Kuroma
Public AI Readiness score cards
Anyone can run a free AI Readiness audit on any public URL. What the audit found is what gets published here: the score, the four category breakdowns, and every check the rubric ran with the evidence it found on the page. Where a card can also say what fixing something is worth in points, it ranks the fixes that way. Where it cannot, it leaves that out rather than guess.
How the 4,748 cards on rubric version 2026-08-13 score
- Barely legible 368 · 8%
- Partly legible 3,819 · 80%
- Mostly legible 558 · 12%
- Highly legible 3 · 0%
A score is how legible one page is to an AI answer engine: whether the thing that has to read it can find the content, understand what the page claims, and attribute it. Barely legible is 0-39; Highly legible is 80-100.
These figures cover the 4,748 of the 5,494 published cards scored on rubric version 2026-08-13. The others were scored under a different generation of the rubric, or before a card recorded one at all, so their scores are not comparable with these: they are listed below and never averaged in.
Most recently audited
How to read this
- This is not a sample of the web. It is the set of URLs people chose to run through our free audit, which skews toward sites whose owners already suspected they had a problem. Treat any distribution here as a description of this corpus, not of the internet.
- Every card is a real audit. Nothing here is generated for the archive. Each card is the unedited result of one run of the same engine the product runs, scored by the rubric in force on the day it ran, and rewritten in place if the URL is audited again.
- Scores are comparable only within a rubric generation. When the rubric changes what a factor measures, its version is bumped and the scores before and after stop being two measurements of the same thing. The current version is 2026-08-13. Each card shows the score it was given. The average and distribution above cover only the cards scored on that version; the rest are listed without being folded into them. The count of published cards is the one figure that covers the whole archive, because counting cards does not depend on the rubric. The industry benchmarks follow the same rule.
- Not every audit gets a card here. An audit that a bot wall blocked, or whose URL did not resolve, has no score to publish and never appears. Neither does one whose stored result kept no per-check detail: that card would show a score with nothing behind it. Those audits still answer at their own address; we just do not list them or ask a search engine to index them.
- Industry aggregates live elsewhere. A public audit has no company behind it, so these cards carry no industry. Per-industry medians and failure rates come from a separate audited-domain corpus that never names a company. The industry benchmarks publish an industry once enough of its domains have been audited and at least one rubric factor has been scored on enough of those audits to rank, and say so plainly while the corpus behind them is still too thin. Read the full methodology.