Developing
Many answers were framed as personal understanding rather than documented fact.
In 60 seconds
- Google converts PDFs to HTML during indexing, after rendering; converted PDFs rank normally and can appear in AI Overviews and AI Mode.
- The Googlebot documentation lists a 64MB fetch limit for PDFs (updated February 3, 2026); the documented HTML cap is 2MB. Neither host recalled the figures.
- PDFs are refetched rarely, so Mueller suggested a new filename for updated files.
- For duplicates, Mueller suggested an HTTP rel="canonical" header or a visible homepage link.
Impact Layer
- What changed
- Google staff have explained how PDFs are handled, while showing how little the figures are known even internally.
- Why it matters
- Many sites publish key content as PDFs, and the limits and canonical options decide whether it is indexed correctly.
- Who is affected
- SEOs and publishers of reports, manuals and whitepapers in PDF form.
- Martech stack area
- SEO & Search, Content
- Stack impact
- Keep PDFs under the documented limit, make sure they have a text layer, and use a new filename when you update an important file.
- Business impact
- Low PPC Land's account of a podcast episode. Several answers were personal understanding, and no recrawl interval or OCR behaviour was given.
- Related vendors
- What to watch
- Updated Google documentation on PDF handling and recrawl timing.
PPC Land’s Luís Rijo reported on October 14, 2026 on episode 115 of Google’s Search Off the Record, “PDFs in Search: Indexing, AI & Pitfalls”, with Martin Splitt and John Mueller. The video page is dated October 13.
What was said
- A converter turns PDFs into HTML during indexing, after rendering. Splitt said converted PDFs rank normally and appear in AI Overviews and AI Mode: “we don’t care if something was a PDF originally.”
- Google’s Googlebot documentation states a 64MB fetch limit for PDFs (February 3, 2026 update). Both hosts said they did not know that figure. The documented HTML cap is 2MB, which Splitt could not recall.
- PDFs are refetched rarely, so Mueller suggested a new filename for updated files.
- Rasterised PDFs lack a text layer. Splitt’s test: select and paste the text.
- Re-hosted copies of a PDF are byte-identical, which complicates canonicalisation. Mueller suggested an HTTP rel=”canonical” header or a visible homepage link.
- Both rejected saving llms.txt as a PDF.
Not stated
Any size limit on the converted HTML, whether OCR is applied to scanned PDFs, a recrawl interval, and whether images in PDFs are processed for image search.
Our read
PDFs are indexed like pages, but they are slow to refresh. Use filenames and canonical headers to manage duplicates.
Sources
- PPC Land: Google podcast hosts can't name the PDF crawl limit it documents as 64MB (opens in a new tab)
Source for the episode's statements, documented limits and gaps.