AI assistants are changing how customers find businesses online, and the requests they send can place a heavier load on websites than visits from people, according to Fastly.
Fastly said more than 51% of AI requests on its network need to reach a website’s original server to retrieve information. Fewer than 9% of human requests do the same. A page visited by a person can often be served from a stored copy nearby, but an AI assistant may need fresh details from the source.
For businesses, that creates two practical questions: Can their websites handle more automated requests, and will an AI assistant find the right information when a potential customer asks about a product or service?
“While the National Fiber Backbone expansion is critical to strengthening the country’s digital infrastructure, readiness is not only about laying more fiber,” said Rachel Ler, area vice president for ASEAN, Greater China, and South Korea at Fastly. “It is also about whether our businesses, government platforms, and digital services are designed to handle a growing wave of machine-to-machine traffic without slowing down the human experience.”
Fastly said AI requests on its network grew roughly 30% from January to May 2026, about 6.5 times faster than human traffic. Most of those AI bot requests came from crawlers that scan websites. In May, crawlers accounted for about 85% of AI bot requests, while “fetchers” made up the remaining 15%.
Fetchers retrieve information in response to a user’s request to an AI assistant. Someone comparing providers, for example, might ask an assistant about prices or product features instead of visiting several websites. The assistant may then fetch the answers from those businesses.
That makes automated traffic a business issue as well as a technical one. Blocking every bot could prevent some AI services from retrieving information that helps customers discover a company. Allowing every bot through could expose the website to excessive traffic, scraping, or abuse.
“The question is no longer simply whether to block bots. Organizations need to understand which machine interactions should be accelerated, managed, challenged, or stopped,” Ler said.
She said businesses should first identify which AI systems visit their websites, how often they make requests, and what information they seek. They can then decide which activity is useful and which needs closer scrutiny.
Businesses also need to keep information such as prices, availability, specifications, opening hours, and policies up to date. An AI assistant answering a customer’s question is only as useful as the information it can retrieve.
“If an AI assistant retrieves outdated information, it can give a customer the wrong answer even when the company’s own systems are up to date,” Ler said.
Fastly also noted caching, which stores information closer to those requesting it, as one way to serve suitable content faster and reduce repeat requests to a company’s original server. The aim is to make reliable information available to AI assistants while keeping websites responsive for customers who visit directly.