SALT agency’s latest analysis challenges the idea that AI-powered answers lead to more engaged visitors on websites. Dan Taylor, head of Technical SEO at SALT, compiled Google Analytics 4 data from dozens of websites covering the period from January to March 2025. The data compares around 671,000 sessions referred by large language model interfaces (ChatGPT, Perplexity, etc.) with about 188 million organic search sessions. Taylor introduced a “Key Event Conversion Rate” (KECVR) – the percentage of visits that triggered a site’s key actions – to compare traffic sources. The results show that, in most industries, LLM referrals convert at much lower rates than traditional search, challenging claims by Google and Microsoft Bing about AI-generated clicks bring higher-intent users. A few exceptions, including health, careers, and catalog sites, saw small advantages for LLM traffic, but overall, organic search performed better.
The study spanned 40 industry categories, including consumer e-commerce, travel, SaaS, and B2B. The analysis revealed that organic search traffic generated about 62.2 million key events, while LLM referrals accounted for 214,600. Taylor accounted for a known GA4 quirk, excluding data sets with “key event bloat,” where the number of recorded key events exceeds visits. This exclusion ensured accurate results and helped reveal broader trends.
Overall, organic search visits showed much higher KECVR than LLM-driven visits. For example, in consumer e-commerce, organic search had a 24.1% key-event rate, while LLM referrals only achieved 17.6%. In travel, search traffic converted at 28.9%, compared to 24.3% for AI-driven visits. Even in sectors with low conversion rates, search still performed better: B2B e-commerce buyers showed a 2.7% KECVR from search versus 2.6% from LLMs, and SaaS products saw near-identical engagement from both sources, around 6.7%. In nearly every major industry, from retail to education, HR to luxury, organic search outperformed LLM traffic in generating conversions.
A few sectors did see small benefits for LLM traffic. Career and job sites saw a 22.3% KECVR from LLMs, compared to 16.6% from search, while health sites recorded 13.2% from LLM referrals versus 12.9% from search. Catalog sites also had a slight advantage for LLMs (2.3% vs. 2.1%), but these were the exceptions. Across most sectors, organic search still yielded higher KECVR than LLM traffic.
The findings contradict Google and Bing's claims that AI-driven search would increase high-intent traffic. The study suggests that LLM-based answers primarily serve as research or informational tools, rather than driving users to click through to websites. AI chats often provide quick answers or summaries, reducing the need for users to explore further. In contrast, users of traditional search engines are often looking for more in-depth information, such as product reviews, price comparisons, or detailed service offerings, leading them to visit websites more frequently. For instance, e-commerce shoppers tend to use search for product specs and reviews, while LLMs provide more general answers that do not necessarily drive purchasing decisions. Similarly, travel queries may begin with AI suggestions, but users often finalize bookings through richer organic content and customer photos.
The analysis also highlights the rapid growth of LLM referral traffic. Over the past year, the share of visits from LLMs has steadily increased, reflecting growing adoption of AI tools. ChatGPT has become the dominant source of LLM-driven referrals, particularly since OpenAI began routing users through chatgpt.com. Perplexity has also grown, though it remains behind ChatGPT. Even Meta’s new AI features have not yet sent significant traffic, but this may change as they roll out more widely. Despite the increase, LLM referrals still account for a small fraction of overall traffic compared to search engines, further emphasizing the importance of organic search.
All in all, the SALT study paints a clear picture showing that AI-driven search is not yet delivering the high-engagement audience that was promised. The Key Event Conversion Rate metric shows that visitors from AI tools like ChatGPT and Perplexity are generally less likely to complete key actions than those coming from traditional search engines. Marketers should note that traditional SEO remains far more effective at attracting engaged users. While LLM referrals can boost traffic and serve informational purposes, this study suggests they should complement, rather than replace, strong organic content if the goal is conversion.
Image: DIW-Aigen
Read next:
• AI Search Engines Favor Product Content, Dismissing Blogs and PR Materials, New Study Reveals
The study spanned 40 industry categories, including consumer e-commerce, travel, SaaS, and B2B. The analysis revealed that organic search traffic generated about 62.2 million key events, while LLM referrals accounted for 214,600. Taylor accounted for a known GA4 quirk, excluding data sets with “key event bloat,” where the number of recorded key events exceeds visits. This exclusion ensured accurate results and helped reveal broader trends.
Overall, organic search visits showed much higher KECVR than LLM-driven visits. For example, in consumer e-commerce, organic search had a 24.1% key-event rate, while LLM referrals only achieved 17.6%. In travel, search traffic converted at 28.9%, compared to 24.3% for AI-driven visits. Even in sectors with low conversion rates, search still performed better: B2B e-commerce buyers showed a 2.7% KECVR from search versus 2.6% from LLMs, and SaaS products saw near-identical engagement from both sources, around 6.7%. In nearly every major industry, from retail to education, HR to luxury, organic search outperformed LLM traffic in generating conversions.
A few sectors did see small benefits for LLM traffic. Career and job sites saw a 22.3% KECVR from LLMs, compared to 16.6% from search, while health sites recorded 13.2% from LLM referrals versus 12.9% from search. Catalog sites also had a slight advantage for LLMs (2.3% vs. 2.1%), but these were the exceptions. Across most sectors, organic search still yielded higher KECVR than LLM traffic.
The findings contradict Google and Bing's claims that AI-driven search would increase high-intent traffic. The study suggests that LLM-based answers primarily serve as research or informational tools, rather than driving users to click through to websites. AI chats often provide quick answers or summaries, reducing the need for users to explore further. In contrast, users of traditional search engines are often looking for more in-depth information, such as product reviews, price comparisons, or detailed service offerings, leading them to visit websites more frequently. For instance, e-commerce shoppers tend to use search for product specs and reviews, while LLMs provide more general answers that do not necessarily drive purchasing decisions. Similarly, travel queries may begin with AI suggestions, but users often finalize bookings through richer organic content and customer photos.
The analysis also highlights the rapid growth of LLM referral traffic. Over the past year, the share of visits from LLMs has steadily increased, reflecting growing adoption of AI tools. ChatGPT has become the dominant source of LLM-driven referrals, particularly since OpenAI began routing users through chatgpt.com. Perplexity has also grown, though it remains behind ChatGPT. Even Meta’s new AI features have not yet sent significant traffic, but this may change as they roll out more widely. Despite the increase, LLM referrals still account for a small fraction of overall traffic compared to search engines, further emphasizing the importance of organic search.
All in all, the SALT study paints a clear picture showing that AI-driven search is not yet delivering the high-engagement audience that was promised. The Key Event Conversion Rate metric shows that visitors from AI tools like ChatGPT and Perplexity are generally less likely to complete key actions than those coming from traditional search engines. Marketers should note that traditional SEO remains far more effective at attracting engaged users. While LLM referrals can boost traffic and serve informational purposes, this study suggests they should complement, rather than replace, strong organic content if the goal is conversion.
Read next:
• AI Search Engines Favor Product Content, Dismissing Blogs and PR Materials, New Study Reveals
