Agentic AI traffic surges as bots dominate the web

Automated bots now generate most web traffic, and agentic AI traffic has risen 7,851% year on year, according to Decodo. That shift creates a new dilemma for businesses that block bots by default.

Its analysis found bots accounted for 57.4% of all web requests, compared with 42.6% from human users. AI-driven traffic grew by 187% during 2025, about eight times faster than human traffic over the same period.

The figures suggest businesses may need to rethink how they treat automated visitors. Security teams have long viewed bots mainly as a source of fraud, scraping and account attacks. Now, AI assistants are also visiting sites to gather information for users researching products, comparing prices and seeking answers before making a purchase.

Commercial pages

According to the analysis, 77% of agentic AI activity reached product and search pages in 2025. Another 8.8% reached account pages, 5% accessed authentication flows and 2.3% reached checkout pages.

That puts automated systems on some of the most commercially sensitive parts of websites, especially for retailers whose pricing, inventory and listings change frequently. For online merchants, the question is no longer just how to stop harmful traffic, but how to avoid shutting out automated systems that may influence what customers see through AI tools.

Vaidotas Juknys, Chief Executive Officer of Decodo, said: "The instinct to block every bot is understandable, but it's like locking your storefront because some visitors don't buy. The agents crawling your site today might be how your customers will discover you tomorrow."

Retail and eCommerce sites featured heavily in the findings. They were the largest commercial category for successful automated requests in Decodo's anonymised network data, accounting for about 13% of total traffic analysed.

Search engines and AI answer engines held the largest overall share at about 72%, underlining the scale of automated systems involved in finding, processing and presenting information online. By contrast, travel, airlines and cargo accounted for about 0.4% of successful requests, while news and media, real estate, and jobs and professional data each represented about 0.2%.

Separate analysis cited by Decodo found retail and eCommerce accounted for 62.5% of training-crawler traffic in 2025. Combined with the concentration of agentic AI activity on product and search pages, that suggests retail websites are becoming a key source of information for both customer-facing AI tools and the systems used to train them.

Geographic split

The United States was the largest source of automated web traffic in the analysis, accounting for 53.5% of global bot traffic. Germany ranked second with 8.2%, followed by the Netherlands with 5.6%, Singapore with 5.4%, and France and China with 3.9% each.

The share of traffic generated by bots also varied sharply by country. Bots made up 43.6% of traffic in the United States and 45% in Germany, rising to 61.3% in the Netherlands and 73.7% in Singapore.

The United Kingdom accounted for 3.3% of global bot traffic in the ranking. Other countries in the top 20 were India, Ireland, Brazil, Japan, the Russian Federation, Iran, Hong Kong, Indonesia, Canada, Vietnam, South Korea, Finland and Australia.

Filtering traffic

The findings come as companies face pressure from several directions at once. Cybersecurity teams are trying to contain malicious scraping, credential stuffing and fraud, while marketing and eCommerce teams increasingly need their sites to remain visible to AI systems that may shape product discovery and recommendation.

Decodo recommended distinguishing between different forms of automated traffic rather than treating all bots the same. It said businesses should assess requests by behaviour and verified identity, allow crawlers that support discovery, limit unknown traffic and stop patterns linked to attacks, fraud and malicious scraping.

The shift also affects analytics. If automated and human activity are not separated, businesses risk drawing the wrong conclusions about website demand, product interest and customer behaviour.

For retailers in particular, the issue is as practical as it is technical. If AI assistants cannot access current product information, stock levels or pricing, businesses may become less visible when consumers rely on automated tools to compare options and narrow buying decisions.