Google’s Martin Splitt and John Mueller just posted a Search Off the Record podcast on the topic of PDFs, it is named PDFs in Search: Indexing, AI & Pitfalls. It goes over how PDFs are converted for crawling and indexing, how Google processes links, images, what are the file limits and do they rank in search and AI search.
The podcast is broken into these chapters:
- 00:50 – Can you save an LLMs.txt file as a PDF?
- 01:56 – Plain text vs. Binary container: What makes PDFs different from HTML?
- 03:08 – Inside Google’s converter pipeline: Turning PDFs into HTML for indexing.
- 05:08 – Extracting images and links from PDF files.
- 06:04 – Crawling limits, file sizes, and testing cut-off thresholds.
- 08:35 – When does it make sense to publish a PDF?
- 09:33 – Do PDFs rank in Search, AI Overviews, and AI mode?
- 11:22 – Technical pitfalls: Rasterized pages, scanned images, and accessibility.
- 12:53 – Mobile connection delays, layout locking, and zoom issues.
- 14:12 – Canonicalization challenges with duplicate “binary blobs”.
- 15:57 – Metadata, password protection, and updating PDF files.
- 18:59 – The problem with dynamic CMS PDF generation.
- 21:08 – The restaurant menu debate: PDF vs. HTML web pages.
- 23:00 – PDFs for GEO (Generative Engine Optimization) & AI grounding.
Truth is, we covered most of this stuff, maybe outside of the GEO/AI search aspect, which Google didn’t say much about.
Here is our old coverage of what Google said on PDFs, I am sure I am missing stuff:
Here is how I’d sum it up after listening to it twice and reading the transcript:
How Google Processes and Ranks PDFs:
- HTML Conversion: PDFs are binary container files rather than plain text. During the indexing pipeline, Google uses a converter component to turn PDFs into an HTML representation, allowing internal search systems to handle them like regular web pages.
- Links and Embedded Media: Because PDFs are converted to HTML, embedded text links are preserved and crawled normally. However, images embedded in PDFs lack standalone URLs, so they generally cannot be processed into Google Image Search the same way as standard web images.
- Search and AI Visibility: Converted PDFs rank and index using the same fundamental systems as standard HTML documents. They can appear in AI Overviews and are accessible to AI models for training and grounding.
- Crawl Size Limits: The crawling limit for PDFs is larger than the limit for standard HTML pages. Google deliberately does not document an exact cutoff because crawl thresholds are adjusted over time as file size norms evolve. Although, Google does document that.
Technical Pitfalls and UX Challenges:
- Rasterized and Scanned Documents: If a PDF is composed of scanned images or rasterized pages rather than text, search engines cannot read the content without OCR. Site owners can test this by highlighting, copying, and pasting text from the PDF into a text editor.
- Mobile and Accessibility Issues: PDFs must download completely before rendering on mobile devices and have fixed print layouts that do not reflow text when zoomed. They can also present barriers for screen readers if not structured accessibly.
- Canonicalization and Duplication: Because identical PDFs are exact binary duplicates, cross-domain canonicalization is difficult. Publishers can mitigate duplicate content issues by using rel=canonical HTTP headers, populating PDF metadata fields, and placing prominent links back to their website in the document header.
- Access Restrictions: Password protection blocks search engine bots from reading or indexing PDF content.
Crawl Frequency and Updates:
- Slower Recrawl Rates: Google treats PDFs as relatively static binary files (similar to images) and recrawls them much less frequently than HTML.
- Changing File Names for Updates: Because recrawls take longer, publishers needing to index an updated PDF quickly should publish it under a new URL or file name rather than overwriting the existing file.
- Avoiding Auto-Generated “Print” PDFs: Automated CMS tools that generate PDF duplicates of live web pages can cause both versions to compete in search results, leaving the PDF version outdated because it is recrawled less often.
SEO Recommendations:
- Default to HTML: HTML is easier to edit, responsive on mobile screens, and faster to index.
- Use PDFs Selectively: PDFs are appropriate when content requires rigid print formatting, downloadable offline forms, or paper documentation. For small businesses like restaurants, a PDF menu is acceptable if creating an HTML equivalent is impractical.
- Keep llms.txt Plain Text: Files intended for machine consumption (like llms.txt) should never be formatted as PDFs, as the entire purpose of the format is to provide simple, lightweight plain text.
Forum discussion at X.
