(Data referenced from Graphite’s research report “AI Is Much Bigger Than You Think”.)
If you’ve been following the AI boom, you’ve probably seen a familiar narrative: AI is rapidly replacing search. Charts comparing ChatGPT traffic to Google searches appear everywhere, and the conclusion often feels obvious. AI goes up, search goes down.
But what if that comparison is fundamentally flawed?
A detailed analysis by Graphite.io suggests something surprising. AI is already enormous, far larger than most public discussions suggest. And at the same time, search is not shrinking the way many people assume. Instead of replacing search, AI appears to be expanding the total amount of information-seeking behavior online.
Once you look at the complete data, including mobile apps, multiple AI tools, and real usage patterns, the picture changes dramatically. AI is not a niche tool anymore. It is already operating at a scale comparable to one of the largest digital behaviors on the internet.
Let’s break down what the data actually shows.
The Scale of AI Is Already Massive
Most people underestimate how big AI usage already is.
According to Graphite’s analysis, AI tools collectively generate 45 billion monthly sessions worldwide and about 5.4 billion monthly sessions in the United States.
To put that into context, the study estimates that AI usage today equals roughly:
- 56% the size of search worldwide
- 34% the size of search in the United States
That alone challenges the common perception that AI is still in its early niche phase. In reality, AI usage has already reached a meaningful fraction of the global search ecosystem.
But this comparison becomes even more interesting once you understand how AI usage is actually measured.
The Biggest Measurement Mistake Everyone Makes
A lot of popular comparisons rely on a simple metric: Google Search web traffic vs ChatGPT website traffic.
That sounds reasonable, but it leaves out several crucial pieces of the puzzle.
First, most AI usage does not happen in the browser. Second, ChatGPT is not the entire AI ecosystem. Third, Google is not the only search engine. And finally, not every AI prompt is actually replacing a search query.
Once you correct for those missing pieces, the scale of AI looks very different.
The biggest factor is mobile usage.
The Mobile App Effect Most People Ignore
If you only measure web traffic, you miss the majority of AI activity.
According to the Graphite analysis, 83% of AI usage worldwide happens inside mobile apps. And, only 17% happens on the web.
In the United States, the split is slightly different but still heavily mobile. In the U.S. 75% of AI usage happens in apps, whereas 25% happens on the web.
This means estimates based only on website traffic undercount AI usage by roughly 4x to 5x.
Think about what that implies. When people compare ChatGPT web visits to Google search traffic, they are leaving out most of the actual AI activity happening globally.
This single factor dramatically changes the size of the market.
AI Is Not Just ChatGPT
Another mistake is assuming that ChatGPT represents the entire AI category.
Graphite’s analysis compares the entire AI ecosystem with the entire search ecosystem, rather than looking at a single product. The study includes the five largest LLM platforms and the six largest search engines.
When these broader markets are included, the numbers shift slightly but meaningfully.
So the real comparison is not Google vs ChatGPT. It is the entire search ecosystem vs the entire AI ecosystem.
Not Every AI Prompt Is a Search
One of the most thoughtful parts of the analysis is how it separates different types of AI usage.
AI prompts are not all the same. Some resemble traditional search queries, while others are more like productivity tasks or personal conversations.
A study by ChatGPT analyzing 1 million anonymized ChatGPT prompts from May 2024 to June 2025 categorized usage into three main types:
- Asking (51.6%): Users requesting information, explanations, or advice.
- Doing (34.6%): Users asking the AI to perform tasks such as writing, coding, summarizing, or extracting information.
- Expressing (13.8%): Conversational prompts involving personal reflection, creativity, or emotional discussion.
Only the Asking category closely resembles traditional search behavior.
When the analysis isolates that portion, AI usage equivalent to search becomes:
- 28% of the size of search worldwide
- 17% of the size of search in the United States
Even that estimate is considered an upper bound, because some “asking” prompts still differ from typical search queries.
Still, the number is large enough to demonstrate that AI has already become a significant channel for information discovery.
Search Is Not Actually Declining
Despite all the attention around AI, search engines are not collapsing.
In fact, the data suggests something surprising.
Search engine usage has been flat rather than declining, and a separate Graphite study published in January 2026 found that traditional search usage has not decreased over the past six years.
There is no clear downward trend in visits to search engines overall.
This means AI is not simply replacing search. Instead, both are coexisting.
Which leads to a more interesting conclusion.
The Internet’s Discovery Market Is Growing
Instead of thinking about AI and search as competitors fighting over a fixed amount of user attention, the analysis suggests something different:
The total discovery market is expanding.
When search engine usage is combined with AI “Asking” prompts, the total volume of search-like activity has increased significantly.
According to the Graphite data:
- Total discovery behavior increased 26% worldwide from Q1 2023 to Q4 2025
- Comparing 2025 to 2024, combined usage increased 26% globally, and 16% in the United States.
In other words, AI is not just redirecting existing search queries. It is also creating entirely new forms of information-seeking behavior.
ChatGPT’s Growing Share of Discovery
Within this growing ecosystem, ChatGPT has emerged as a major player.
After adjusting for the proportion of prompts that resemble search queries, ChatGPT accounts for:
*The data as of Q4 2025
That is a remarkable shift for a product that only launched publicly in late 2022.
At the same time, Google’s share of global, and U.S search traffic has declined modestly:
United States Search Market ShareThis does not mean Google is collapsing. It simply means that for the first time in years, a new discovery platform has grown large enough to register in the market share data.
ChatGPT Still Dominates the AI Market
Even though multiple AI tools exist, ChatGPT remains overwhelmingly dominant.
According to another Graphite data:
- ChatGPT accounts for 89% of worldwide AI usage
- ChatGPT accounts for 86% of AI usage in the United States
*The data as of Q4 2025
Gemini, Perplexity, Grok, and Claude together represent the remaining portion of the market.
So while the AI ecosystem is expanding, ChatGPT is currently the primary gateway for most AI activity.
AI Adoption Is Even Larger Outside the United States
Another surprising finding involves geography.
AI usage worldwide is more than seven times larger than AI usage in the United States.
Many discussions around AI focus heavily on US product launches and Silicon Valley trends. But the data suggests that global adoption is the real driver of scale.
So basically, the AI story is not just happening in the US. It is happening everywhere.
Global Growth Has Slowed, But the US Is Still Accelerating
Interestingly, AI growth patterns differ by region.
According to Graphite, worldwide AI sessions have plateaued since July 2025 across major LLM platforms.
But in the United States, growth continues to accelerate.
The Graphite study also emphasizes that AI usage in the US in December 2025 was 300% higher than in December 2024.
So while global usage may be stabilizing temporarily, regional growth dynamics are still evolving rapidly.
Why Earlier Forecasts Got the Numbers Wrong
Many widely shared forecasts predicted that AI would soon surpass search.
Ironically, Graphite argues those projections may actually underestimate AI, not overestimate it.
The problem lies in flawed assumptions, such as:
- Measuring only web traffic
- Ignoring mobile apps
- Comparing only Google and ChatGPT
- Ignoring other search engines and LLMs
- Assuming search traffic is already declining
- Treating the market as zero-sum
One example highlights how inaccurate some traffic estimates were.
According to Graphite, some forecasts estimated Google search traffic in the US at about:
- 4 billion monthly visits in January 2021
- 3.7 billion visits by July 2024
But Similarweb data suggests the real numbers were far higher:
- 15.3 billion visits in January 2021
- 14.4 billion visits in January 2024
Google also stated in March 2025 that it processes more than 5 trillion searches per year, reinforcing the scale of the existing search ecosystem.
How the Data Was Validated
The analysis relies heavily on Similarweb traffic estimates, which were cross-checked against first-party analytics data from websites using Google Search Console and Google Analytics between September 2024 and November 2025.
The comparison showed a median correlation of 0.86, indicating strong alignment between Similarweb estimates and real traffic.
Additional comparisons also supported the scale estimates. For example:
- Similarweb estimated 852.3 billion Google visits in 2024
- Average pages per visit were 6.7
- This implies roughly 5.7 trillion pageviews annually
That aligns closely with Google’s public statement of over 5 trillion searches per year.
For ChatGPT, OpenAI CEO Sam Altman stated in October 2025 that the product had 800 million weekly active users.
Similarweb estimates around September 2025 showed:
- 133 million weekly users on Android
- 310 million weekly users on iOS
- 459 million monthly unique web visitors
Together, these total roughly 902 million users, although duplication across platforms likely exists.
Why the Study Uses Sessions Instead of Users
The analysis focuses on sessions rather than unique users.
This is because many people use AI tools across multiple devices and platforms. Measuring sessions helps avoid some of the duplication issues that appear when counting users.
In this framework:
- A web visit represents a session ending after 30 minutes of inactivity or at midnight.
- A mobile session represents active usage inside an app.
While these measurements are not perfectly identical, they offer the best available way to estimate total activity levels.
The Bigger Idea: The Zero-Sum Bias
The entire discussion ultimately comes down to a psychological pattern called zero-sum bias.
Humans tend to assume that if something new grows, something old must shrink.
This mindset appeared during the rise of mobile apps when people claimed “the web is dead.” In reality, the web continued to thrive while apps grew alongside it.
The same pattern may be repeating with AI.
The data suggests that AI is not simply replacing search. Instead, both systems are expanding the overall discovery ecosystem.
The Real Takeaway
AI is already operating at an enormous scale.
With 45 billion monthly sessions worldwide, usage equivalent to 56% of global search, and a discovery market that has grown 26% since 2023, AI is no longer an experimental tool.
At the same time, search engines remain massive and stable.
Rather than replacing search, AI appears to be creating new ways for people to ask questions, solve problems, and explore information online.
The discovery landscape is changing.
But the most important insight might be this: The pie isn’t shrinking. It’s getting bigger.