JSON-LD
JSON-LD (JavaScript Object Notation for Linked Data) is a format with which you embed structured data into a web page. In a machine-readable block, it describes what is on the page: for example a company, a product, or an FAQ. Search engines and AI systems read this block directly and thus understand the content faster and more unambiguously.
Why JSON-LD matters for AI visibility
AI assistants like ChatGPT, Perplexity, or Google AI Overviews have to grasp in seconds what your page is about. Pure running text is ambiguous: is "Sonnwendhof" a place, a name, or a hotel? JSON-LD removes this ambiguity because it cleanly marks up facts, such as name, address, opening hours, or prices. The clearer a machine understands your content, the more likely it is to draw on it as a reliable source and name you in an answer. JSON-LD does not replace good content, but it makes existing facts tangible for AI systems. This increases the chance that your brand is cited correctly and with the right details, instead of not appearing at all or appearing wrongly.
How JSON-LD works technically
JSON-LD usually sits as its own code block in the head area or at the end of your HTML page, embedded in a script tag. Unlike older methods (microdata), it does not mix into the visible text but lies separately alongside it. This leaves your layout untouched and makes the markup easier to maintain. In terms of content, JSON-LD follows the vocabulary of Schema.org: it defines which types exist (such as Organization, Product, Event) and which properties belong to them. So you specify which type is present and then fill fields like name, description, or date with values. Search engine crawlers and AI crawlers read out this block and link the details to their knowledge graph.
Common mistakes
The classic mistake is that the marked-up data does not match the visible page content. If your JSON-LD names a price of 89 euros but the page shows 129 euros, Google and AI systems devalue this as manipulation. Just as common: required fields are missing, date entries are formatted wrongly, or an unsuitable type is chosen. Outdated details also cause harm, such as opening hours that no one maintains anymore. Another stumbling block is syntax errors, because a single wrong comma can render the whole block unusable. Therefore check your markup with a validation tool before you publish it, and keep the data as current as the visible page itself.
Relation to AI recommendations
In AI search, the unambiguity of your data often decides whether you get cited. JSON-LD delivers exactly this unambiguity: it tells the machine not only that a text exists, but what it means. An FAQ schema, for example, signals directly which question belongs to which answer, ideal, because AI assistants search precisely for such question-answer pairs. This increases the citability of your content. JSON-LD is thus a building block of generative engine optimization: you make it as easy as possible for AI systems to take over your facts correctly. This replaces neither reach nor authority, but it measurably increases the probability that your brand appears as a clean, reliable source in answers.
Example
Imagine a small bakery that wants to be found online. On its home page there is a JSON-LD block that marks it up as a local business: name, address, phone number, opening hours, and the info that it carries gluten-free bread. If someone asks an AI assistant "Where do I get gluten-free bread in Regensburg on Sunday?", the system can read the opening hours and the range directly from the markup. Without JSON-LD, the AI would have to laboriously guess these facts from running text, with it they are unambiguously available and the bakery is more likely to be named.
Common questions
Do I need programming skills for JSON-LD?
Basic knowledge helps but is not a must. Many content management systems and plugins generate JSON-LD automatically from your entries. What matters is that you keep the details current and check them with a validation tool before they go live.
Is JSON-LD the same as Schema.org?
No. Schema.org is the vocabulary that defines which types and properties exist. JSON-LD is the format in which you technically write this vocabulary on your page. Both belong together but are not identical.