AI Course Import Now Keeps the Images in Your Word Document

If you have used AI course import on a document full of screenshots and wondered why the finished course was a wall of text, this post is the explanation and the apology. The import read your document as plain text. Every image in it was discarded at the door, before the AI ever saw the document.

What was actually going wrong?

There were two import paths and they behaved differently. The tagged import — where you mark up a Word document with tags like [image] — handled images properly. The AI import, which reads an ordinary unmarked document and works out the structure for itself, extracted only the text.

The machinery to place images already existed. The AI was capable of positioning images, and the validator would accept them. Nothing ever supplied the list of images, so every image the model tried to place was silently dropped. Your diagrams were not lost in conversion; they were never in the conversation.

What happens to images now?

Images embedded in your Word document are extracted, stored, and written into the text at the exact position they occupied in the original. The AI sees where each image sat in relation to the surrounding paragraphs, and places it back near the content it illustrates.

That positional detail matters more than it sounds. A bare list of images tells the model nothing about which section a diagram belongs to, so it would either scatter them or ignore them. Seeing an image marker between two paragraphs is what lets it put the picture back where the author had it.

Will it add images I did not supply?

Where a module has no images of its own, the AI can now request one stock photograph on a concrete subject drawn from the content — the activity being described, the workplace, the object under discussion. Never more than two in a module, and it is asked to skip the image entirely rather than reach for a generic photograph of people smiling around a laptop.

Any stock image is fetched, re-hosted on our storage so it cannot disappear later, and credited in the caption. You can remove any of them in the preview before the course is created, and all of them afterwards in the editor.

Do I need to change how I write my documents?

No. Insert images in Word the way you normally would and they will come across. A few things help:

• Place images near the text they illustrate. The AI uses that position, so an image dumped in an appendix will end up in an appendix.
• Keep them a sensible size. Very large images make the file heavy without improving the course. Anything over 8MB in a single image is skipped.
• Watch the total file size. Uploads are capped, and an image-heavy document reaches that cap quickly.

What if I do not want stock images?

Delete them. The preview shown before the course is created now includes a thumbnail strip of every image in each module, with a remove button on each one. Nothing appears in a finished course that you have not had the chance to look at first.

What else changed at the same time?

Fixing images meant measuring the import properly, and that surfaced a larger problem: documents over roughly four and a half thousand words were being silently truncated, producing courses with sections quietly missing. That is fixed too. A 7,500-word document that previously took twelve minutes and lost two modules now converts in about three minutes with everything intact.

How do I check what actually came across?

Look at the preview before the course is created rather than afterwards. It now lists every module with a thumbnail strip of its images, so you can see at a glance which modules got pictures, which did not, and whether anything obviously wrong slipped through.

Two things are worth a second look. First, images that were decorative in the original document — logos in headers, dividing lines, background elements — occasionally survive as content images and are better removed. Second, the alternative text is inferred rather than written, so anything a learner genuinely needs the image to understand deserves a proper description in the editor.

What this means for your next import

If you tried AI import before and got a text-only course from a document full of diagrams, try it again. Nothing about your document needs to change. If you have documents you did not bother importing because you assumed the images would not survive, they will now.