How the score is produced, and what it does not claim
Published September 14, 2026
The whole point of a readiness score is that you can check it. This page lays out every rule, so nothing about your report is a mystery.
What we measure
The audit fetches a site the way a plain AI crawler does, without running JavaScript, and scores it across four categories. Each category holds a handful of checks, and each check quotes the exact content it read. The weights below add up to 100. This is rubric v0.2, the version named in the methodology link printed on every report generated today.
| Category | Weight | What it covers |
|---|---|---|
| Crawl access | 30 | Whether AI crawlers are allowed in, whether a sitemap and an llms.txt guide exist, and whether the site offers enough readable pages to sample. |
| Structured data | 25 | Whether the pages carry machine-readable labels (JSON-LD) that name the business and describe what it offers, so assistants quote facts instead of guessing. |
| Legibility | 30 | Whether the important text is present in the raw HTML without JavaScript, plus sane page titles, descriptions, and link-preview tags. |
| Contactability | 15 | Whether a phone number, an email address, and a physical address are discoverable in the pages, so an assistant can tell someone how to reach the business. |
How evidence works
Every deduction quotes the site itself. When a check marks something down, the report shows the words it read: the robots line that blocked a crawler, the JSON-LD fields that were missing, the title that ran too long, the contact detail it could not find. There is no hidden judgment and no opinion. If you disagree with a finding, you are disagreeing with a quote from your own page, which means you can fix it and re-run.
The same site, the same score
The audit is deterministic. It runs the same fixed checks in the same order, and it chooses which pages to sample by a stable rule rather than at random, so the same site scored twice returns the same number. That matters most for monitoring: when your score moves month to month, the movement reflects your site changing, never the tool changing its mind.
The complete-sample rule
The audit samples your homepage plus a few key pages reached from its navigation. If a page fails to load even after a retry, the run is marked as an incomplete sample and recorded as such. We only ever compare two scores when both came from complete samples. An incomplete run is never quietly treated as a drop in your score, because the gap was ours, not yours.
What we do not measure yet
Readiness is not the same as visibility, and we keep them separate on purpose. Whether an AI assistant actually names your business is a visibility probe, reported on its own and never folded into the readiness score on this site. Off-site reputation signals, such as reviews and mentions across the wider web, are on the roadmap and are not part of any score today. When a signal is not measured, we say so rather than implying it counts.
Score versioning
Every report carries the version of the engine that produced it. When we change a check or a weight, we advance the version, and we never restate old numbers under new rules. A score is only compared with another score from the same version. This is the promise that keeps a rising or falling number honest over time.
The same rules, in plain language
If you would rather read what all this means for a working business than how the arithmetic is done:
- How do I check what ChatGPT says about my business?
- Why doesn't ChatGPT recommend my business?
- What is llms.txt and do I need one?
- How do I explain AI findability to my web design clients?
The full rail is at Findability, and what an audit covers is one short page.
Removal policy
The audits and probes behind these pages were run over real businesses, and this site publishes what they showed at cohort level only. No page here lists an individual business with its score, and none ever will: the counts are reported per cohort, per date, with the method above attached.
If your business was audited or probed and you want its data excluded from the dataset and from anything we publish from it in future, email info@vonzie.studio and we take it out promptly. No reason required.
Frequently asked questions
- How is the score calculated?
- Each check earns its full weight when it passes, half when it needs work, and nothing when it fails. A check we could not measure is set aside and left out of the total, so a network hiccup never reads as a bad website. A few checks are marked for information only: they report a choice you made, such as whether you allow AI training crawlers, and they are worth zero points either way. The four category weights are crawl access 30, structured data 25, legibility 30, and contactability 15, summing to 100.
- Do you measure whether AI assistants actually name my business?
- That is a separate measurement called a visibility probe, and it is reported on its own, never blended into the readiness score. Readiness is about what a site makes readable. Visibility is about what an assistant answers on a given day. Mixing them would hide which one moved, so we keep them apart.
- Can a business have its data removed?
- Yes. No page here lists an individual business with its score: the counts are reported per cohort, per date, with the method attached. If your business was audited or probed and you want its data excluded from the dataset and from anything we publish from it in future, email info@vonzie.studio and we take it out promptly. No reason required.
- What happens to my score if you change the checks?
- Every report records the engine version that produced it. When we change a check or a weight, we change the version, and we do not silently restate old numbers under new rules. A score is only ever compared with another score from the same version and a complete page sample.
Your own site
Find out what an AI reader sees on yours
The same audit behind every number on these pages, run on your address, with the exact line quoted behind every deduction.