Edip Kilinc is the founder of Webwavers. Together with his team, he develops websites, brands, and digital presences for ambitious companies – with over 250 completed projects, and honored with awards including the German Design Award 2026.
The idea dates back to 2024 and is quickly explained: a text file in the root directory of the domain, similar to robots.txt, in which a company lists its most important content in plain text and Markdown. The thought behind it: language models shouldn't have to wade through navigation, cookie banners, and layout code, but rather get a clean summary directly.
That sounds reasonable. The problem is the commitment. There is no authority that has defined this, no promise from a provider to read the file, and to this day, no resilient proof that any of the major systems actually evaluate it.
It is a proposal with support in the specialist community, but without buyers. That could change – it is just not a basis for an investment today.
The idea dates back to 2024 and is quickly explained: a text file in the root directory of the domain, similar to robots.txt, in which a company lists its most important content in plain text and Markdown. The thought behind it: language models shouldn't have to wade through navigation, cookie banners, and layout code, but rather get a clean summary directly.
That sounds reasonable. The problem is the commitment. There is no authority that has defined this, no promise from a provider to read the file, and to this day, no resilient proof that any of the major systems actually evaluate it.
It is a proposal with support in the specialist community, but without buyers. That could change – it is just not a basis for an investment today.
Schema.org is the counter-proposal: established since 2011, jointly supported by Google, Microsoft, Yahoo, and Yandex, documented, and verifiable in practice. The markup resides as JSON-LD in the source code and explains to machines what a piece of content is – an article, a company, a question-and-answer collection, a review, a product.
For search, this has been the basis for rich snippets for years: review stars, additional details, event dates. For generative systems, it is something different and more important: a clear statement about who is behind a piece of content, when it was created, and what it refers to.
This is precisely where it is decided whether a system classifies content as citable. A text with no recognizable author, no date, and no connection to an organization is, in case of doubt, a second-choice source.
You don't need the entire vocabulary. Five types cover the needs of a service provider.
Organization – once centrally: name, logo, address, founding date, profiles on social networks. This is the anchor to which everything else refers.
WebSite – the domain itself, with language and reference to the organization.
BlogPosting or Article – per post, with headline, description, image, date, and author as a person.
BreadcrumbList – the breadcrumb path. It shows machines where a page fits into the structure and is done in just a few lines of code.
FAQPage – for pages with real questions and answers. Important: The answers must be visible on the page, otherwise the markup is not permitted.
Everything else is fine-tuning. Anyone who has these five set up cleanly is further ahead than most competitors.
The most common cause of invisibility is neither a missing file nor missing markup. It is more banal: the content is not in the HTML.
Many modern pages load their texts later via JavaScript. In the browser, everything looks correct. In the delivered document, there is nothing in that place. Search engines can partially compensate for this because they render pages. The tools that generative systems use to retrieve pages usually do not do this.
The test takes ten seconds: open the page, display the source code, and search for a sentence from the middle of the text. If it is not found, the content does not exist for these systems – regardless of how well it is marked up.
This particularly affects expandable areas, tabs, and lazy-loaded lists.
First, the HTML test for all important page types. If something is missing here, any further measure has no effect.
Then, implement Organization and WebSite once centrally. Effort: one hour. Effect: the foundation for everything else.
Then, Article and BreadcrumbList on the content pages, ideally via the template, so that new posts automatically include them.
Then, FAQPage where questions are actually answered.
Finally, if time permits, llms.txt. The file is written in twenty minutes and does no harm. It is a bet that the proposal will catch on – a cheap bet, but a bet nonetheless.
Anyone who reverses the order and starts with llms.txt ends up with a file that nobody reads and a website that machines cannot categorize.
For structured data, there are two free testing tools from Google: the Rich Results Test and the Schema Markup Validator. The former tells you whether a markup is suitable for rich results, the latter whether it is formally correct. Both should run without errors.
A third test is more important and rarely done: fetch the page with a simple request without JavaScript and look at what arrives. This shows what a language model actually gets to see.
After implementation, it is worth checking in the Search Console. Detected markups appear there with a delay of a few days to weeks, along with errors and warnings. Warnings are usually non-critical; errors mean that the markup is ignored.
After that, a quarterly check is sufficient.
