Automated traffic on the web has surpassed human traffic for the first time, and the agents generating that traffic do not read pages the way people do. In early June 2026, Cloudflare reported that automated requests accounted for roughly 57 percent of page requests, compared with 43 percent from humans. A fast-growing share of that automated activity comes from AI agents acting on behalf of users, and those agents depend on a stripped-down model of a page called the accessibility tree — the same structure browsers have built for screen readers for two decades.
The accessibility tree is a short list of interactive elements, their roles, their names, and their states. It does not know what a color is. It looks for well-labeled structure. If a site lacks that structure, agents from ChatGPT, Claude, Gemini, Grok, and Perplexity may not be able to read the page at all. For years, accessibility work served a small, often-overlooked audience and was defended on principle alone. That argument has now expanded: the audience reading through page structure is no longer a rounding error but a majority of web traffic.
Major vendors have taken different approaches to how their agents consume pages. Anthropic’s Computer Use is vision-first, with Claude looking at screenshots and counting pixels to place the cursor, which is flexible but fragile when layouts shift. Its browser extension, Claude in Chrome, leans the other way: its read_page tool takes the accessibility tree first, with a get_page_text tool for article-style pages and screenshots as a fallback. Microsoft’s Playwright MCP, the official server for letting a model drive a browser, works from structured accessibility snapshots and skips vision entirely, on the stated premise that the structured view beats a screenshot. Its consumer surface is the opposite: in Copilot Mode in Edge, Copilot Actions read and act inside a signed-in session, leaning on what is on screen.
Google’s Gemini agents sit in the hybrid middle, blending structured channels with a visual pass to confirm layout. That shows up for users as Gemini in Chrome and its Auto Browse mode, with Project Mariner as the research agent behind it. The most telling case is OpenAI. Its Computer-Using Agent, which drives both Operator and the Atlas browser, started from screenshots. OpenAI’s own publisher guidance now tells site owners that Atlas uses ARIA tags — the same labels and roles that support screen readers — to interpret page structure, and points them to standard accessibility practices. Even the company whose agent leads with pixels is telling site owners to fix their accessibility tree.
Accessibility does not matter equally on every page. For a page that is mostly prose, an agent can usually cope; the words are the content, and even a messy structure leaves them readable. The stakes rise the moment an agent has to act — to find the right button, fill the correct field, choose among options, and move through a flow without losing its place. Transactional, multi-step, and dynamic interfaces are where a broken structure stops being an inconvenience and starts being a brick wall. If a keyboard user cannot finish the task, the agent usually cannot either.
The evidence is getting concrete. In the A11y-CUA study presented at CHI 2026, researchers at UC Berkeley and the University of Michigan ran Claude Sonnet 4.5 through 60 everyday web and desktop tasks under different conditions. With normal access, the agent finished about 78 percent of them. Constrained to keyboard-only navigation — the path an accessibility-tree agent depends on — success fell to 42 percent, and the work took roughly twice as long. Under a magnified viewport, it dropped to 28 percent. The same structural weaknesses that block assistive technology users are now blocking the agents that a growing share of web traffic depends on.
The practical takeaway is that inclusive design has become load-bearing for the agent era. The questions worth asking are which agents read what, when accessibility becomes foundational, and which small set of categories is worth fixing first. The semantic structure that screen readers have used for decades now determines whether the major AI agents can read a page at all.