What is AI agents?

Definition: AI agents are AI systems that carry out multi-step tasks on a user's behalf, such as browsing websites, comparing options, filling in forms or making bookings, rather than only answering a question.

From answering to acting

A chatbot answers a question. An agent is given a goal, such as “find a dentist near me who takes my insurance and book the earliest slot”, and works through the steps itself: searching, opening sites, reading prices, filling in forms. Several AI providers now offer agents of this kind inside their assistants and browsers.

How agents visit your site

Agents reach websites in two ways. Some fetch pages over HTTP and read the HTML, much like a crawler. Others drive a real browser, so they run JavaScript, scroll and click. Some identify themselves with user-triggered agents such as ChatGPT-User or Claude-User; others present as an ordinary browser, which makes them hard to tell apart from people in your analytics.

What makes a site easy for agents

Most of what helps agents is accessibility and clean front-end work:

Real links and buttons rather than clickable div elements.
Form fields with visible labels and sensible input types.
Prices, stock, opening hours and policies stated as text, not only in images.
Structured data for products, offers and organization details.
No CAPTCHA or bot challenge on pages that only display information.

Chrome’s Lighthouse has started adding audits aimed at agentic browsing, including a check on llms.txt (Chrome for Developers). Our view: if a screen reader user can complete a task on your site, an agent probably can too.

Allow or block?

Agents act for real customers, so blocking them can mean losing a booking or a sale that a person started. The cases for restricting them are narrower: protecting account areas, preventing scraping of priced inventory, or limiting automated form submissions. The website audit flags crawlability and markup issues that trip up agents as well as search engines.