How the score is calculated

Scoring rules

• Three outcomes per check. Fine — full weight, needs work — half, blocking — zero.
• What cannot be checked is not penalised. If a check does not apply (no language versions) or is not run in quick mode (loading speed), its weight drops out of both the earned score and the maximum. Otherwise a single-language site would lose to a multilingual one for no reason.
• Critical errors are shown separately. Some failures are not a low score but a breakage: the page does not exist for a machine. Those are lifted to the top of the report, and they carry the heaviest weights in the scale so the number and the text say the same thing.
• One check carries no weight. Analytics events do not affect whether the site gets found: we show it as an observation but do not score it.

Why these weights

Weight is the damage to visibility when a check fails — not how hard it is to fix and not the order of the line. That is why noindex costs more than sitemap.xml: the first removes the page from the index entirely, the second merely helps a robot find sections faster that it would reach through links anyway.

For the same reason llms.txt gets a cautious weight: the standard is young, its effect on citation is not yet proven, and inflating it would pass a hypothesis off as a fact. And AI crawler access carries a high weight but not the highest: a block zeroes visibility, yet most sites block no one, so the line confirms the norm more often than it tells sites apart.

The weighting is our expert judgement, not an industry standard. It will change as measurements accumulate, and every change will land on this page — because the page reads the engine rather than retelling it.