Format and Extraction

The PDF Catalogue and AI Reading

Yes, an AI system can find your PDF catalogue, because, according to Google Search Central, PDF is one of the indexable file types alongside Word, Excel, PowerPoint, HTML, XML and CSV. The real question is not whether it gets indexed — it is whether it gets read cleanly. A scanned page, a specification table exported as an image, or text with no selectable characters can sit inside an indexed file and still be invisible to extraction. The problem is structure, never the format by itself.

Diagnose My Technical Content All questions on Industrial Knowledge Architecture

What Actually Breaks Extraction

Scanned page

An image of text with no selectable characters underneath

Table as image

A specification grid exported as a picture instead of real rows and columns

Indexable

According to Google Search Central, PDF is a supported, indexable file type

No headings

Visual formatting that looks like a heading but carries no semantic tag

01The Direct Answer

Can AI read the PDF catalogue on my site?

Direct Answer

It can find it. Whether it can extract a usable answer from it depends on how the PDF was built, not on the fact that it is a PDF.

According to Google Search Central, Google indexes PDF (Adobe Portable Document Format) alongside Word, Excel, PowerPoint, HTML, XML and CSV. That settles the indexability question. What it does not settle is quality of extraction: a PDF made of scanned images, with a specification table pasted in as a picture, gives a crawler or a language model nothing to parse, even while the file itself sits in the index.

02Google Does Index PDF

Retire the wrong argument before making the right one

"Google cannot read PDF" is a claim that circulates in industrial marketing and it is not true. Google Search Central lists PDF explicitly among the indexable file types. Making the wrong argument — that the format itself is invisible — leads to the wrong fix, usually a costly full rebuild of the catalogue as HTML pages, when the actual fix is often cheaper: rebuilding the same PDF with real text and real tables instead of scanned images.

Keep the argument precise: the format is not the barrier. The barrier is whatever prevents a machine from extracting text and structure from the file, and that barrier can exist inside a PDF, inside an HTML page, or inside any format at all.

03The Real Problem

Structure and extraction, not visibility

What to Check in Your Own Catalogue

1. Can you select the text?

Open the PDF and try to select a sentence. If nothing highlights, the page is an image, not text.

→

2. Is the specification a real table?

Try to copy a row of numbers into a spreadsheet. If it pastes as one block instead of separate cells, it is a picture of a table.

→

3. Does a heading behave like a heading?

Bold, larger text that carries no semantic tag reads as a paragraph to a machine, not as a section break.

These three checks take less than five minutes and usually explain more than switching formats ever would. When the file is read but the assistant still states wrong facts about the company, see the AI gets facts wrong about my company.

04Not Every PDF Is Equal

A native PDF and a scanned PDF are different files

Native PDF

Exported from a document, real text and tables

Selectable text, structured tables, sometimes tagged headings. Extraction quality can approach a well-built HTML page.

Scanned PDF

A photograph or scan of a printed page

An image saved with a .pdf extension. Indexable as a file, unreadable as text, unless optical character recognition was run and the result was embedded.

Most catalogues in industry accumulate both kinds over the years, and the first useful task is simply sorting which datasheet is which. That sorting decides where to spend effort first.

05What to Do with the Catalogue

Fix the worst offenders, then decide the format

Start by re-exporting the scanned datasheets as native PDFs with real text and real tables — a fix that touches the file, not the platform it lives on. Once the worst offenders are fixed, the separate question of whether to publish the specification as HTML instead of PDF is worth its own answer.

See that comparison in datasheet in HTML or in PDF, or read a full walkthrough of building a catalogue that converts in how to create a technical catalogue that sells, or the technical catalogue guide.

06FAQ

FAQ

Can AI read the PDF catalogue on my site? +

It can find it, according to Google Search Central, which lists PDF among the indexable file types together with Word, Excel, PowerPoint, HTML, XML and CSV. Whether it can extract a usable answer depends on whether the PDF has real, selectable text and structured tables, not on the format itself.

Is it true that Google cannot read PDF? +

No. According to Google Search Central, PDF is explicitly listed as an indexable file type. What limits extraction is not the format but whether the specific file has scanned images instead of real text, or tables built as pictures instead of rows and columns.

How do I know if my datasheet is a scanned image? +

Open the file and try to select a line of text with your cursor. If nothing highlights and you cannot copy it, the page is an image saved as a PDF, not real text a machine can read.

Should I run OCR on old scanned catalogues? +

It is one option, and better than leaving the page unreadable, but optical character recognition on a low-quality scan often introduces errors in numbers and units. For a specification table, re-typing the values directly from the source document is usually safer than trusting OCR on a technical figure.

Is a specification table exported as an image a problem? +

Yes. A table saved as a picture cannot be parsed into rows and columns by a crawler or a language model, even inside an otherwise well-built PDF. The fix is rebuilding the table as real, structured content in the source file before exporting again.

Do I need to rebuild the whole catalogue as HTML? +

Not necessarily, and often not first. Fixing scanned images into native PDFs with real text is usually cheaper and faster than a full HTML rebuild, and it solves most of the extraction problem. Whether HTML is worth it for your highest-traffic datasheets is a separate decision.

Does the file name or URL of the PDF matter? +

A descriptive file name and a stable, linkable URL help a crawler and a person find the file, but they do not fix an extraction problem inside the document. Naming is worth doing well and is not a substitute for real text inside the PDF.

Where should I start fixing my catalogue? +

Start with the three checks: can you select the text, does a table copy into separate cells, and does a heading carry real markup. Fix the datasheets that fail all three first, since those are the ones a machine is currently unable to read at all.

Find out which datasheets are unreadable to a machine

The diagnosis checks the catalogue file by file: selectable text, real tables and structured headings.

Check My FunnelData first, media afterwards