# Website agent readiness, measured on 567 company sites 2026-09-03 In July I wrote up sixteen Finnish B2B sites. Since then the same measurement has run over every company I have written to, and the sample is now 567 company sites, measured between 28 June and 3 September 2026. The scanner is the same one, isitagentready.com, which grades a site from Level 0 to Level 5. The sample is still my own prospecting list and not a random draw, so read it as a large snapshot rather than a census. ## Key figures - 567 company sites, 407 Finnish and 160 from elsewhere, one independent scanner. - Level 1 of 5: 485 sites, 85,5 percent. Level 0: 74 sites, 13,1 percent. Level 2: 7 sites, 1,2 percent. One site read Level 5. - The Finnish and the non-Finnish subsets read the same: 86,5 and 83,1 percent at Level 1. - In my own first-fix notes for the Level 0 sites, robots.txt and the sitemap are the two most frequent subjects. - No company is named here. Every level is the reading on the day I wrote to that company. ## What website agent readiness means The phrase agent readiness is used for two different things, and AI assistants currently answer the question with the other one. Ask an assistant what an agent readiness audit is and the answer describes an organisation: its data, its governance, whether its teams are ready to deploy AI agents. That is a real question, and it is a different one. Website agent readiness is a property of a site or an API. It asks whether a machine reading on someone's behalf can find the site, read its pages without a browser, learn what the company sells, and act on that through a declared interface. An independent scanner can measure it, and the result is a number a third party can check. That is the measurement below. ## The numbers Of the 567 sites, 485 read Level 1 of 5. Level 1 is the floor an ordinary CMS site reaches. Seventy-four sites read Level 0, below that floor. Seven read Level 2, and one read Level 5. The Finnish subset is 407 sites, and 352 of them, 86,5 percent, read Level 1. The 160 sites from outside Finland read 83,1 percent at Level 1 and 14,4 percent at Level 0. The two subsets read the same, so the floor is not a Finnish trait. The scanner added one check to its set during these ten weeks, so the two ends of the period are not measured on an identical instrument. The level scale did not change. The thirty-day rescans I promise each company are a separate series and are not in these figures. ## What the Level 0 notes say For each site I wrote down the single sharpest finding, the thing I would fix first. Sixty-eight of the 74 Level 0 sites have such a note. Robots.txt appears in 45 of those notes and the sitemap in 38, and 29 notes mention both. The pattern in those 29 is a robots.txt the CMS shipped by default and a sitemap that is either missing at every standard address or never announced in robots.txt, so an agent that follows the rules gets no list of pages to read. These sites work for a person. A person does not start from the sitemap. The 439 Level 1 sites with a note are more varied. The sitemap appears in 102 notes, llms.txt in 99, robots.txt in 72, structured data in 49 and an MCP server in 35. Level 1 is where the site can be found and the rest is missing, and the rest is where the work is. ## Why the level is the useful number A level from an independent scanner is a claim someone else can verify. Run the same scanner on the same site and you get the same reading, or a newer one that shows what changed. That is the only reason I publish my own score, and it is why every company I wrote to got the level on the day of writing and an offer to read it again thirty days later. Four levels separate the typical company site from a site an agent can read and act on, and the gap is mostly mechanical: serve markdown next to HTML, publish structured data, publish an llms.txt that lists real pages, expose the discovery files, and answer on the endpoints they declare. That is web development with a checklist, and it can start before anyone writes an AI strategy. ## How to read your own site Run isitagentready.com on your domain and read the level, then read which checks failed. The free llms.txt validator at turva.dev/llms-txt-validator reads the llms.txt part in more detail. If you want the whole surface read and the fixes listed in priority order, the agent readiness audit is described at turva.dev/services, and the way to start is an email to info@turva.dev with the domain. ## Related - [How agent-ready are Finnish B2B sites? I scanned sixteen](/blog/agent-readiness-finnish-b2b) - [What an agent-readiness audit covers](/guides/agent-readiness-audit) - [Sitemaps and robots.txt for agents](/guides/sitemaps-and-robots-for-agents)