Structured data for Google and ChatGPT

JSON-LD is a dictionary next to the HTML. Google uses it for rich results. Agents use it to see what you are. It does not slide you to the top on its own. Dumping five types on one URL “just in case” makes it vaguer.

One Organization, same name as the footer and llms.txt. Article with headline, date, publisher. FAQPage only if the question and answer sit in the HTML, not in a script the visitor never sees. Product on real product pages, price that is true. Breadcrumbs that match the path you actually click. Review stars you do not show on the page: leave them out. That is how you earn a manual action.

I have seen five Organization blocks that disagree more often than I would like. Microdata and JSON-LD with two brand names too. Rich Results Test pulls that out in a minute. Run it after every theme update. Schema gone is a classic.

Mark only what the page is. A contact page is not a Product. A blog post is not a FAQ because you invented two questions at the bottom. Honest schema is dull. Fake schema is work later.

After a plugin update I check whether a second JSON-LD block showed up. Two Organization names, two addresses: that is how rich results vanish while the HTML looks the same. One source, one block, the same text as on the page. More is not an extra chance. It is noise.