Google Says Markdown Hurts AI SEO

On a recent episode of Search Off the Record, Google’s John Mueller and Martin Splitt challenged the growing claim from some AI SEO discussions that stripped-back, content-only formats such as Markdown are a better way to optimise for AI search. They argued that removing parts of a webpage actually strips away information that search systems rely on.

Non-content elements still matter

Splitt explained that HTML is designed not just for machines, but also for browsers and accessibility tools such as screen readers. Because of this, it contains far more than just visible text content.

He noted that some people are drawn to Markdown because it appears “cleaner” and removes what they see as unnecessary code. From an AI perspective, fewer tokens and simpler formatting can seem more efficient when a page is viewed in raw form.

However, he also pointed out that Markdown still carries structure in a readable way, such as link formatting and basic hierarchy. This simplicity, he suggested, is part of why it is often assumed to be more suitable for machines consuming content.

HTML is already easy for systems to process

Mueller and Splitt pushed back on the idea that HTML is difficult for crawlers or AI systems to interpret. While HTML may look complex to humans when viewed as raw code, they emphasised that parsing it is a well-established and straightforward process.

Mueller explained that web crawlers have been working with HTML for decades and already have robust tools and libraries for extracting usable text. From a technical standpoint, converting HTML into clean text is not considered a difficult task.

Because of this, the argument that Markdown simplifies crawling was seen as less compelling, since modern systems already handle HTML efficiently.

The importance of site structure and discovery

Another key point raised was that Markdown tends to focus purely on content, while stripping out important structural elements such as navigation, internal links and hierarchy.

Splitt highlighted that these elements are not decorative extras but play a key role in how search engines understand and navigate websites. Internal linking and structure help systems discover related pages and interpret how content fits within the wider site.

By removing these elements, Markdown-based approaches risk losing signals that help with discovery and context across a website.

Why context matters for SEO

The discussion also touched on a broader SEO principle: search engines do not view websites as isolated pages, but as interconnected systems of content within a wider web ecosystem.

This means individual pages are interpreted in relation to the rest of the site, including categories, internal links and overall structure. Removing structural information can therefore limit how well a page is understood in context.

From this perspective, both fine-grained content and wider site architecture are important for search performance.

Trust and practical limitations

The hosts also raised concerns around trust and standardisation. In the past, simple metadata systems such as keyword tags were heavily abused, which led search engines to reduce reliance on them.

They suggested a similar issue could emerge if simplified content formats were treated as primary sources, since anything easier to manipulate may eventually become less reliable as a signal.

For this reason, they argued that search engines are unlikely to treat Markdown versions as more authoritative than original HTML, especially when HTML already contains the full, verifiable structure of a page.

Final thoughts

Overall, the discussion reinforced the idea that simplifying content for AI systems is not always beneficial. While Markdown may look cleaner and more lightweight, it removes structural and contextual information that plays an important role in discovery, interpretation and SEO.

Mueller and Splitt’s core message was clear: the elements often seen as “extra” in HTML are actually part of what helps search engines understand the full picture of a website.