A playbook for making a site legible to AI.
Not good enough to be taken as rules of the web. Good enough to run on the next site.
1 site · 17 versions · 28 observations (9 replicated)
(n) = observations behind the entry
reach (8)
Did the system get your page at all?
- A cautious answer usually means nothing was found(1)
- Similar names can inherit each other's reputation(1)
- The chatbot and the API do not fetch the same way(1)
- A page can render perfectly and be broken for machines(2)
- An assistant can confidently describe outdated copy(1)
- If it needs JavaScript, it arrives empty(1)
- Your sources may not be reachable(1)
- Conclusion: Check what the crawlers get. The rest isn't yours.(1)
understanding (0)
Having got it, did it parse what you meant?
- No observations were filed here. Every one landed in reach or credence, which is reason to think parsing is not where this fails.
credence (9)
Having parsed it, did it treat your claims as true?
- Put the source next to the claim(1)
- Publish the raw, not the conclusion(1)
- A gap you leave gets filled by the reader(1)
- Say which claim you can't prove(1)
- Name the thing you don't have(1)
- Predict the discrepancy before someone finds it(1)
- Well-formed data reads as audited(2)
- Text aimed at AI reads as manipulation(2)
- Conclusion: Aim to be auditable, not believed(3)
method (6)
How to measure this without fooling yourself.
reference (5)
The tags, fields and files the rest of this is about.