Overview
If companies are going to invest in AI product recommendation, the first question is not simply which large language model is the most intelligent. It is where consumers are actually using AI at meaningful scale.
AI product recommendation ultimately depends on human behavior. A recommendation only matters commercially if a person is there to ask the question, receive the answer, refine what they want, compare alternatives, and potentially make a purchasing decision.
This is why mass-market adoption matters.
Before looking at the data, the intuitive picture is relatively straightforward. ChatGPT has become a household name. It was many people’s introduction to conversational generative AI, and the behavior it introduced is remarkably natural: ask something, receive an answer, ask another question, refine the answer, and continue the conversation. OpenAI now reports that ChatGPT has surpassed one billion weekly active users, while its own usage research suggests adoption has continued to broaden across age groups, geographies, and recurring everyday use cases. OpenAI — Expanding access to AI with ChatGPT ads
Google occupies the other obvious position. Google Search was already deeply embedded in everyday human behavior before generative AI arrived. Google is now introducing conversational AI directly into that existing behavior through AI Mode, which has surpassed one billion monthly users. Google also says AI Mode queries have more than doubled every quarter since launch. Google — A new era for AI Search
The adoption data therefore largely confirms the intuition: ChatGPT and Google’s AI Search experience are currently the two clearest mass-market AI environments for product recommendation. Gemini is also becoming increasingly important as a standalone destination, while Claude and Perplexity remain considerably smaller consumer platforms today.
That does not mean the latter platforms should be ignored. It means companies should prioritize according to where consumers actually are.
Mass-Market Adoption Matters for AI Product Recommendation
AI product recommendation (AIPR) is different from simply measuring AI technological capability.
A model might perform extremely well on benchmarks. A particular AI system might be preferred by developers, researchers, or technical professionals. But if relatively few ordinary consumers use that system to research products, its immediate importance to mass-market AI product recommendation is inherently smaller.
For AIPR, what matters is the intersection of:
AI capability + consumer adoption + recommendation-oriented behavior.
The third part is especially important.
A person asking:
“What’s the capital of France?”
is using AI, but there is little commercial decision-making involved.
A person asking:
“I need a reliable car for everyday commuting that won’t cost much to maintain. What should I buy?”
has created a decision narrative.
They can then continue:
“What if reliability matters more than fuel economy?”
Then:
“Compare Toyota and Hyundai.”
Then:
“Which one would you choose for my situation?”
That conversational refinement is where AI becomes particularly important for product recommendation. The AI is no longer merely retrieving information. It is participating in the consumer’s process of defining, evaluating, and narrowing a decision.
So when determining which AI systems deserve investment, companies should look for platforms where this kind of human behavior can happen at mass-market scale.
Mobile Adoption Is One of the Cleanest Signals of Human AI Habits
Mobile usage is particularly valuable because smartphones are deeply embedded into everyday behavior.
People carry them everywhere. They use them while shopping, commuting, sitting on the couch, traveling, working, comparing prices, looking something up in a store, or simply wondering about something.
Installing and repeatedly using an AI application therefore provides a useful behavioral signal that the product has become part of someone’s normal digital toolkit.
The Digital 2026 Mid-Year Global Update similarly treats mobile-app usage as an important way to understand regular and sustained AI use, since users have actually gone through the process of installing an application and continuing to use it. We Are Social — Digital 2026 Mid-Year Global Update
Similarweb’s App Intelligence data estimated the following worldwide mobile audiences in February 2026:
ChatGPT: 592 million MAU
Gemini: 152 million MAU
Perplexity: 37.9 million MAU
Claude: 31.6 million MAU
These figures are third-party estimates rather than exact first-party counts, but their value is that the same basic methodology is being applied across competing mobile apps. We Are Social — Digital 2026 Mid-Year Global Update
The difference is substantial.
ChatGPT is not simply leading the mobile AI market. In this dataset, its standalone app audience is almost four times Gemini’s and many times larger than Claude’s or Perplexity’s.
Google AI Mode requires somewhat different treatment because it is not a standalone AI application. It operates through Google Search, including Google.com and the Google app. Google reports more than one billion monthly AI Mode users. For our directional mobile-versus-non-mobile analysis, broader Google.com behavior can be used as a proxy: Semrush estimated that roughly 70% of Google.com traffic was mobile across recent 2026 measurements. Semrush — Google.com Traffic Analytics
That does not measure the exact device distribution of AI Mode itself. It gives us a useful directional proxy for the environment in which Google AI is being adopted.
The broader pattern is difficult to miss:
ChatGPT and Google AI Mode already operate at genuinely mass-market scale.
Web Adoption Tells a Similar Story
Mobile is only one part of the picture.
People also use AI heavily through browsers on laptops and desktop computers. That environment can be particularly relevant for longer research sessions, work, comparison shopping, high-consideration purchases, and situations where users want more screen space or are already working at a computer.
Similarweb’s Web Intelligence data estimated that in February 2026:
ChatGPT.com attracted approximately 460 million unique monthly visitors.
Gemini’s web experience attracted approximately 253 million.
Claude attracted approximately 37.2 million.
Perplexity attracted approximately 24.3 million.
The same dataset was published in the Digital 2026 Mid-Year Global Update, giving us one comparable measurement framework across the major standalone AI websites. We Are Social — Digital 2026 Mid-Year Global Update
Again, the broad structure resembles what we see on mobile.
ChatGPT leads.
Gemini represents a meaningful second standalone consumer destination.
Claude and Perplexity remain considerably smaller.
That does not necessarily make them less capable. It means their current mass-market consumer distribution is smaller.
And distribution matters enormously in product recommendation.
Google AI Mode Is Particularly Important Because It Combines Search and Conversation
Google AI Mode deserves special attention because it sits at the intersection of two behaviors:
traditional search and conversational AI.
That is a strategically powerful position.
Search engines remain deeply embedded in purchasing behavior. Consumers already use Google to compare products, look up specifications, find reviews, research brands, evaluate alternatives, and locate merchants.
AI Mode effectively adds a conversational layer to that established behavior.
Instead of performing five independent Google searches, a consumer can ask a broader question, receive an AI-generated synthesis, ask a follow-up, introduce new constraints, compare alternatives, and progressively refine the decision.
That begins to look much more like the decision-narrative process we see in ChatGPT.
Google describes AI Mode as part of its effort to combine the capabilities of Search with generative AI, and reports that the product has surpassed one billion monthly users. Google — A new era for AI Search
At the model layer, AI Mode is powered by Google’s Gemini models. Google continues to update which Gemini model powers the experience as the underlying model family evolves. Google — AI Mode in Search
The particular model version will inevitably change. The more important strategic fact is that Google Search and Gemini are increasingly interconnected.
Google AI Overviews Should Be Treated Separately
Google AI Overviews is important, but it represents a different form of adoption.
Google reports that AI Overviews reaches more than 2.5 billion monthly active users. Alphabet — June 2026 Investor Presentation
Meanwhile, Similarweb estimates that Google.com attracts roughly 3.2 billion unique monthly visitors. We Are Social — Digital 2026 Mid-Year Global Update
These figures are measured differently, so they should not be interpreted as a literal penetration calculation. But they illustrate something important about the nature of AI Overviews:
AI Overviews is embedded AI adoption.
A person does not necessarily wake up and decide:
“I’m going to use Google AI Overviews today.”
They Google something.
Google determines that the query would benefit from an AI-generated response, and the Overview appears within the search results.
That creates extraordinary distribution, but it is different from deliberate conversational engagement.
Someone may read an AI Overview carefully.
Someone else may scan two sentences.
Someone may click one of the cited sources.
Someone else may scroll straight past it.
Unless we have behavioral data showing how deeply users actually engage with those results, the 2.5 billion figure should primarily be understood as evidence of massive AI exposure inside Google Search.
For AIPR, AI Mode is arguably the stronger indicator of deeper recommendation behavior because the consumer has entered an environment designed around asking questions, receiving synthesized answers, and continuing the conversation.
AI Overviews still matters enormously. But it belongs in a different category.
Primary AI Systems Companies Should Focus On
ChatGPT Should Be a Primary AIPR Investment
If a company can only prioritize a small number of AI environments today, ChatGPT should be one of them.
The quantitative case is strong: more than one billion weekly active users overall, hundreds of millions of mobile users, and hundreds of millions of unique web visitors. OpenAI — Expanding access to AI with ChatGPT ads
The behavioral case may be even stronger.
ChatGPT normalized the idea of asking AI.
A consumer does not need to learn an entirely new research workflow. They can simply ask:
“What’s a good laptop?”
And then continue talking.
The important strategic target is therefore ChatGPT as a consumer decision environment, rather than one particular model version. The underlying GPT model will continue changing much faster than the consumer habit of opening ChatGPT and asking it a question.
Google AI Mode Should Be Another Primary Investment
The same applies to Google AI Mode.
Google already owns one of the most deeply established information-seeking behaviors in the world:
Google something.
AI Mode does not need to create that behavior from zero.
It needs to extend it.
The consumer who previously searched:
“best reliable compact SUV”
can now have a conversation around the same decision.
That makes Google AI Mode unusually important for product recommendation because it combines Google’s enormous Search distribution with the synthesis and conversational refinement associated with large language models.
Google’s own adoption figure—more than one billion monthly AI Mode users—and the continued growth of AI Mode make the platform difficult for brands to ignore. Google — A new era for AI Search
Gemini Is Worth Investing in Before It Fully Reaches ChatGPT Scale
Gemini deserves a slightly different argument.
Its standalone mobile and web audiences remain meaningfully smaller than ChatGPT’s in the comparable Similarweb datasets.
But Gemini has something that Claude and Perplexity do not have at the same scale:
Google’s distribution ecosystem.
Consumers are already becoming familiar with Gemini intelligence through Google Search and the wider Google ecosystem. Google explicitly ties its broader AI products and Search experiences to the Gemini model family. Alphabet — June 2026 Investor Presentation
As familiarity with AI grows, the behavior itself becomes less novel.
Someone who becomes accustomed to asking AI Mode increasingly complex questions may eventually be perfectly comfortable opening Gemini directly.
Someone who uses ChatGPT regularly may become more willing to try a second assistant.
And someone who encounters Gemini throughout Google’s ecosystem may eventually form a direct habit around the standalone Gemini product.
So even if Gemini is not yet at ChatGPT’s standalone consumer scale, companies should probably treat it as an investment ahead of the curve rather than something to revisit later.
Claude and Perplexity Matter, but They Are Secondary Mass-Market Priorities Today
Claude and Perplexity should not be dismissed.
They simply occupy different positions today.
Anthropic’s own usage research shows Claude remains particularly concentrated around professional and technical tasks, with coding representing a significant portion of usage. Its Economic Index also distinguishes consumer Claude conversations from more programmatic API activity. Anthropic — Economic Index
Perplexity has deliberately positioned itself around research, web retrieval, citations, and answer-engine behavior. It describes itself as an answer engine built around finding and synthesizing information from the web. Perplexity — Getting Started
For certain categories, these audiences may be disproportionately valuable.
A technical B2B product could care significantly about Claude.
A complex financial, technology, or research-heavy purchase could make Perplexity particularly relevant.
But from the perspective of general mass-market AI product recommendation, their current consumer mobile and web audiences remain far below ChatGPT, AI Mode, and Gemini.
That makes them secondary priorities today—not irrelevant ones.
One AIPR Framework
Companies should build one AIPR Framework, not five completely separate strategies
Prioritizing certain systems does not mean building an entirely different strategy for each one.
ChatGPT, Gemini, Google AI Search, Claude, and Perplexity differ in architecture, product design, retrieval systems, source selection, personalization, and model behavior.
But from the perspective of a company trying to improve its probability of being recommended, many of the underlying strategic inputs overlap.
The systems need knowledge about the company and its products.
They need evidence supporting relevant product claims.
They may rely on parametric knowledge learned from training.
They may retrieve current information from the web.
They need sources that corroborate claims.
They encounter reviews, publishers, communities, company websites, product documentation, retailers, expert sources, and other parts of the information environment.
And ultimately, they have to connect a user’s decision narrative with candidate products that appear to satisfy it.
That means the long-term objective should not be:
“Create a ChatGPT strategy, then separately create a Gemini strategy, then separately create a Claude strategy.”
It should be:
Build an AI Product Recommendation framework that improves the company’s information environment and recommendation position across AI systems, while measuring and adapting to the behavior of each major platform individually.
What Companies Should Focus on Today
If the question is where companies should concentrate their AIPR investment today, the current evidence suggests a relatively clear hierarchy.
ChatGPT and Google AI Mode should receive the greatest attention.
They already combine extraordinary human adoption with exactly the kinds of conversational information-seeking behaviors that can influence product decisions.
Gemini should also be a strategic priority.
Its standalone audience is smaller than ChatGPT’s today, but its growth and integration into Google’s broader AI ecosystem give it considerable upside.
Claude and Perplexity should remain part of the measurement and optimization framework, particularly in industries where their audiences or use cases are disproportionately relevant, but their present mass-market consumer reach makes them secondary priorities.
Google AI Overviews should be monitored separately as an enormous embedded AI exposure layer inside Search rather than treated as equivalent to an intentionally used conversational assistant.
The exact model names underneath all of these products will continue changing. The more durable strategic question is simpler:
Where are humans increasingly turning to AI to ask questions, evaluate products, refine decisions, and decide what to buy?
Today, the strongest answers are ChatGPT and Google AI, with Gemini increasingly important as a direct consumer destination.
For companies building an AI product recommendation strategy now, that is where the center of gravity should be.
Leave a Reply