Google on PDFs: converted to HTML, recrawled less often

Martin Splitt and John Mueller devoted the 13 October 2026 episode of Search Off the Record, PDFs in Search: Indexing, AI & Pitfalls, to PDFs. Search Engine Roundtable’s Barry Schwartz summarized it. We did not listen to the episode, so the points below are the episode description and Schwartz’s account.

How does Google index a PDF?#

A PDF is a binary container, not plain text. Google’s pipeline converts it into an HTML representation so its systems can handle it like a regular page. Text links in a PDF are preserved and crawled. Images embedded in a PDF have no standalone URL, so they generally cannot show up in Google Images.

Do PDFs rank like web pages?#

Converted PDFs use the same core systems as HTML pages, per the report, and can appear in AI Overviews and be used by AI models for grounding. Splitt and Mueller still advised defaulting to HTML, and keeping an llms.txt file as plain text, not a PDF. Our AI search hub covers how those answers are built.

Which PDFs cause problems?#

• Scanned or rasterized PDFs without OCR are hard to read. A quick test is to copy text out of the file.
• PDFs must fully download before rendering on a phone, and fixed layouts do not reflow when zoomed.
• Identical PDFs are exact binary duplicates, which makes canonicalization across domains hard. Mitigations named were rel=canonical in an HTTP header, metadata fields and links back to the publisher’s site.
• Password-protected PDFs block search engine bots.

How often are PDFs crawled?#

Schwartz reports crawling is much less frequent than for HTML. To get an updated file indexed quickly, publish it under a new URL. PDFs get a larger size limit than HTML pages, but Google does not publish an exact cutoff because thresholds change as file sizes change.

When should you still use a PDF?#

The speakers’ guidance, as reported, is to use a PDF for rigid print formatting, offline forms or paper documentation, and HTML otherwise. Auto-generated print PDFs of web pages can compete with the HTML version and go stale, so they deserve a check. See our indexing and technical SEO hubs for the wider picture, and the earlier Mueller sitemap answers from the same podcast.