In the three years since ChatGPT launched to the public, artificial intelligence (AI) has taken the world by storm, with daily media headlines, soaring stock valuations, and hyperbolic social media posts indicating that AI has become today’s “hottest” digital topic.

However, while individual platforms have announced a number of notable milestones, there are still precious few evidence-based figures for how many people might actually be using these tools overall.

This article will attempt to address that gap, offering detailed analysis to demonstrate that well over 1 billion people now use standalone AI tools each month.

Definitions

Use of the abbreviation “AI” has come to encompass many things over the past couple of years, so – for the purposes of this article – it’s worth clarifying that we use “AI” to refer primarily to large language models (LLMs) and generative AI applications (GenAI) such as ChatGPT, Google Gemini, Claude, Perplexity, and Grok.

We also refer to “standalone AI tools”, which are distinct from broader applications that make use of generative AI technologies.

And for the avoidance of doubt, our reference to “standalone AI tools” does not include:

AI Overviews in Google search results
The use of AI capabilities within software and tools such as Gmail, Microsoft Office, Canva, Adobe Photoshop, or Grammarly
The use of “AI companion” chatbots such as Character.AI
Use of Meta AI within apps such as Facebook, WhatsApp, and Instagram, because such use is indistinguishable from these platforms’ generic search functions.

Quantifying global AI use

Despite the furore surrounding AI, it’s still remarkably challenging to put an exact figure on current levels of global AI adoption and use.

On the one hand, many of the top AI companies like OpenAI and Anthropic remain “private” – as opposed to being publicly traded companies like Google and Meta – so they tend to make fewer announcements about their active user numbers.

Meanwhile, media headlines can resort to hype and speculation, resulting in inflated figures that overrepresent actual use.

On the other hand though, the available user data may underrepresent actual use, partly because active user figures aren’t reported on a regular schedule.

And furthermore, platform figures typically only reference the use of a single tool, and our research indicates that a meaningful number of active AI users do not use the world’s most widely used AI platform.

For example, Kepios analysis of data from Similarweb suggests that roughly 4 in 5 active users of a selection of the top LLMs outside of China use ChatGPT at least once each month.

However, that same figure also reveals that roughly 1 in 5 active AI users does not use ChatGPT each month, instead preferring platforms like Google’s Gemini, Microsoft’s Copilot, and “independent” platforms like Deepseek, Perplexity, and Claude.

And when viewed in totality, this data offers compelling evidence for a ten-figure active AI user total.

That’s a fairly sweeping claim though, so let’s break this number down into its constituent parts.

Use of top AI platforms

To avoid confusion, it’s worth noting that Similarweb’s figures for ChatGPT use are somewhat lower than those reported by OpenAI

Moreover, the figures you’ll find in our Digital 2026 Global Overview Report relate to separate use of each platform’s web tools and mobile apps.

Even then though, Similarweb’s data indicate that more than 550 million people use ChatGPT’s mobile app each month, while close to 500 million unique “device identities” visited ChatGPT’s web platform in August 2025.

There will inevitably be some degree of overlap between ChatGPT’s mobile app users and web platform visitors of course, but nonetheless, these figures are already impressive.

However, OpenAI’s own figures indicate significantly higher levels of ChatGPT use.

For example, a chart in an academic study published by the National Bureau Of Economic Research (NBER) in partnership with OpenAI indicates that weekly active ChatGPT users had already passed the 750 million mark by September 2025.

Meanwhile, OpenAI’s CEO, Sam Altman, reported that ChatGPT already had 800 million weekly users during an event in early October 2025.

And furthermore, Kepios analysis of August 2025 data from Similarweb suggests that ChapGPT’s monthly active user figure should be roughly 1.3 times higher than its weekly active user figure, which puts the likely monthly user figure for ChatGPT around 1 billion in its own right.

But as we saw above, Similarweb’s data also shows that ChatGPT’s user base “only” accounts for 4 in 5 active users of a selection of the top LLMs outside of China.

So, despite OpenAI’s impressive market share, data suggests that somewhere between 100 million and 200 million active AI users do not use ChatGPT each month.

And if we add these users of other platforms to the ChatGPT figure we saw above, we’re already well into ten-figure territory.

Getting to a global figure

However, it’s also important to remember that ChatGPT remains blocked in China, so – in order to identify a global AI user figure – we also need to factor separate AI use in China.

For initial perspective, Alibaba announced that its Quark AI platform had 150 million worldwide monthly active users in April 2025, and while that figure undoubtedly includes users outside of China, our analysis suggests that a meaningful proportion of Quark’s users live within its home country.

But perhaps more importantly, a July 2025 report indicated that China is now home to 250 million “AI agent users”, which is more than enough to push our global user figure well above 1 billion.

So, while an exact figure remains elusive, the available evidence indicates that there are now well over 1 billion people around the world using standalone AI platforms each month.

And in fact, the figures we’ve seen so far suggest that the global figure could already be somewhere in the region of 1.5 billion.

But for broader perspective, the reason we reference “standalone” AI platforms is that Google’s AI search summaries already reach more than 2 billion active users each month.

These overall figures for AI adoption don’t offer much insight into AI use though, so let’s explore some other metrics that can help us build a more informed picture of AI’s place in people’s everyday lives.

We’ll start by looking at how people feel about AI, before looking at the world’s most popular AI platforms, and then finishing up with a closer look at what people are actually using these tools for.

Excitement about AI

The latest research from GWI reveals that just under half of the world’s online adults are “excited” about artificial intelligence, with that share figure edging up over recent months.

However, it’s worth highlighting that this same figure might imply that more than half of connected adults are “not” excited about AI [even if the lack of an affirmative does not necessarily imply a negative].

Moreover, GWI’s data shows that enthusiasm for AI varies significantly by country and culture, and this particular dataset reveals some important surprises.

At the top end of the spectrum, more than three-quarters of online adults in Nigeria (77 percent) say that they’re “excited” by AI, and the same is true of more than 7 in 10 internet users over the age of 16 in Ghana and Turkey.

Positive emotions are more muted in the world’s two most populous nations though, with fewer than half of connected adults in China (47.8 percent) and India (47.3 percent) expressing enthusiasm for AI technologies.

And – strikingly – fewer than 1 in 3 Americans (33.0 percent) say they’re excited about AI, while the figure falls below 30 percent in the United Kingdom.

Indeed, Brazil is the only one of the world’s ten largest economies where more than half of online adults say they’re enthusiastic about artificial intelligence.

And these geographic trends may have significant implications for AI’s financial outlook.

Bubble, bubble…

For context, it seems that investors are considerably more bullish than the public at large when it comes to AI – at least at the time of writing.

Indeed, AI had become one of the biggest stories in financial markets by early October 2025, with Michael Cembalest, chairman of market and investment strategy at J.P.Morgan, stating that:

“AI-related stocks have accounted for 75 percent of S&P 500 returns, 80 percent of [company] earnings growth, and 90 percent of capital spending growth since ChatGPT launched.”

With eye-watering numbers like that, it’s perhaps not surprising that some analysts suggest we’re now in “AI bubble” territory.

And in fact, even Open AI’s CEO, Sam Altman, has said he believes investors may be “overexcited” by AI.

So, can current valuations be justified?

[Note: nothing in this article or in our Global Digital Reports constitutes any form of investment advice]

Monetising AI models

It’s not yet clear whether the majority of AI revenues will eventually come from subscription fees or from ad-supported models, but it seems reasonable to assume that the industry may embrace both.

However, the geographic trends in AI sentiment that we explored above may be cause for concern for AI companies and investors alike.

For example, whilst Netflix operates in a very different industry to OpenAI, the video streaming giant derived more than three-quarters (76 percent) of its Q2 2025 revenues from users in North America and Europe, suggesting that subscription-centric digital businesses are still largely dependent on more developed economies for their financial success.

And that geographic revenue skew is even more pronounced amongst digital companies that depend on advertising.

For example, Google’s parent company, Alphabet, earned nearly half (48 percent) of its Q2 2025 global revenues from users in the United States, while more than two-thirds of Meta’s Q2 2025 revenues came from North America and Europe.

Of course, adoption rates, use cases, and revenue models may all turn out to be quite different for AI companies, but as a recent Reuters article highlights,

“Fewer than 2 percent of the 800 million or so people using ChatGPT pay for the service, and a growing number of them live in low-income countries such as India.”

Moreover, OpenAI itself recently published the following statement:

“ChatGPT has also become a broadly accessible global tool, with especially rapid growth in low- and middle-income countries. By May 2025, ChatGPT adoption growth rates in the lowest income countries were over 4 [times higher than] those in the highest income countries.”

For added perspective here, OpenAI recently launched a special subscription package for its flagship platform in India, with a monthly subscription fee of roughly USD $4.60.

And critically, that India-specific fee is more than four times cheaper than the “standard” ChatGPT Plus monthly subscription fee of USD $20 in the United States.

Impact on digital behaviours

But geographic variations in AI’s appeal aren’t just of interest to investors.

Indeed, many of the countries where we see the highest rates of “AI excitement” are also the countries that are experiencing the most rapid growth in overall internet adoption.

Consequently, many of the people coming online for the first time today across Africa and the Indian subcontinent may be “AI native”, and as a result, they may adopt quite different online behaviours compared with users who have been online for more than a couple of years.

In particular, many of those newer users may never adopt the same kind of “online search” behaviours that have defined people’s internet journeys over the past 25 years, and that evolution could have important consequences for how tomorrow’s internet users discover, research, and buy brands.

For perspective, GWI’s latest research indicates that search engines like Google remain the most important source of new brand discovery amongst the world’s online adults, with, roughly 1 in 3 respondents selecting this channel from a range of predefined options [however, note that GWI’s list of options doesn’t yet include LLMs].

Influence on traditional search

But AI may reshape “traditional” search behaviours amongst more seasoned internet users too, especially now that more than 2 billion users see AI Overviews within Google results each month.

And more broadly, Google’s CEO, Sundar Pichai, has also suggested that broader AI use has resulted in longer search queries.

However, third-party analysis using Similarweb data suggests that this particular claim may be overstated, at least when it comes to user behaviour within Google search itself.

Indeed, one study found that the average length of Google Search queries in the US only increased by about 3 percent over the past two years, from 3.27 keywords in early 2023 to 3.37 keywords in early 2025.

And furthermore, the same research found that the average length of Google search queries in the UK actually declined after the introduction of AI Overviews, from an average of 3.18 keywords in August 2024, to 3.17 keywords in February 2025.

For perspective, Similarweb’s analysts have informed us that the average length of a ChatGPT prompt is already close to 100 characters – i.e. 30 times longer than the typical Google search query.

Despite these findings, however, the growing popularity of AI-centric (re)search platforms like Perplexity suggests that the world’s online search behaviours are indeed evolving.

And as a result, marketers will want to keep a close eye on the keywords that drive organic search traffic to their brand’s website – and to the websites of their competitors.

The world’s most popular AI platforms

As we noted above, most AI companies tend to be quite conservative when it comes to disclosing user figures, so we need to rely on third-party data in order to make sense of comparable trends across different AI platforms.

Fortunately however, Similarweb’s Web Intelligence and App Intelligence offerings allow us to build a detailed understanding of platform adoption and use.

Data for China remains elusive though, so note that the following analysis does not include coverage of AI activity within the country.

ChatGPT dominates AI mobile apps

OpenAI’s primary platform remains the top choice across the web and mobile apps, accounting for more than four times as many users as its next nearest rival in both environments.

Interestingly, Similarweb’s data suggests that the ChatGPT mobile app now attracts a greater number of unique monthly users than the ChatGPT website, with the platform’s Google Play and iOS apps attracting a combined total of 557 million monthly active users.

That’s almost eight times greater than Google Gemini’s 70 million monthly app users, but that figure is still enough for the Alphabet offering to claim second place in our latest ranking of the most popular “standalone” AI mobile apps.

Deepseek ranks third, with close to 60 million monthly active users outside of China.

Perplexity and X’s Grok round out the top five AI mobile apps, with 39.4 million and 38.9 million MAUs (respectively).

Note that we’ve removed Microsoft 365 Copilot from the figures that we report for Microsoft Copilot, because the “365” app is more focused on augmenting Office 365 products like Word and PowerPoint, as opposed to the AI chatbot focus of the more “standalone” Microsoft Copilot app.

The top AI web platforms

The same group of platforms claim the top five spots in our AI web ranking too, although their respective order is a little different.

ChatGPT still holds the top spot, with Similarweb’s data indicating that chatgpt.com attracted 489 million unique visitors during August 2025.

Google Gemini ranks second again too, with 122 million unique visitors to gemini.google.com in August.

And Deepseek also claims third spot in the web rankings, with deepseek.com enjoying 45.2 million unique visitors during the same month.

Grok and Perplexity swap places in our web ranking though, with grok.com attracting 26.3 million unique visitors in August 2025, and perlexity.ai attracting 24.8 million.

AI referrals to third-party web properties

But if we’re to compare AI platforms more directly with search engines, we need to understand how platform use translates into traffic referrals to third-party properties.

And interestingly, new data suggests that – specifically amongst AI platforms – ChatGPT refers the greatest volume of traffic to third-party websites, with analysis from Statcounter indicating that the OpenAI site is responsible for more than 4 in 5 referrals (80.92 percent) originating from the top LLMs.

Meanwhile, despite only placing fifth in the web user rankings, Perplexity ranks second in Statcounter’s ranking of AI referrals, with the search-oriented platform responsible for 8.12 percent of AI referrals.

And Microsoft’s Copilot ranks third, with Statcounter’s data indicating that the platform delivers 5.17 percent of AI referrals to third-party websites.

Orders of magnitude

However, if we compare AI referral volumes with traffic referrals from traditional search engines like Google, the findings are… sobering.

For initial context, Similarweb reports that fewer than 3 in 10 visits to reddit.com come from “direct” traffic, meaning that more than 7 in 10 visitors to the Reddit’s website come either from search referrals, or from link clicks on other websites and social platforms.

But here’s the kicker: for every 71 “non-direct” visitors to reddit.com in August 2025, more than 68 came from organic search, meaning that all other forms of referral combined contributed just 3 percent of Reddit’s web traffic

More specifically, Similarweb reports that organic search platforms delivered roughly 2.8 billion of Reddit’s total 3.95 billion total visits in August 2025.

So how does AI compare?

Well, the simple answer is that it doesn’t.

Indeed, Similarweb’s data suggests that chatgpt.com contributed just 8.7 million visits to reddit.com’s traffic in August 2025 (0.24 percent of total traffic), while the figure for perplexity.ai was just 2.1 million (0.06 percent of total traffic).

However, for perspective, social channels delivered just 49.6 million visits to reddit.com (1.22 percent of total web visits), so the key story here is that traditional organic search is still absolutely essential for web success.

What are people using AI for?

Interestingly, data reveal that the world’s 1 billion AI users may not be using AI platforms in the same ways that they use search engines, even if there are some inevitable overlaps in behaviour.

For clarity, we haven’t yet seen any comprehensive research across multiple countries and platforms that offers consistent and comparable insight into people’s AI behaviours, but there are some valuable studies that can help us make sense of evolving AI behaviours.

Common AI use cases

The first of these is a detailed study conducted by Marc Zao-Sanders that appeared in Harvard Business Review (HBR), with the first iteration published in March 2024, and an updated version with new data published in April 2025.

The latest version of Zao-Sanders’s study concluded that “therapy and companionship” represents the most prevalent use case for AI tools like ChatGPT in 2025, with this activity moving up one place in the rankings compared with his 2024 study.

It’s important to stress that this dataset includes broader AI applications like “companion chatbots” though, so this ranking is not exclusive to the use of “standalone” AI tools.

“Organising my life” ranks second in Zao-Sanders’s 2025 list, while “finding purpose” ranks third, and it’s interesting to note that both of these activities did not appear in Marc’s 2024 ranking.

Next up, “enhanced learning” ranks fourth, and “generating code” rounds out the top five, with both of these activities seeing their ranking increase between 2024 and 2025.

Making sense of people’s AI use

At first glance, it’s perhaps unsurprising to see activities related to learning and generating code amongst the top five.

For example, studies have found a marked drop in LLM and GenAI use during “summer holiday” periods for schools and universities in the Northern Hemisphere, as well as a marked resurgence in activity when students return to class.

Meanwhile, recent analysis published by Anthropic indicates that “computer and mathematical” activities dominate the primary use cases of its Claude platform:

“Since December 2024, computer and mathematical uses of Claude have predominated among our categories, representing around 37-40 percent of conversations.”

But the other three activities in Zao-Sanders’s top five are perhaps more interesting.

For initial context, a recent study exploring ChatGTP use published by OpenAI stated the following:

“ChatGPT consumer usage is largely about getting everyday tasks done. Three-quarters of conversations focus on practical guidance, seeking information, and writing—with writing being the most common work task, while coding and self-expression remain niche activities.”

Aside from its unabashed use of the em dash [did AI write that summary?!], there’s plenty to explore in this statement.

First of all, OpenAI acknowledges that coding remains a “niche activity” on ChatGPT – in contrast to the Anthropic statement we saw above – which reveals important differences in the use of individual AI platforms.

Secondly, the reference to “getting everyday tasks done” is also particularly relevant to Zao-Sanders’s HBR ranking that we introduced above, where “organising my life” placed second.

And it’s especially interesting to note that this particular activity is a “new entry” in this year’s HBR list, despite its prominent second place in the ranking.

But it’s positions 1 and 3 in this year’s HBR ranking that really caught my attention.

Can AI be a “friend”?

Based on various data points that we’ve been tracking over recent months, seeing “companionship” at the top of this ranking didn’t come as a total surprise, but its position relative to some other activities is still truly remarkable – and more than a little unnerving.

It’s important to highlight at this point that OpenAI’s own research suggests that “companionship” activities may not be as prevalent as Zao-Sanders’s study concludes:

“[T]he share of messages related to companionship or social-emotional issues is fairly small: only 1.9 percent of ChatGPT messages are on the topic of Relationships and Personal Reflection, and 0.4 percent are related to Games and Role Play. In contrast, Zao-Sanders (2025) estimates that Therapy and Companionship is the most prevalent use case for generative AI.”

However, there’s plenty of evidence beyond these two studies to support the conclusion that “companionship” is indeed a popular AI use case.

For example, Similarweb’s App Intelligence data indicates that a number of “AI companion” apps – notably Character.AI – have attracted millions of monthly active users.

And this data suggests that many people may turn to dedicated “conversation bots” when seeking “AI companionship”, rather than using more “general” AI platforms like ChatGPT.

But it’s also important to view the prevalence of “AI companionship” through the lens of rising rates of social isolation.

For context, the World Health Organisation (WHO) has declared loneliness to be a “pressing global health threat”, and in a comprehensive study published in 2023, the organisation concluded that loneliness can have an alarming impact on human health:

“Loneliness is far more than just a bad feeling – it harms both individual and societal health. It is associated with a greater risk of cardiovascular disease, dementia, stroke, depression, anxiety, and premature death. The mortality impact of being socially disconnected is similar to that caused by smoking up to 15 cigarettes a day, and [is] even greater than that associated with obesity and physical inactivity.”

For added context, as part of the launch of a commission to address the “loneliness epidemic”, a WHO spokesperson also revealed that:

“[Loneliness] transcends borders and is becoming a global public health concern affecting every facet of health, wellbeing and development. Social isolation knows no age or boundaries.”

Friend or foe?

These insights may go some way to explaining why so many people are turning to AI platforms for companionship.

And in theory, if AI can help people address feelings of loneliness, these activities could potentially deliver tangible health benefits.

However, it remains to be seen whether AI can help to mitigate people’s longer-term feelings of loneliness, or whether such tools may actually exacerbate it, by reducing human interaction and driving people further into isolation.

Alarmingly, some have even alleged that AI may actively encourage people to self harm.

Privacy remains a key consideration too, and there is a danger that people will increasingly share “sensitive” data that could eventually be used to their disadvantage, especially when it comes to therapy-oriented conversations.

Worryingly, the latest research from GWI indicates that most people don’t seem to be as worried about how companies might use their personal data as media headlines might have you believe.

And in fact, at a global level, just 30.7 percent of survey respondents across 54 of the world’s largest economies expressed concerns about companies’ potential use of their data.

That figure varies meaningfully by geography though, from a high of more than 50 percent in Spain, to barely 10.5 percent in Nigeria.

However, with research indicating that younger people may be particularly prone to the effects of loneliness – especially younger men – it’s worth highlighting that these age groups may also be disproportionately at risk of divulging “sensitive” data online.

Younger generations already demonstrate lower levels of concern and perceived risk when it comes to sharing personal information online, and as this quote cited in Zao-Sansers’s HBR study observes,

“Unfortunately, data safety is not a concern when your [mental] health is deteriorating, and survival is the ‘morning agenda’.”

Does AI understand the meaning of life?

But even more than “AI companionship”, there’s something particularly striking about the prevalence of people asking a machine questions about “purpose”.

And if I’m honest, I find it somewhat unsettling that “finding purpose” ranks third in Zao-Sanders HBR study.

Having said that, this particular behaviour has clear digital precedent, and Google Trends data shows that asking machines questions about the “purpose” isn’t a new phenomenon.

For example, worldwide Google search volumes for the “meaning of life” appear to have reached a crescendo during 2023, but they’ve also risen again sharply over the past two months, following a few months of decline.

I’m not a trained psychologist, so I won’t attempt to derive any specific meaning in this data.

However, from a layman’s perspective, when viewed together with the prevalence of “finding purpose” in the HBR ranking we saw earlier, the trends in this data suggest that there may be merit in further research of these topics by more qualified individuals.

And moreover, the fact that so many people are turning to AI to ask profound, existential questions may have wide-reaching consequences.

Will AI replace search engines?

But turning our attention back to more prosaic concerns, the question remains: are AI platforms likely to replace search engines?

At this point, it’s important to highlight that Zao-Sanders’s HBR study is informed by the volume of conversations about specific AI topics in forums like Reddit and Quora, so it may not be fully representative of actual AI query volume or frequency.

However, it’s still interesting to note his finding that references to “specific search” have actually declined over the past year, slipping from tenth place in his 2024 HBR ranking, to 13th place in his more recent 2025 list.

Of course, some of the other activities in this year’s list may well have replaced activities that would previously have relied on search engines, with “healthier living”, “cooking with what I have”, and “troubleshooting” all being potential examples.

However, as the US National Bureau of Economic Research (NBER) and OpenAI discovered in their analysis of 1.5 million distinct “conversations” that took place on ChatGPT around the world between March 2024 and June 2025,

“The three most common Conversation Topics are Practical Guidance, Seeking Information, and Writing, collectively accounting for about 77 percent of all ChatGPT conversations. Practical Guidance has remained constant [over the past year] at roughly 29 percent of overall usage. Writing has declined from 36 percent of all usage in July 2024 to 24 percent a year later. Seeking Information has grown from 14 percent to 24 percent of all usage over the same period.”

More specifically, roughly a quarter (24.4 percent) of all ChatGPT queries in July 2025 involved “seeking information”, which the authors state includes looking for “specific information”, researching “purchasable products”, and asking questions related to “cooking and recipes”.

The marketing opportunity

The same NBER report also explores a more granular sample of 1.1 million ChatGPT conversations initiated by roughly 130,000 distinct users, and this particular dataset reveals an even more nuanced story.

Overall, “seeking specific information” was the single most prevalent topic of “conversations” with ChatGPT between May 2024 and June 2025, accounting for more than 1 in 6 user queries, or 18.3 percent of the total.

And for added perspective, the NBER report also indicates that ChatGPT users are sending billions of messages each day:

“By July 2025, ChatGPT [users]… were collectively sending more than 2.5 billion messages per day, or about 29,000 messages per second.”

Putting these two sets of figures together, it seems reasonable to deduce that ChatGPT handles more than 450 million conversations “seeking specific information” every day, which equates to almost 14 billion such conversations each month.

Meanwhile, the NBER data indicates that conversations relating specifically to “purchasable products” accounted for 2.1 percent of total conversations during the study period, suggesting that ChatGPT may be addressing more than 50 million user conversations relating specifically to purchasable products every single day, or 1.6 billion every month.

Despite these impressive numbers, Similarweb’s data indicates that google.com still attracts almost 15 times as many visits as chatgpt.com, so there’s a very real possibility that search engines still deliver well over 90 percent of the answers to people’s online questions about potential purchases.

However, monthly visits to chatgpt.com have more than doubled over the past year – from 2.63 billion in August 2024, to 5.85 billion in August 2025 – whereas monthly visits to google.com only increased by 0.4 percent during the same period, from 83.5 billion in August 2024 to 83.9 billion in August 2025.

So, marketers may well want to explore how they can ensure their brands are more visible in AI “consumer conversations”.

Conclusion: AI offers mostly new opportunities

But despite the rapid rise in the use of LLMs like ChatGPT, our analysis suggests that the use of “traditional” search engines like Google is currently only declining at a rate of between 1 and 2 percent per year.

And as a result, asking whether LLMs might “replace” traditional search engines might be a misplaced – and commercially dangerous – distraction.

Indeed, the available evidence does not appear to suggest that the majority of AI user activities has been a direct substitute for activities that would previously have gone to platforms like Google or Bing.

And in fact, my reading of the data indicates that a significant proportion of “commercial conversations” that take place in LLMs are incremental: they’re additions to our audiences’ activities in search engines, not a direct substitute or replacement.

In other words, LLMs likely represent an additional opportunity for marketers, and the current data does not indicate that “AI vs. search” is a zero-sum equation.

Of course, the specific impact of these trends will vary by industry, by geography, and also by brand, so marketers will need to conduct their own due diligence, and keep a close eye on trends in their own metrics and performance.

However, for me, the top takeaway in all of this data is that LLMs offer marketers opportunities to deliver new kinds of value, and as a result, AI should not be viewed as a “substitute” for online search within the marketing mix.

Disclosure: Simon Kemp is a brand ambassador for GWI.

About the author

Simon Kemp

Simon is Ru-Dark’s chief analyst, and CEO of Kepios.
Click here to see all of Simon’s articles, read his bio, and connect with him on social media.

Digital 2026: more than 1 billion people use AI