leave us?

AI is clearly part of search behaviour in the UK.

Its influence is wider than the number of people deliberately using ChatGPT, Claude or Gemini because AI-generated answers are now inside conventional search results too.

But there is no good evidence that AI has replaced search engines, websites or the need for people to verify information.

People use it for particular jobs. It appears especially useful when they need to understand something complicated, combine information or compare options. Search engines remain important for finding websites, checking facts and getting closer to an actual transaction.

For SEO and marketing, the question is not whether we should choose Google or AI.

Customers are already using both.

The job is to understand where each one enters the decision, what information the customer needs at that point and whether the business is represented properly.

Less exciting than promising to “rank in ChatGPT”, perhaps.

But rather more useful.

Sources and methodology

The studies in this article measure different populations, periods and behaviours. AI usage, platform preference, satisfaction and the method people use most often should not be treated as different versions of the same market-share figure.

YouGov research:

Other research:

Not sure whether any of this matters for your business?

A free 15-minute call is usually enough to work out whether AI search deserves your attention, or whether the sensible priority is somewhere else. No pitch deck. No invented visibility crisis.

Book a free 15-minute call

How People in the UK Use AI – and What It Means for SEO and Marketing