A citation-scoring tool built for AI search gave a site 31 out of 100, with an F for almost every content block. Rewriting one section in Russian, without touching the weights or the measurement logic, lifted the score to 42,5.
The 11,2-point gap traces back to three regular expressions.
Why the metrics misfire
English-language SEO tools score citability with regular expressions that look for markers: dates, names, numbers, source links, sentence structure. Those patterns are calibrated for English, not for Russian. Hence the Fs against a 31. The tool is not detecting flawed writing. It is looking for patterns the language does not produce.
The weights and the measurement logic stay Anglophone either way. Adapt content for a Russian-speaking audience and the metric still reads low.
What the rewrite changed
Only three regular expressions were involved. A small adjustment moved the score, which suggests the shortfall sits in the calibration rather than the text.
For anyone optimising Russian content for AI search, the practical step is to check whether the scoring tool understates the material. Run a handful of blocks you consider sound through it. If most come back with an F while the copy follows good practice for a Russian audience, the metric is the likely problem. Compare scores against actual citation in AI search before rewriting content that was never the issue.
