Schema everything
Across a research network of 50 live sites in 9 industries (200+ days of logs), six schema types produce about 97% of all AI schema engagement: Organization and Person (~28%), Article and NewsArticle (~22%), Product and Offer (~18%), FAQPage and QAPage (~14%), BreadcrumbList (~10%) and Recipe (~5%). The other 700-plus types share the remaining 3%. Adding fifteen types per page costs hours, bloats every response, and raises the chance that two blocks contradict each other, which is one of the four things that make bots discard schema entirely. Less, but correct, wins.
HowTo
Google retired the HowTo rich result in September 2024. In the logs AI fetchers read it inconsistently and rarely. The markup is heavy, has to be kept in sync with the visible steps, and when it drifts it contradicts the page. Write the steps as clear headings in the HTML instead. That is what bots actually extract.
Speakable
Proposed for voice assistants in 2018. No current AI surface in the logs uses it. It is dead weight on every page that carries it.
FAQ spam
FAQPage schema is useful, which is exactly why it gets abused. It works when the questions and answers are visible on the page, written for a real reader, and match what the page is about. It fails when a plugin stuffs ten keyword questions into JSON-LD that nobody can see, or pastes the same FAQ block onto every page. Schema that contradicts or invents visible content is one of the four mistakes that nuke schema: bots notice, and the inconsistency can suppress citation of the whole page. Google restricted FAQ rich results to government and health sites in 2023 for the same reason. This plugin prints no FAQPage itself. It flags FAQ schema whose questions do not appear in the visible text of the page.
llms.txt as a lever
Zero requests from any AI bot across a research network of 50 live sites and 200+ days of logs. No AI lab has adopted it. The plugin can generate one because it is free and harmless, and the card says plainly that it does nothing yet. The full write-up: Why llms.txt is dead.
Multiple disjoint JSON-LD blocks
Bots struggle to merge separate blocks that do not reference each other, and may keep one and drop the rest. One graph, cross-referenced by @id.
JavaScript-injected schema
It shows in your browser's developer tools and is absent from the raw HTML. Some fetchers never run JavaScript. This plugin prints schema in the server's HTML only.
Rate limiting tuned for humans
Security plugins that throttle "crawlers" return 429 errors to AI fetchers that expect to retry. This plugin names the setting instead of pretending it is not there.
Promising citations
The dashboard measures the conditions for citation. Citation is the model's choice. Anyone promising a specific placement in ChatGPT is guessing.
