Schema Markup Vs Llms.Txt For Ai Visibility

• Schema markup is structured data embedded in a page's HTML that tells machines what an entity is, such as a product, an author, or an organization.
• The measured answer is mixed and, for Google surfaces, close to flat.
• No documented evidence shows that an llms.txt file makes AI engines cite a site more often.
• Crawler behaviour is the layer where the two files diverge most.
• Schema is worth keeping for regular search rich results even if its AI citation effect is unproven. For instance, llms.txt was proposed by Jeremy Howard of Answer.AI and first published on 3 September 2024, per llmstxt.org as described by aischemagen.com; the spec was updated to version 2 on 10 August 2026.

Schema markup is structured data embedded in a page's HTML that tells machines what an entity is, such as a product, an author, or an organization. Its job is to remove ambiguity from content that already exists on the page. An llms.txt file works at a different layer. It is a plain Markdown file placed at the site root that proposes a curated map of key pages for language models. Under the llmstxt.org spec proposed by Jeremy Howard of Answer.AI, first published on 3 September 2024, the only required element is an H1 with the site or project name. A blockquote summary, background text, and H2 sections of curated links are optional. Schema describes what is on a page; llms.txt points at which pages to read. Confusing the two is the most common mistake in this debate.

Google's guidance on structured data is documented on its own developer pages; the llms.txt rows reflect the spec and Google's stated position. Ahrefs' June 2026 study of 137,210 domains found that 97% of llms.txt files that existed received zero requests during May 2026, according to Ahrefs (as reported in the aischemagen.com post).

The measured answer is mixed and, for Google surfaces, close to flat. In a test of 1,885 pages that added JSON-LD schema between August 2025 and March 2026, compared against about 4,000 similar pages without it, Ahrefs found that Google AI Overviews citations fell about 4.6% relative to comparison pages. Google AI Mode and ChatGPT citations rose about 2.4% and 2.2%, according to lpageo.com's summary of Ahrefs. Those movements could not be distinguished from zero. Google's documentation adds that no special schema.org markup is required to appear in AI Overviews or AI Mode, according to Google's AI features documentation as cited by lpageo.com. Specifically, Microsoft's Fabrice Canel said at SMX Munich in March 2025 that schema markup helps Microsoft's LLMs understand content. The defensible reading is that schema aids comprehension, not a documented citation switch.

• Ahrefs test: 1,885 schema pages versus about 4,000 comparison pages
• Google AI Overviews: about 4.6% decline relative to comparison pages
• Microsoft: schema helps its LLMs understand content

No documented evidence shows that an llms.txt file makes AI engines cite a site more often. A study of nearly 300,000 domains by SE Ranking found no link between having an llms.txt file and how often a site is cited as a source in AI answers, according to SE Ranking as cited by aischemagen.com and Search Engine Journal. However, Ahrefs' June 2026 study of 137,210 domains found a stronger signal of disuse. According to Ahrefs as reported in the aischemagen.com post, 97% of llms.txt files that existed received zero requests during May 2026. For example, a file that no crawler fetches cannot influence the crawler's answer. Specifically, teams that publish an llms.txt file for control or documentation reasons can reasonably do so. A citation case built on llms.txt, however, currently rests on assumption rather than measurement.

• SE Ranking: no link between llms.txt and citation frequency
• Ahrefs: 97% of existing llms.txt files received zero requests in May 2026
• Ahrefs study covered 137,210 domains

Crawler behaviour is the layer where the two files diverge most. OpenAI, Anthropic, Perplexity, Google (Google-Extended), and Apple (Applebot-Extended) each publish named AI crawler user agents that site owners can allow or block in robots.txt, according to the storyworks.marketing guide. That documented control surface concerns access, not llms.txt. However, Google's stated position is that its search systems ignore llms.txt, according to Google as reported by pixelmojo.io. Schema, by contrast, is documented by Google as the way to describe entities and enable rich results, according to pixelmojo.io and lpageo.com. For example, a site owner can check each named crawler in robots.txt. Specifically, before weighing either file, verify which crawlers you permit and confirm that your pages render their content to a non-JavaScript fetch.

• OpenAI, Anthropic, Perplexity, Google-Extended, Applebot-Extended
• robots.txt is the documented control surface for AI crawler access
• Google: schema is the documented route to describing entities

Schema is worth keeping for regular search rich results even if its AI citation effect is unproven. Google's documentation treats structured data as the documented route to rich results, so the case does not depend on AI. However, prioritise types that describe the entity a page is about. For example, use Organization and Person for authorship and identity, Product or SoftwareApplication where the page sells something, and Article for editorial content. Specifically, use FAQPage only where the questions genuinely appear on the page. Keep the markup truthful to visible content, because schema that describes text a reader cannot see is a credibility risk regardless of engine. Treat schema as a clarity layer on good content rather than a substitute for it.

• Organization and Person: identity and authorship
• Product or SoftwareApplication: pages that sell something
• Article: editorial content
• FAQPage: only where questions appear on the page A study of nearly 300,000 domains by SE Ranking found no link between having an llms.txt file and how often a site is cited as a source in AI answers, according to SE Ranking as cited by aischemagen.com and Search Engine Journal.

Fix access and rendering first. Schema and llms.txt both assume a crawler can reach and read the page. Check robots.txt against the named AI user agents, including OpenAI, Anthropic, Perplexity, Google-Extended, and Applebot-Extended, and confirm critical content appears in the server response rather than only after client-side JavaScript runs.

However, the most important step is making sure canonical URLs and internal links expose the pages you want cited. A July 2026 review of 45 studies on generative engine optimisation found no technique with a stable, long-term effect across AI platforms on how often a site is found, according to arXiv paper 2607.14035 as summarised by lpageo.com.

That review is a reason for humility, not inaction. Basic crawlability is the prerequisite every other lever depends on. None of these steps guarantees a citation, and that limit should shape what you promise stakeholders.

The specific figures and claims on this page are grounded in the following sources, reviewed at the time of writing:

• Schema Markup for AI Citations: The Technical Implementation Guide
• Schema.org vs llms.txt: which AI search signal matters most?
• Do Schema and llms.txt Help AI Citations? What the Evidence Says
• llms.txt vs. Schema Markup: Which One Do AI Engines Actually Use?
• llms.txt & Schema: The Files AI Looks For on Your Site
• Schema Markup for AI Overviews: The Technical Guide

Related guides

• Schema Markup for AI Search Engines: Implementation Guide
• Schema Types That Matter for AI Search Visibility
• Does Schema Markup Help AI Citations? What the Tests Show