I used this searx instance to search for “metric chart for 10mm ribbed plastic anchors in masonry”:

https://search.fabiomanganiello.com/search

First hit was this pdf, which names DDG as the source. Since DuckDuckGo is merely a boot-licking Microsoft syndicate, MS was really the one who crawled and indexed that PDF – which is not OCRd! It’s interesting that the preview text on the search page shows the text that we can see in that pdf. When I fetch the pdf and run pdftotext, there is nothing.

Anyway, thought this was notworthy. Bing is going to the trouble of OCRing PDFs to build their index, but the OCR effort is somewhat wasted because users who fetch the PDF don’t get the text.

So when a raster PDF traverses Microsoft’s e-mail servers, it stands to reason that they may be surreptitiously doing OCR on those scanned docs as well, for snooping driven by ad surveillance.