Google's AI Overviews are surfacing raw markdown files as cited sources inside the snippet experience, according to reports tracked by Search Engine Roundtable. Screenshots shared by search professionals show AI Overviews citing and linking directly to .md files — plain-text documents written in markdown formatting — rather than to rendered HTML pages.

The behavior appears unintentional. Markdown files published on the web do render in browsers as unstyled text in many cases, complete with visible hash symbols for headers, bracketed link syntax, and asterisks for emphasis. When AI Overviews pull such a file as a supporting source, users see links that lead to documents that look broken or unfinished, not to polished web pages.

Search Engine Roundtable, which aggregates anecdotal reports from the SEO community, highlighted the issue after multiple independent examples circulated on social platforms. The site's coverage characterizes the phenomenon as "unexpectedly" showing markdown files — a framing that signals this is an observed glitch rather than a documented feature rollout.

What is confirmed versus what is speculation

There is no confirmation from Google that this behavior is intended, announced, or even acknowledged. Google has not issued a statement on the matter as of the reporting. Everything known about the situation comes from user-submitted screenshots and community observations — tool-signal-level evidence, in other words, not official search-engine communication. Treat the reports as credible sightings of a rendering or source-selection quirk, not as a policy change or product update.

The speculation among observers is mechanical: AI Overviews select supporting links from Google's index, and markdown files are indexable content like any other document. If a .md file ranks well for a query — something increasingly common as developers publish technical documentation directly as markdown on GitHub and similar platforms — the AI Overview generation pipeline may cite it without distinguishing between a rendered page and a raw file.

Who is likely to notice

The impact skews toward technical and developer-oriented queries, where markdown documentation is abundant. Sites that publish software documentation, API references, and open-source project guides in .md format are the most probable sources of these citations. For publisher sites with conventional HTML pages, the practical effect is minimal — the issue is less about lost traffic and more about the quality of the answer experience Google presents.

The more significant implication is reputational. Each glitch of this kind feeds the running critique that AI Overviews surface low-quality or malformed sources. For a feature Google has positioned as the future of search results, citing documents with exposed formatting syntax is the kind of detail that erodes user trust in the summaries.

What to watch next

The key monitorables: whether Google addresses the behavior publicly or quietly patches source selection to filter raw markdown files from citations; whether the frequency of .md citations grows or remains a scattered edge case; and whether site owners of heavily cited documentation respond by blocking indexing of raw markdown versions. Search Engine Roundtable's ongoing tracking of community reports is the best early signal — if examples multiply in the coming weeks, that pattern would suggest a systemic gap in AI Overview source filtering rather than an isolated anomaly.