Can ChatGPT and Google read a PDF catalogue at all?
They can read a text PDF, but it is a weak home for the facts that win enquiries. Google indexes PDFs that contain real text. The trouble lies in three places: PDFs that are scans with no text, facts buried on page 37 of a 60-page file, and the slow rate at which search engines come back to check whether a PDF has changed.
So the title of this article needs a qualifier. ChatGPT is not blind to your catalogue. But a fact that sits as text on a web page is easier for any reader, software or human, to find, quote and keep current than the same fact inside a download.
EbizIndia's study of 30 Indian manufacturer websites, read in October 2026, found a downloadable brochure, catalogue or datasheet on 11 of them. On many factory websites, that file is where the size range, grades and standards actually live.
What did Google say about PDFs in Search Off the Record?
Google's search team discussed PDFs at length in an episode of its podcast, Search Off the Record, titled "PDFs in Search: Indexing, AI & Pitfalls", with Martin Splitt and John Mueller. Barry Schwartz's write-up on Search Engine Roundtable, published on 13 October 2026, summarises the main points:
- Google converts PDFs to HTML for indexing, treating them as binary container files rather than plain text.
- PDFs are recrawled much less often than HTML pages, because they are treated as relatively static files.
- Scanned or rasterised PDFs cannot be read without text recognition (OCR). The simple test is to try copying text out of the file.
- Images embedded in a PDF have no web address of their own, unlike a photo placed on a web page.
- On phones, a PDF must download fully before anything shows, and it can be hard for screen readers unless built with accessibility in mind.
The write-up recommends defaulting to HTML and keeping PDFs for documents that need fixed print layout, offline forms or paper records. That recommendation is the author's reading of the episode rather than a quoted Google rule, but it fits Google's wider advice. Its guide to generative AI features says Google may discover, crawl and index many kinds of files in addition to HTML, but the facts you most want quoted are safest on the page itself.
Why does an updated price list or spec sheet stay out of date in AI answers?
Because the old version is what was last fetched. If you upload a revised catalogue with a new pressure rating or a wider size range, and the PDF is recrawled much less often than your web pages, the search index can keep the old figures for a long time. AI answers built on that index repeat them.
A web page that states the same facts gets rechecked more often and, if it carries an updated date, tells readers and crawlers that something changed. A new PDF uploaded under the same file name can easily go unnoticed, because nothing on the page says it changed. If the facts must live in a PDF for now, at least announce the revision on the page that links to it.
What does each format hide, and what fixes it?
Each common format hides something different, and each has a plain HTML fix. The table sets them out.
| Format | What it hides | HTML version that fixes it |
|---|---|---|
| Text PDF catalogue | Facts buried deep in a long file; slow to refresh in the index | One product page per product line with its own spec table, linking to the PDF as a download |
| Scanned PDF catalogue | Everything: there is no text to read | Retype the specs as page text; replace the scan with a text PDF made from the original file |
| Spec table as an image | Every number in the table | A real HTML table with headings for size, rating, material and standard |
| Certificate scan | Standard, number, issuer and validity date | Those four details typed beside the image |
| Datasheet behind an enquiry form | The whole datasheet: crawlers do not fill forms | Key specs on the open page; keep the form for drawings or detailed quotes |
| "Price on request" | Any sense of cost, minimum order or lead time | A price band or price basis, minimum order and typical lead time in text |
The certificate row has its own guide on showing certificates as text, and the product page layout is set out in what an industrial product page needs.
What does the fix look like for one product?
Take a made-up 50 mm ball valve. In the before state, the product page shows a photo, one line ("Ball valves in all sizes, contact us for details") and a link to "Catalogue.pdf", a 14 MB scan of the printed brochure. An assistant asked "who makes 50 mm CF8M ball valves rated to Class 300?" finds nothing on that page to match.
In the after state, the page states in text: "Two-piece and three-piece ball valves, 15 mm to 200 mm, bodies in WCB and CF8M, pressure Class 150 and Class 300, flanged and screwed ends", followed by an HTML table of sizes and ratings, the standards the valves are made and tested to, and the lead time. The PDF link stays at the bottom, now a text PDF exported from the design file. Every fact the buyer might ask about is now on the page, with nothing to download.
When is a PDF still the right format?
When the reader needs to print, sign, file or forward it. A full catalogue for a distributor, an installation manual, a test certificate a buyer must attach to a file, or a form filled in offline all suit PDF. The rule is about order of priority, not removal: the web page carries the facts, and the PDF repeats them for those who want a document.
A checklist for whoever maintains your site:
- List every PDF on the site and test each for copyable text.
- Replace scanned PDFs with text PDFs exported from the original design files.
- For each product PDF, make sure its specs, standards and size range also appear as text on the matching product page.
- Rebuild image spec tables as HTML tables.
- Move datasheets out from behind forms, or put their key figures on the open page.
- When a catalogue is revised, update the web page too and show the revision date.
- Give each PDF a descriptive file name and link text, such as "Ball valve catalogue 2026 (PDF, 3 MB)".
Which facts to move first is covered in the factory facts AI assistants look for, and why they matter before a buyer ever visits is the subject of your website now has two visitors. EbizIndia has a shorter note on product facts kept only in a PDF catalogue, and the AI Visibility section has the full series.
Questions founders ask
Does Google index PDF files?
Yes. Google converts text-based PDFs into HTML for indexing, so their words can be found in search. Scanned or image-only PDFs are a different matter: without a text layer there is nothing to index unless text recognition succeeds. Google also recrawls PDFs much less often than ordinary web pages.
How can I tell whether my PDF catalogue is a scan?
Open it and try to select a line of text, such as a row in a spec table, then copy and paste it into a note. If nothing highlights, or the pasted text is garbled, the page is an image. Catalogues made by scanning printed brochures are usually like this.
Should I remove the PDF catalogue from my website?
No. Buyers still forward, print and file catalogues, and a PDF does that job well. Keep it, but make sure every important fact in it also appears as text on the matching web page, and link to the PDF from that page as a download.
Our prices change often. Should we publish them?
Publish what you can stand behind, even a price band or a basis such as "ex-works, per tonne, valid this month". If prices cannot be published at all, state the factors that set them, the minimum order and the typical lead time. That gives an assistant something to say beyond "price on request".