The difference between a schema score of 60 and 90: what actually matters to AI crawlers?

I've spent enough time in the trenches of structured data to know that a schema score is essentially a confidence interval, and most organizations treat it like a credit score when it's really more like a blood pressure reading. The schema must not lie. Here's what actually matters: a 60 versus a 90 isn't just a 30-point gap in compliance. It's the difference between "the crawler understands *most* of what you're saying" and "the crawler understands what you're *trying* to say." Google's own documentation doesn't emphasize this enough. At 60, you're missing critical properties—likely in context fields like `datePublished`, `inLanguage`, or `articleBody`. Your rich snippets still render, sure, but you're leaving interpretation gaps. At 90, you're approaching what I'd call "unambiguous intent." The crawler doesn't have to guess whether your event is virtual or hybrid, or whether your review is actually a recipe. The difference compounds downstream, especially when AI systems start reasoning about your data rather than just indexing it. What really gets me is that most teams obsess over reaching 90 as if it's a finish line. It's not. I've seen 85-rated schemas outperform 95-rated ones because the 85 was *semantically precise* while the 95 was just technically complete—stuffed with optional properties that contradicted each other. The schema must not lie, and a lie by addition is still a lie. @Echo Zhang has been preaching this at every standup, and she's right: quality of signal beats quantity of fields, especially as LLMs become your real audience. The uncomfortable truth? Beyond 85, you're optimizing for crawler psychology, not actual understanding. You're gilding. At 60-75, you're genuinely limiting discoverability and reasoning. The sweet spot most people miss is 78-84: specific enough to be useful, constrained enough to mean something. So here's my challenge: show me a case where hitting 95+ actually changed your CTR or conversion rate, *not* just your crawl efficiency. I suspect the correlation breaks down faster than most vendors would admit. @Kai Ostrowski, @Maya Chen—have you seen this in the wild?