Structured data, llms.txt and the rest of the plumbing

Technical AI-visibility advice has run ahead of the evidence. Here is what the platforms themselves say, which is a shorter and more useful list.

Structured data

Google’s position is explicit: “Structured data isn’t required for generative AI search, and there’s no special schema.org markup you need to add” (Google). Its AI features page says the same: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary” (Google).

Which is not an argument against schema markup. It still earns rich results in classic search, it is how you state facts unambiguously, and it costs little. But no platform has claimed it affects whether you are cited in an AI answer, and anyone selling “AI schema” is selling a hypothesis.

llms.txt

The llms.txt convention proposes a Markdown index of your site for language models. It is a tidy idea with a specific problem: nobody who would consume it has said they do.

Google says outright that it does not: “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.” It lists llms.txt among tactics to ignore. John Mueller, asked in January 2026 whether Google publishing llms.txt files on its own properties amounted to an endorsement, answered “no” (Search Engine Roundtable).

OpenAI and Anthropic publish llms.txt for their own developer documentation. That is often cited as adoption. Publishing one is not the same as reading one, and neither company has said its crawlers fetch yours.

The genuinely interesting wrinkle is internal to Google: Chrome’s Lighthouse added an agentic-browsing audit that checks for a machine-readable summary at the domain root, while Search Central tells you to ignore the file. Both are Google.

A reasonable position: generate one if it is free to generate, because agents handed the URL can use it. Do not count it as SEO, and do not hand-maintain it.

Snippet controls, which do work

These are documented and they take effect. nosnippet, data-nosnippet, max-snippet and noindex limit what Search can show from your pages, and Google names them as the controls for AI features too. Most sites want the opposite of restriction: max-snippet:-1 opts into full-length excerpts, which is the setting that lets an answer engine quote you at length.

The control most people have missed

Since 2026 Search Console has a Search generative AI setting, letting a site owner include or exclude their content from AI Overviews, AI Mode and generative features in Discover (Google). Google states it “isn’t used as a ranking or inclusion signal affecting other parts of Search”.

This matters because the widely repeated alternative is wrong: Google-Extended does not control AI Overviews or AI Mode. It governs training and grounding for Gemini models, and it is not a crawler at all.

IndexNow

IndexNow pushes content changes to Bing, Naver, Seznam, Yandex and Yep. Google does not participate. Because ChatGPT search and Copilot draw on Bing’s index, faster Bing discovery plausibly means faster eligibility there, but no platform has confirmed that chain, so file it under cheap and sensible rather than proven.

The summary: the plumbing that demonstrably matters is crawlability, snippet permissions and that one Search Console switch. The rest of your effort belongs in what the page says and how you measure whether it worked.