2025 in Review: How AI Web Referrals Are Reshaping Traffic Acquisition 

Over the course of its ascent, AI has uprooted brands across industries, forcing companies to rethink how they approach customer acquisition. For some, this is a positive; in many cases, AI is able to streamline processes, increase personalization, and ultimately improve outcomes. But for other stakeholders, AI is a challenge to be dealt with. By now, everyone has seen the headlines boldly proclaiming that LLMs are replacing traditional search engines — and much of the evidence does seem to point that way. 

In 2025, in the United States, ChatGPT saw greater net growth in both unique visitors and total overall visits than Google. While Google’s performance still dwarfs ChatGPT’s by most metrics (it had more than four times the traffic and triple the hours spent last year), the rate at which OpenAI’s brainchild is expanding suggests that this hierarchy could change in the future. Given the surging importance of LLM referrals in driving web traffic, determining the best way to capture these visits has become something of a hot topic.

To help frame the discussion, we analyzed the sites that gained the most AI traffic in 2025, hoping to shed some light on the characteristics that set LLM-favored destinations apart from other high-performers. 

Methodology: Comparing AI-Favored Sites to the Web’s Most Popular Destinations 

This blog focuses on the 25 sites with the highest year-over-year (YoY) gains in AI-referred visits, zeroing in on the United States. For the purposes of this analysis, our control group is the 25 sites with the highest traffic overall throughout 2025. This comparison provides a baseline for how AI-favored destinations compare to the web’s largest and most established players, distinguishing characteristics associated with AI visibility from those driven primarily by scale. 

Notably, more than half of the sites with the largest AI traffic gains also rank among the most-visited, suggesting that AI referrals last year largely amplified existing web leaders rather than reshuffling the competitive landscape. Nevertheless, eleven sites appeared in the AI-favored group without being present in the control — meaning that it is possible for non-incumbents to harness LLMs for online visibility they may not have otherwise had. 

Which Website Categories Are Gaining the Most AI-Referred Traffic?

AI traffic is not evenly distributed across the web, and while it tends to favor many of the same dominant categories that drive overall traffic, it also surfaces some sites that don’t typically appear in raw traffic rankings. In total, five different site categories were represented within the AI-favored group. More than a third (9 of 25) were software platforms, including three Gen AI tools. A fifth (5 of 25) were social media platforms, and five were shopping sites — a distribution broadly consistent with the control group. 

Where the samples diverged was in the appearance of several government, employment, and academic sites among the AI-favored sites. While no government website, job search platform, or academic journal ranked in the top 25 by overall traffic, outlets such as the National Institutes of Health, Indeed, and Science Direct did register among the largest YoY gainers in AI-referred visits. The dataset is not large enough to be conclusive — but it does show the ability of LLMs to spotlight heavily-researched, high-quality material that users may not gravitate to as frequently at their own behest — or when subject to the whims of the Google algorithm. 

How Much Traffic Do LLMs Actually Drive?

On average, LLM-referred traffic made up 2.2% of the AI-favored group’s total visits — in other words, around 1 in every 50 site visits for these brands came from Generative AI, with a median YoY increase of 58.3 million AI visits. In comparison, the control group only saw 0.2% of visits referred from LLMs. AI is still small in absolute terms, but highly concentrated, and disproportionately important for a subset of sites. 

AI-Favored Sites Rely More On Organic Discovery (Not Paid Traffic)

Crucially, the AI-favored group also exhibited a much higher portion of organic traffic compared to the control (45% of total visits vs. 14%). The control group, meanwhile, saw 5% of their traffic come from paid channels, while Gen AI-favored sites received a negligible amount of visits (<0.1%) from paid sources. 

Interestingly, while the AI-favored sites benefited from an average of 2,200% more LLM visits in 2025 than in the year prior, they also saw traffic decreases across every other channel, including direct (-25%), paid (-23%), and organic (-23%). The control group, on the other hand, saw its paid visits increase by 12%, along with a 229% increase in AI visits. 

To recap, the sites that were most favored by LLMs also showed a higher overall propensity to see high levels of organic traffic, and lower levels of paid traffic. This could suggest that the characteristics that drive organic authority (e.g. depth of content, topical relevance) are the same ones that win AI citations. In other words? At least in the current landscape, AI visibility is not something brands can simply throw money at, and expect to win. 

Where AI Traffic Actually Comes From

Taking a closer look at the LLM users actually traveling to these favored destinations, we found that 89% of them were referred from Google’s AI Overview, with 8% coming from ChatGPT. Rufus and Gemini claimed the remainder (~1% each). We saw a similar breakdown for our control group, where Google’s AI Overview dominated even more heavily — accounting for 91% of LLM-referred users. 

Also potentially useful - taking note of the demographic breakdown of the audiences who are using ChatGPT and other LLMs, a group which is naturally overrepresented in the traffic funneling from these platforms. While Google (and its built-in AI Overview) host an even split of men and women, ChatGPT and other Gen AI platforms tend to skew male, with men making up 67% of ChatGPT’s user base. Forty percent are between the ages of 25 and 34, and they are 89% more likely than the general population to use productivity apps. The top-growing app for this population is Google Gemini, with downloads of the Gen AI chatbot increasing 27% YoY.  

Finally, the data on mobile versus web traffic shows that mobile users spent similar amounts of time in both subsets (114 minutes per month for the AI winners and 117 for the control) while a larger difference was seen among web users, who spent an average 499 minutes per month on the control group’s sites, and only 384 on the AI-favored sites. We can cautiously conclude that AI referrals are still web-first, site mix likely matters, and usage intent may differ. 

What AI Web Referrals Mean For Brands in 2026

LLM referrals still make up a small share of overall traffic. But for a subset of sites, AI already accounts for a meaningful (and growing) portion of visits. Many of the biggest AI traffic gainers already rank among the web’s most visited sites, but incumbents can, and do, break through. And, at least preliminarily, AI-favored sites look structurally different from the biggest sites overall, skewing more organic-first, and relying far less heavily on paid traffic. The good news for Google? AI traffic is still largely clustered within its ecosystem. 

OpenAI has recently announced its plan to incorporate advertisements within ChatGPT. As this new sponsored content starts to make its foray into the LLM content market, we’ll be here with the insights you need to understand this brave new world. 

Want to explore the data for yourself? Request a free demo of Web Insights — we’d be happy to show you around.