And turns out that actually there were almost no false-positives here at all.
Some of the flagged things were rather minor, like missing files that are not really necessary in my case, or too long meta descriptions, but technically all is flagged correctly even if severity is low.
But on the other hand, it has also correctly flagged my tracking settings or overly strict rules for AI crawlers, so this were genuinely useful for me to find out now.
And the only real false-positive is actually a recognised AI builder signature for Bolt.new - not sure where that one comes from, as Bolt.new was not used here. I've only used Claude Code, mostly with Anthropic + GLM by Z.AI models.
Hope that feedback helps you, as I have defnitely found it helpful on my end. Thanks!
reply