Google Explains How It Handles PDFs in Search and AI

Google has explained how its search engine handles PDF files, in a Search Off the Record podcast episode published on October 13, 2026. It converts PDFs to HTML during indexing, so they can rank and appear in its AI features. Google still recommends HTML. This matters for any site that publishes menus, catalogs or price lists as PDFs.

What we know

Who and where: Martin Splitt and John Mueller of Google's Search Relations team, in the Search Off the Record episode "PDFs in Search: Indexing, AI & Pitfalls," dated October 13, 2026, according to its official page.
The conversion: the episode page says Google converts the PDF, a binary container, into an HTML page during indexing, so the rest of its pipeline can process it normally.
Search and AI: according to the same official summary, a PDF's indexable text becomes HTML, which is why it can rank normally and appear in AI features. One chapter asks whether PDFs show up in Search, AI Overviews and AI Mode, and the episode covers their use in AI systems for training and grounding.
HTML is better: the official summary says HTML is more mobile-friendly, reflows when users zoom, and loads faster.
A simple test: copy text out of the PDF and paste it into a text editor. If nothing copies, the page may be an image and Google cannot easily index it, the episode page says.
Restaurant menus: a PDF menu is better than no menu, but a responsive HTML page performs better, is more accessible and is easier to use, according to the official summary.
The same PDF on several sites: the episode notes that identical PDFs on different sites make canonicalization harder, and that a rel="canonical" HTTP header helps.
llms.txt: saving it as a PDF is a bad idea, the episode page says.
What one outlet adds: according to Search Engine Roundtable's summary, text links in PDFs are crawled; embedded images have no URL of their own and generally do not reach Google Images; Google recrawls PDFs less often than HTML; password-protected PDFs cannot be read; and CMS-generated "print" PDFs can compete with the HTML page. The same outlet says little was said about optimizing for AI search.

The official transcript could not be opened for this story, so no direct quote from Splitt or Mueller is presented here as read.

What changes and what doesn't

There is no new rule. Google's documentation already lists PDF among the file types it can index, and its Googlebot page says it reads the first 64MB of a PDF, compared with the first 2MB of other supported file types.

What is new is that Google explains it publicly and ties it to AI: a well-built PDF can end up cited in an AI summary. Google also states that, for users, HTML is still the better choice.

How to check if this affects you

Find your PDFs. In Search Console, open the Performance report and filter by page with URLs containing ".pdf". You will see which ones get impressions and clicks.
Run the copy test. Open each important PDF, select a paragraph and paste it into a text editor. If nothing appears, it is a scanned image: export it again from the original document.
Move what people search for to HTML. A menu, a price list or opening hours work better as a web page with visible text. Keep the PDF as a download, not the only version.
Check the file size. Google only reads the first 64MB of a PDF. A very large catalog is better split or compressed.
Handle duplicates. If the same PDF lives on your site and on others, Google's documentation lets you declare the canonical URL with a `Link` HTTP header using rel="canonical". Whoever manages your server or hosting sets it up.
Keep out what shouldn't be in Google. A PDF has no robots meta tag. Google says to use the X-Robots-Tag HTTP header with "noindex" and gives Apache and NGINX examples.

Sources

Updates: if Google changes its PDF documentation or the transcript adds new details, they will be added here with a link.