Your PDF brochure is invisible to AI assistants. Here is what to do with it

Picture a machine shop whose website says, in a PDF linked from the footer, on page 14, that it machines stainless steel. Nowhere else. An assistant asked whether they do stainless steel says no, because the only page it could read was the homepage, and the homepage talks about passion for precision.

We see some version of this in a large share of the SME sites that run an analysis here. The company has written down exactly what it does, carefully, with specifications. It has put that text in the one container that the systems answering questions about it are least likely to read.

Why PDFs get skipped

It is not that AI systems cannot read PDFs. They can, and when a user uploads one, they do it well. The problem is retrieval. When an assistant answers a question with web search, it gets a list of candidate pages from a search index, fetches a few, and extracts the text. At each step a PDF is a worse candidate than an HTML page:

• Search indexes rank PDFs lower for most queries, and often do not index them at all when the file is large or image-based.
• A 40-page brochure is one URL. The assistant has a limited budget of text per source; it takes the first part and stops. Page 14 does not exist.
• Scanned brochures and design-heavy exports have no text layer at all. The crawler sees a sequence of pictures.
• PDFs carry no structured data, no headings a parser trusts, no author, no date. Everything that helps an engine decide a page is relevant and reliable is missing.
• The typical brochure PDF is also out of date, because updating it means asking the agency.

So the text that most precisely describes the business is the text least likely to be in the answer.

The two-hour fix

You do not need to delete the PDF. People print it. You need the same facts to exist as HTML pages.

Pull the facts out. Open the brochure and list every statement a customer might ask an assistant about: services, materials, capacities, certifications, delivery areas, minimum order, lead times, the team. For most SMEs this is twenty to forty sentences.

Give each topic a page, or a section with its own heading. "Stainless steel machining" gets a page whose first paragraph says you machine stainless steel, which grades, to what tolerances, for which industries. The specification table from the PDF becomes an HTML table. If this feels like a lot of pages, it is the right number; the brochure had them too, as chapters.

Put the date and a name on them. Who is responsible for this information, when was it last checked. Engines weigh that; so do customers.

Link the PDF from the page, not instead of it. "Download the full brochure (PDF, 4 MB)" under the HTML version. The PDF becomes a convenience, not the source.

Then check the file itself. If you keep PDFs online, export them with a text layer, give them a real title in the document properties, and keep them under a few megabytes. It makes them a slightly better candidate when nothing else matches.