Chapter 5. File formats, at the most basic level — extensions, mojibake, scanned PDFs

Key points

• Knowing what a file extension is tells you what you can hand to an AI
• Word and PDF, as they are, are enough to start with
• The one thing to watch: files where the text is really an image

What an extension is

The extension is the part at the end of a filename — .docx, .pdf — that says what kind of file it is.

• .docx — a Word document
• .pdf — a PDF (a document with fixed layout)
• .pptx — PowerPoint slides
• .xlsx — an Excel spreadsheet
• .csv — tabular data (comma-separated; opens in Excel)
• .txt — a plain text file
• .md — Markdown (the star of the second half of this guide)

You may rarely think about it, but knowing this lets you judge whether a file can go to an AI, and how to hand it over.

Word and PDF are fine to start

Be reassured: there is no need to learn exotic formats first.

Most people use Word, PDF, PowerPoint and Excel already. That is enough.

• have the AI read a Word document
• summarise a PDF
• organise the content of a slide deck
• look at the tendency in an Excel file

That alone is very practical. Markdown and Obsidian come later, slowly, once you are comfortable.

One thing to watch: text that is really an image

There is one pitfall worth knowing when handing files to AI:

files where the text is an image

For example:

• a PDF that is only a scan of a paper document
• a screenshot

To your eye there is text on the page; to the computer it is a photograph, not text. So the AI may not read it accurately.

Telling them apart is easy:

Open the file and try to select the text with the mouse.

• selectable → text data, easy for AI to read ◎
• not selectable (the whole thing selects as one image) → it is an image △

Even if all you have is an image PDF, recent AI has become much better at reading text from images (OCR), so it is worth trying. Just be aware that misreadings become more likely.

Try this

Open a PDF you have and try to drag-select the body text. If it selects cleanly, that is a file an AI will read well.

Chapter 5 summary

• Knowing extensions lets you judge what can go to an AI
• Word and PDF are enough at first
• Only files where the text is an image need special care