Google warns agencies against markdown SEO shortcuts for AI visibility

Google’s John Mueller and Martin Splitt pushed back against parallel markdown versions of websites as an AI SEO tactic, arguing that converting or duplicating pages into markdown is a complicated workaround for a simple job: keep the site structured, accessible, and easy to process. The June 23 discussion landed as agencies were already hearing more requests for AI-specific visibility hacks, and it put duplicate infrastructure on the wrong side of the trade-off. HTML, they stressed, is for browsers and screen readers, while markdown strips out non-content elements rather than improving the site itself.

That warning matched Google Search Central’s June 2026 guidance on generative AI features. Google said those features are rooted in core Search ranking and quality systems, and that SEO best practices still apply. It also said its generative AI features use retrieval-augmented generation and query fan-out to ground answers in indexed web pages, which is a long way of saying the answer still depends on the same crawlable site structure agencies have spent years optimizing. Google’s advice stayed blunt: focus on foundational SEO, valuable non-commodity content, and a clear technical structure.

Google’s structured-data documentation pushes the same point from a different angle. It says structured data helps Google understand page content and can enable rich results, but it does not guarantee that a page will appear that way. The documentation also cites performance gains that are hard to ignore: Rotten Tomatoes saw 25% higher click-through rates, Food Network recorded a 35% increase in visits after converting 80% of pages, Rakuten posted 1.5x longer time on page and 3.6x higher interaction rates on AMP pages with search features, and Nestlé reported an 82% higher click-through rate for rich-result pages. Google’s guidelines also say structured data should not be hidden from users and should not be blocked from Googlebot if a page is meant to qualify.

The markdown debate was already moving before Mueller and Splitt weighed in. In February 2026, Google and Bing representatives were being read as against separate markdown pages for LLM use, with the concern that serving one version to bots and another to users could be treated as cloaking. Cloudflare added to the pressure that same month with Markdown for Agents, which auto-converted HTML to Markdown for AI crawlers and was described as cutting token usage by up to 80%. That is exactly why the agency pitch is shifting back to basics: audit the existing site, tighten the HTML, reinforce the hierarchy, and keep one version of the page. The stronger brief is one maintainable site, not a parallel markdown layer that doubles upkeep and fragments trust.