Generative AI is quickly reshaping how people search for information.
A recent Statistica and Semrush .Trends report indicates that the generative AI market was valued at $67 billion in 2024 and is expected to grow annually by 24.4% through 2030.
This growth is making waves in the search engine sector, where platforms like ChatGPT enable more conversational and context-aware search experiences.
In a previous SearchGPT study, we found that when compared to Google and Bing, SearchGPT tended to return smaller domains in its results, hinting at a shift in how users could discover new content.
Note: OpenAI has since retired “SearchGPT” in favor of ChatGPT’s integrated search feature, which we’ll refer to as ChatGPT or ChatGPT Search throughout this article.
Though traditional search engines like Google still receive the bulk of search traffic, ChatGPT introduces a fresh approach to searching for information.
These changing behaviors inspired our investigation into ChatGPT’s impact on website traffic and digital marketing. By analyzing millions of clickstream records, we explored how AI-driven search affects both traffic distribution and user navigation patterns.
Search industry expert Kevin Indig puts these changes in perspective:
“The question is if you want to invest now for a potential first-mover advantage. I think it will pay off like it did with SEO, but if you’re looking for quick results, just focus on doing good SEO.
Tech SEO for LLM Crawlers, listicles, and driving more mentions across the web help with Generative AI visibility because LLMs still use search results from Google or Bing to ground their answers. But for how much longer? I think LLMs will soon develop their own weighting algorithms.”
Methodology
Semrush analyzed more than 80 million lines of global clickstream data from the second half of 2024 to identify how people interacted with ChatGPT.
Clickstream data is the digital trail of a user’s activity as they navigate through websites, apps, or other digital platforms. This captures detailed information about user interactions, such as the sequence of pages visited, time spent on each page, and referral sources.
With this data, we were able to see:
- The specific ChatGPT prompts people queried
- Where they would navigate to after interacting with ChatGPT
- Whether or not ChatGPT’s “Search” feature was enabled during the interaction
From there, we examined the average length and purpose of the queries along with the types of websites ChatGPT directed users to. We also used clickstream data from our .Trends platform to compare audience demographics between ChatGPT.com and Google.com.
Growth of ChatGPT as a Traffic Referrer
Understanding how users navigate from ChatGPT to other websites reveals its growing influence on internet traffic. In early July, ChatGPT sent traffic to less than 10,000 unique domains in a single day.
This number expanded dramatically by November, reaching over 30,000 unique domains in a day – an increase that demonstrates ChatGPT’s rapidly expanding role in directing users across the internet.
Number of Unique Domains Receiving Referral Traffic from ChatGPT
This chart shows the growth in the number of unique domains that received referral traffic from ChatGPT over time.
Each point on the chart reflects how many distinct websites were referred traffic from ChatGPT on a given day. The upward trend demonstrates that more domains are benefiting from referrals over time, indicating a broader distribution of traffic across the web.
Interestingly, this number spiked in August—months before the October 31st search feature launch. This early surge suggests that users were already exploring ChatGPT’s capabilities to discover and navigate content, even before the enhanced search functionality was available.
As ChatGPT’s search capabilities continue to develop, we may see even greater shifts in how users discover and engage with online content.
Analysis of ChatGPT Searches
How do people interact with ChatGPT compared to traditional search engines like Google?
This section explores the prompts people submit to ChatGPT and compares them with traditional search engine queries—like those on Google.
Are users experimenting with more creative requests or asking questions that simply don’t fit the standard keyword approach?
Understanding these interactions helps us grasp how generative AI is shifting search behaviors, potentially influencing future developments in search technology and digital marketing strategies.
SearchGPT On vs Off
Currently, ChatGPT operates without SearchGPT by default, handling prompts and generating responses solely based on its internal knowledge. However, users have the option to enable web searches by clicking the globe icon when they want ChatGPT to pull in information from the internet.
There are also instances where ChatGPT will automatically activate the web search feature for certain types of queries. When this happens, a “Searching the web” notification appears while ChatGPT retrieves the information.
In our analysis, we found that about 54% of ChatGPT queries were handled with SearchGPT turned off, while 46% utilized the web search feature.
This split highlights that users are engaging with ChatGPT in both modes—leveraging its built-in capabilities and its ability to access live web data.
Let’s take a closer look at these usage patterns.
Length of ChatGPT Prompts
When users interact with ChatGPT without the SearchGPT feature, the prompts tend to be more detailed. During this period, we observed an average prompt length of 23 words, with some reaching up to 2,717 words.
Conversely, when SearchGPT is enabled, the prompt length drops significantly to an average of just 4.2 words, with the longest prompt capped at 301 words.
This shorter query length aligns more closely with traditional Google searches, suggesting that users switch to a more direct search mode, likely seeking specific, concise answers.
These distinguishable prompt lengths reflect ChatGPT’s dual use: longer prompts for in-depth discussions or instructions and shorter ones for quick searches.
Search Intent Analysis
Beyond prompt length, understanding why users turn to ChatGPT—their search intent—reveals deeper patterns in how people use AI to find information and solve problems. For marketers and content creators, these patterns guide content strategy and user experience design.
Traditional search engines like Google typically serve four types of intent:
- Navigational (finding specific sites)
- Informational (learning about topics)
- Commercial (researching products)
- Transactional (making purchases)
But ChatGPT is breaking this mold. Only 30% of ChatGPT prompts in our dataset could be identified as one of these traditional categories. The other 70% consisted of unique queries rarely seen in standard search engines—suggesting users are finding new ways to solve problems and gather information.
This indicates that many ChatGPT queries represent entirely new types of intent, potentially related to problem-solving, brainstorming, or exploratory inquiries that don’t fit neatly into traditional categories.
But before we explore the broader implications of ChatGPT’s unique queries, let’s focus on the 30% of prompts we could categorize.
Search Intent on Google vs ChatGPT
Our analysis compared search intent between Google and ChatGPT (the 30% that we could label by intent)—examining what users were trying to accomplish with their queries on each platform. The data showed that informational and navigational intents dominate both Google and ChatGPT.
Note: The search intent of only 30% of the ChatGPT queries we studied is known and displayed in the chart on the right. The other 70% of queries are not searched enough on Google to be found in the Semrush database.
However, Google exhibited a higher proportion of navigational intent, while informational intent was more prevalent in the identifiable ChatGPT prompts.
To understand these differences further, we explored how these trends varied when ChatGPT’s SearchGPT feature was either enabled or disabled.
Search Intent on ChatGPT: SearchGPT On vs Off
When examining how the activation of ChatGPT’s SearchGPT feature affects search intent, we noticed several differences.
With SearchGPT enabled, the intent distribution more closely mirrored that of Google, with a noticeable increase in commercial and transactional searches.
With SearchGPT disabled, ChatGPT prompts leaned heavily towards informational intent. However, it’s likely that many fall into the “unknown” intent category due to their longer, more detailed nature.
The “Unknown” Intent of ChatGPT: What It Means for SEOs and Marketers
Our research underscores a significant difference in how users engage with ChatGPT compared to traditional search engines like Google. While we could categorize the search intent of 30% of ChatGPT queries that matched terms in the Semrush keyword database, the remaining 70% had “unknown intent.”
This vast disparity indicates that many ChatGPT queries are pioneering new ways of seeking information and interacting online.
This variation stems from the fact that ChatGPT is not just another tool for finding websites or straightforward answers. Instead, it positions itself as a dynamic and creative resource.
Take a look at the suggestions highlighted on its main page:
- Create image
- Make a plan
- Surprise me
- Summarize text
- Get advice
- Help me write
- Brainstorm
- Analyze images
These capabilities suggest uses that go beyond the conventional search intent categories—transactional, informational, commercial, and navigational—that SEOs typically consider. Rather, ChatGPT introduces novel intents focused on creation, conversation, and critical thinking facilitated by AI.
This evolution in user interaction may require SEOs and marketers to rethink their strategies and consider these unique intents in their digital marketing plans.
Industry experts are beginning to explore potential strategies for this shift. Chris Long, VP of Marketing at Go Fish Digital and Creator of The Splash Newsletter, shares his perspective on optimizing for Large Language Model (LLM) driven search:
As the search landscape continues to be decentralized – optimizing for LLM-driven is going to be a must-have skill for any SEO in the future. LLMs are able to cater to a user’s extremely personalized needs, thus making brand mentions and traffic from these platforms invaluable.
For OpenAI, SEOs will need to understand the types of queries that trigger ChatGPT Search. What we’ve found is that listicle articles tend to perform quite well for brand mentions. Creating your own article can give your brand power to influence the response.
And Michael King emphasized that “conversational search” is more of a branding channel than a performance channel.
We absolutely need to optimize for SearchGPT, but we don’t need to measure it by the amount of referral traffic it drives as though it’s a performance channel. I think our community, or whatever community emerges to support Conversational Search channels, will realize that these are branding channels. There is value for brands to show up here so they gain and maintain mindshare. It’s my hope that those of us doing marketing here will not undervalue it in the way that those that started SEO devalued Organic Search.
While the full impact of these search patterns is still emerging, we are starting to see patterns in how ChatGPT is reshaping traffic distribution. Let’s take a look at which domains and industries are currently affected by this shift.
Analyzing Traffic Preferences: Which Sites Gain More from ChatGPT Than Traditional Search Engines?
Identifying which websites receive more traffic from ChatGPT compared to traditional search engines like Google and Bing can help pinpoint shifts in user behavior and uncover new opportunities emerging from AI-driven search trends.
From our recent study on our Evolution of Online Search study, we identified several domains that receive significant traffic from ChatGPT.
Interestingly, popular sites such as YouTube and Facebook also rank high in traffic from Google.
This isn’t surprising, as they rank among the internet’s most visited domains and receive substantial traffic from other search engines as well.
However, our analysis focused on domains that receive more traffic from ChatGPT than from traditional search engines.
Here’s what the data showed us:
Websites Receiving More Traffic from ChatGPT Than Google
The data reveals that domains related to technology are experiencing a surge in traffic from ChatGPT.
OpenAI-related domains (oaiusercontent.com) receive nearly 14x more traffic from ChatGPT than Google. In addition, other tech and AI-focused platforms like aiprm.com and gptinf.com also show significantly higher ChatGPT traffic.
Note: openaiusercontent.com is a domain used by OpenAI to serve user-generated content (generated images, documents, etc), which explains the high level of traffic.
But how does this pattern compare to Bing? Let’s take a look.
Websites that Receive More Traffic from ChatGPT than Bing
When we look at traffic from ChatGPT versus Bing, tech domains lead in raw traffic numbers. However, a clear trend emerges in academic and educational sites receiving more traffic from ChatGPT than Bing:
Academic Publishers & Research
- Major publishers (Springer, MDPI, OUP)
- Research institutions (WHO, World Bank)
Education & Technical Resources
- Learning platforms (Instructure, Blackboard)
- Universities (CUNY, UCI)
- Technical documentation (Python.org)
This data suggests ChatGPT has become a prominent traffic source for academic and educational content. But to understand the full scope of ChatGPT’s impact across the internet, let’s examine how traffic is distributed across all major industries.
Top Industries Receiving Traffic from ChatGPT
A closer look at ChatGPT’s referral traffic reveals which sectors draw the most user engagement.
Our analysis examined the 10,000 most-visited domains after ChatGPT interactions, categorizing each by industry. The data represents unique sessions—meaning individual visits that ended on websites within each category.
Online services dominate with nearly 10M sessions, but educational content shows a strong presence—education and distance learning together account for over 9M sessions.
Information technology and software development similarly stand out, driving 7.8M combined sessions.
These numbers suggest ChatGPT frequently guides users to platforms offering practical tools and learning resources.
Traditional sectors like retail (1.9M), finance (1.6M), and healthcare (1.5M) show notable but lower traffic volumes.
But what’s driving these patterns? To understand this, we need to look more closely at who is using ChatGPT.
Audience Demographics: Who is Using ChatGPT and Google?
Understanding user demographics helps reveal how these platforms serve different segments of the online population.
Scale and Reach
According to our Semrush .Trends data, Google reached 6.5B unique worldwide visitors in December 2024, while ChatGPT drew 566M visitors. These numbers show that while Google remains dominant, ChatGPT has built a substantial user base in its first year.
Demographics at a Glance
The demographic data reveals distinct patterns between the platforms. While Google shows broad representation across age groups and occupations, ChatGPT’s user base shows higher concentrations among certain segments.
Age and Gender
ChatGPT’s user base skews younger and more male compared to Google.
Occupation
ChatGPT’s audience is skewed more towards students.
While Google shows higher representation among:
- Full-time workers
- Homemakers
- Retirees
These demographic patterns help explain the traffic trends we observed earlier, particularly the high volume of traffic to education and distance learning sites.
This suggests an opportunity for universities and academic institutions. With the right strategy, these institutions could expand their reach through ChatGPT more than through traditional search alone.
What This Means for Your Digital Strategy
ChatGPT is changing how users interact with online content, shifting from passive discovery to active creation and refinement of ideas. And having a clear brand with information that’s accessible and retrievable for LLMs will be more important than ever.
As SEOs we do not need to abandon the tactics we’ve always relied on but we do need to evolve them. Where in the past we looked at crawlability, indexability and stop at rankability, we now need to add in Retrievability. That is, taking extra steps to make sure that core information about our brands is available, accessible and prioritized for LLMs. What works is having content in search engines, websites and knowledge graph entities that form part of the training data.
Our analysis of 80 million clickstream records, combined with demographic data and traffic patterns, reveals three key changes in online content discovery:
- Traffic Distribution: ChatGPT drives notable traffic to educational resources, academic publishers, and technical documentation, particularly compared to Bing.
- Query Behavior: While 30% of queries match traditional search patterns, 70% are unique to ChatGPT. Without search enabled, users write longer, more detailed prompts (averaging 23 words versus 4.2 with search).
- User Base: ChatGPT shows higher representation among students and younger users compared to Google’s broader demographic distribution.
For marketers and content creators, this data reveals an emerging reality: success in this new landscape requires a shift from traditional SEO metrics toward content that actively supports learning, problem-solving, and creative tasks.
Organizations that adapt their content strategy to serve these “creation-first” intents will be better positioned to connect with audiences, particularly younger users and students.