Google's Deep Dive on PDFs: Ranking, Crawling, and AI Visibility

Google's Martin Splitt and John Mueller released a new Search Off the Record podcast on October 13, 2026, dedicated entirely to PDFs in search. Titled *PDFs in Search: Indexing, AI & Pitfalls*, the episode is one of the most detailed explainers Google has published on how its systems handle PDF files, covering everything from the conversion pipeline and crawl limits to AI Overviews visibility and GEO (Generative Engine Optimization). According to Google Search On PDFs: Conversion, Links, HTML, Ranking, Limits, Crawling..., the podcast was summarized and reported on by Barry Schwartz at Search Engine Roundtable.

If you've been wondering whether PDFs are worth publishing for SEO purposes, or how they behave differently from HTML in Google's eyes, this episode answers a lot of those questions directly. And a few of the answers might surprise you.

The Details: What Google Actually Said

The core technical revelation is this: Google does not index PDFs as PDFs. According to the podcast, Google uses a converter component in its indexing pipeline to turn PDF files into an HTML representation. Once converted, those documents are handled by the same internal systems that process regular web pages. That's actually good news for anyone who has PDFs sitting on their site and wondered if they were being ignored.

Links embedded in PDFs are preserved through that conversion and are crawled normally. So if you've linked to another page inside a PDF, that link matters. Images, however, are a different story. Because images embedded in PDFs don't have standalone URLs, they generally can't be processed into Google Image Search the way a standard web image would be.

On the question of AI visibility: PDFs can appear in AI Overviews and are accessible to AI models for training and grounding purposes. The podcast even dedicated a chapter to PDFs for GEO, which Schwartz flagged as the one area that hasn't been covered as thoroughly in Google's prior communications. If you're thinking about how your content shows up in generative AI results, PDF format doesn't automatically disqualify you.

Crawl size limits also came up. Google's crawl limit for PDFs is larger than the limit for standard HTML pages. The exact cutoff isn't documented publicly, according to the podcast, because Google adjusts those thresholds over time as file size norms evolve.

The recrawl story is less encouraging. Google treats PDFs as relatively static binary files, similar to how it treats images, and recrawls them much less frequently than HTML. If you update a PDF and need Google to pick up the changes quickly, the recommended approach from the podcast is to publish it under a new URL or file name rather than overwriting the existing file.

What This Means For You

Picture this: you're an agency and a client asks whether they should keep their white papers as PDFs or convert them to HTML landing pages. This podcast gives you a much cleaner framework to answer that question. HTML is still the default recommendation from Google, because it's easier to edit, mobile-responsive, and faster to index. But PDFs aren't a dead end.

The real problems come from how PDFs are often published, not from the format itself. A few pitfalls worth calling out:

• Scanned or rasterized PDFs: If a PDF is made of scanned images rather than actual text, Google can't read it without OCR. You can test this yourself by trying to highlight and copy text from the file. If you can't, neither can Google.
• Password-protected PDFs: Any password protection blocks Google's bots completely. If you want a PDF indexed, it needs to be publicly accessible.
• Auto-generated CMS print PDFs: A lot of CMS platforms will auto-generate a PDF version of a live web page. According to the podcast, this creates a situation where both versions compete in search results, and the PDF version tends to fall behind because it gets recrawled less often.
• Mobile experience: PDFs require a full download before rendering on mobile and have fixed print layouts that don't reflow when zoomed. That's a real UX problem for users on slower connections.
• Canonicalization: Because identical PDFs are exact binary duplicates, cross-domain canonicalization is hard. The podcast recommends using rel=canonical HTTP headers, filling in PDF metadata fields, and placing links back to your website in the document header to help Google connect the dots.

One specific call-out for AI search: the podcast discussed llms.txt, the emerging convention for helping AI models understand a site's content. The guidance was clear: never format an llms.txt file as a PDF. The whole point of that format is lightweight plain text that machines can read easily. A PDF defeats the purpose entirely.

And yes, the restaurant menu debate got its own chapter in the podcast. The answer was essentially: if building an HTML version of your menu is impractical, a PDF is acceptable. But if you can do HTML, do HTML.

What to Do Now

1. Audit your existing PDFs. Find out which ones are indexed by searching site:yourdomain.com filetype:pdf in Google. Then ask: are these pages where the content matters for organic traffic? If yes, consider migrating them to HTML.
2. Test every PDF for readability. Open each file, try to highlight and copy a block of text, then paste it into a plain text editor. If the text doesn't transfer cleanly, your PDF is likely rasterized and invisible to Google.
3. Remove or block auto-generated print PDFs. Check your CMS settings. If your platform is spinning up PDF duplicates of your web pages automatically, either disable that feature or block those URLs in your robots.txt to prevent the two versions from competing.
4. Update PDFs by publishing a new URL. Don't overwrite an existing PDF file if you need the update indexed quickly. Change the file name or path, update any internal links pointing to it, and let Google discover the new version.
5. Use rel=canonical headers for PDFs you need to keep. If a PDF must live on your site and has a matching HTML page, implement a canonical HTTP header pointing to the HTML version. This tells Google which version to prefer.
6. Keep llms.txt as plain text. If you're building out an llms.txt file for AI discoverability, keep it in plain text format, not PDF.

Background and Context

Google's guidance on PDFs isn't entirely new. Schwartz noted in his summary that Search Engine Roundtable has covered most of the technical ground before: how Google converts PDFs to HTML, how links in PDFs pass PageRank, the crawl file size limits, the duplicate content nuance between PDF and HTML versions of the same content. These positions have been consistent for years.

What makes this podcast different is the AI angle. The GEO chapter and the explicit confirmation that PDFs can appear in AI Overviews and be used for AI model grounding are newer territory. As AI-powered search results continue to take up more real estate on the results page, knowing that your PDF-based content can still show up there is useful context. I've watched SEOs panic about PDF content becoming invisible in AI search, and this episode pushes back on that assumption.

That said, the structural disadvantages of PDFs haven't gone away: slow recrawl rates, mobile rendering problems, canonicalization headaches. The podcast reinforces what most experienced SEOs have recommended for years: default to HTML, use PDFs only where the format genuinely serves a purpose, and manage the technical details carefully when you do.

If you want to track which of your PDF URLs are showing up in search and whether they're pulling traffic compared to your HTML equivalents, site audit tools can surface that data quickly and help you prioritize what to migrate first. It's a cleaner way to make the PDF vs. HTML call than guessing.