Google Explains How PDFs Are Indexed, Ranked, and Crawled

On October 13, 2026, Google's Martin Splitt and John Mueller published a new episode of the Search Off the Record podcast titled PDFs in Search: Indexing, AI & Pitfalls. According to Google Search On PDFs: Conversion, Links, HTML, Ranking, Limits, Crawling..., the episode covers how PDFs are converted for crawling and indexing, how Google processes links and images inside PDFs, what the file size limits are, and whether PDFs rank in both traditional search and AI-powered features like AI Overviews.

The episode is unusually detailed for a Google podcast. Splitt and Mueller walk through the full technical pipeline, not just surface-level guidance. If you've ever wondered why your PDF rankings behave differently from your HTML pages, this one is worth your time.

The Details: How Google Actually Handles PDFs

The core reveal is straightforward but important. PDFs are binary container files, not plain text. Google runs them through a converter component that transforms each PDF into an HTML representation. From there, the internal search systems treat them essentially like regular web pages. So when people ask whether PDFs can rank, the answer is yes, because by the time Google's systems see them, they have already been converted.

Links embedded in PDFs survive this conversion process and are crawled normally. Images, though, are a different story. Because embedded PDF images do not have their own standalone URLs, Google generally cannot surface them in Google Image Search the way it would a standard web image. If image visibility matters to your strategy, that is a real limitation worth knowing.

On AI visibility specifically, the podcast confirms that converted PDFs can appear in AI Overviews and are accessible to AI models for training and grounding purposes. That is meaningful for anyone thinking about generative engine optimization. Your whitepapers and reports are not invisible to AI search, as long as they can actually be read.

Crawl size limits came up too. The podcast notes that the crawling limit for PDFs is larger than the limit for standard HTML pages. Google deliberately avoids publishing an exact cutoff because the threshold is adjusted over time as file size norms evolve. Plan for large PDFs to be processable, but do not assume there is no ceiling.

One of the more practical topics in the episode is recrawl frequency. Google treats PDFs as relatively static binary files, similar to images, and recrawls them much less frequently than HTML. If you update a PDF and need Google to index the new version quickly, the guidance is clear: publish it under a new URL or file name rather than overwriting the existing file.

What This Means For You

Picture this: you have a client whose most authoritative content lives in a PDF library. Research reports, product guides, regulatory filings. They assume that because the files exist on the site, Google is treating them the same as their blog posts. This podcast is a useful reality check on that assumption.

The slow recrawl rate is probably the biggest practical issue I see agencies overlook. If a CMS is auto-generating PDF versions of live web pages, both versions can end up competing in search results, and the PDF version will almost always be stale. The podcast flags this explicitly: avoid automated CMS tools that generate PDF duplicates of live pages.

The AI Overviews angle matters more than most people are treating it right now. If your PDFs are well-structured and machine-readable, they are in play for AI grounding. If they are scanned image files with no selectable text, they are essentially invisible. You can test this yourself: try highlighting text in the PDF and pasting it into a text editor. If nothing pastes, a search engine cannot read it either.

Canonicalization is also trickier with PDFs than with HTML. Because identical PDFs are exact binary duplicates, cross-domain canonical signals are harder to apply. The podcast recommends using rel=canonical HTTP headers, populating PDF metadata fields, and placing prominent links back to your website in the document header. These are easy wins that most site owners skip.

What to Do Now

1. Audit your PDF library for scanned or rasterized files. Try the highlight-and-paste test on any PDF you think Google is indexing. If the text does not copy, it is not being read. Fix it or replace it with an HTML equivalent.
2. Stop overwriting PDFs when you update them. Publish updated versions under a new URL or file name so Google picks up the new content on the next crawl rather than waiting for a recrawl of the old URL.
3. Check whether your CMS is auto-generating PDF duplicates of live pages. If it is, disable that feature or set up canonical HTTP headers to point to the HTML version.
4. Treat your structured, text-based PDFs as AI visibility assets. If you publish research, reports, or authoritative guides, make sure they are machine-readable and linked from your main site so they can surface in AI Overviews and AI mode.
5. Default to HTML wherever you have a choice. It is faster to index, responsive on mobile, and easier to update. Reserve PDFs for content that genuinely needs rigid print formatting, downloadable forms, or offline documentation.
6. Keep llms.txt as plain text. The podcast specifically notes that files intended for machine consumption should never be formatted as PDFs. The entire point of that format is lightweight, simple plain text.

Background and Context

None of this guidance is entirely new. Google has been indexing PDFs for years, and Googlebot's ability to follow links inside PDFs has been documented for a long time. What makes this episode valuable is the consolidation. Splitt and Mueller walk through the full pipeline in one place, including the newer AI Overviews and AI mode angle that prior documentation never addressed.

The GEO section of the podcast is the freshest territory here. As AI Overviews become a more significant source of zero-click visibility, understanding which content formats are accessible to those systems matters more. The fact that converted PDFs are confirmed as eligible for AI grounding is something worth building into your content strategy if you publish a lot of document-style assets.

The mobile accessibility point is also worth revisiting. PDFs have fixed print layouts that do not reflow text when zoomed, and they must download completely before rendering on mobile. For a user on a slow connection, that is a genuinely bad experience. Google has flagged PDFs as not mobile-friendly in the past, and nothing in this episode suggests that has changed.

If you want to track how your PDFs are actually performing in search compared to your HTML pages, rank tracking broken out by URL type can surface patterns you would otherwise miss. Knowing which PDFs are ranking and which are being ignored by the crawler is the first step to prioritizing the right fixes.

The bottom line is this: PDFs can rank, they can appear in AI Overviews, and they can pass link equity. But they come with enough limitations around mobile, recrawl frequency, and canonicalization that HTML should be your default. Use PDFs with intention, not as an afterthought.