Vibeleaderboard
← All Intel
Intel / article

An AI agent runs ten web errands a day, every transcript published

Source
agentability.org
Author
khalidsaidi
Date

Context

Agentability is a public experiment that tests how well an can get answers from ordinary websites using only plain page requests. According to the site, each day an AI producer turns search trends into ten errands, and a second model, from the DeepSeek family, tries them with read-only requests: no JavaScript, logins or forms. It applies no retries or edits and publishes every transcript.

For its 10 October 2026 episode the site reports that the agent finished all ten errands but met nine bot walls, meaning blocks aimed at automated visitors, while reading 131 pages across 47 sites. A weekly audit of the 1,215 most visited sites gives an average readiness score of 51 out of 100, with about 13 percent publishing an file and 27 percent blocking at least one AI crawler.

These are the site's own figures from its own scoring. Its top rank is a 25-way tie of perfect scorers and no operator is named, so the results are best read as a running demonstration of what a limited agent runs into, not a rigorous survey.

Terms in this piece · Glossary
  • AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
  • llms.txt — A file at a site's root that gives AI systems a curated, plain-text map of its most useful content — robots.txt's counterpart for comprehension rather than permission.
Read the source agentability.org
Recommended reads
Comments

Checking sign-in…

Loading comments…