Accessibility has to be built, not retrofitted

And it improves the course for everybody.

Captions, keyboard operability, sufficient contrast, meaningful text alternatives and a logical heading structure are the baseline. Each is cheap when designed in and expensive to add afterwards, particularly captions on a large video library and keyboard operation of interactions built with a mouse in mind.

The learners who need these are not a small minority, and the provisions help far more people than they were designed for. Captions are used heavily by people watching in noisy environments or in a second language. Clear headings help anybody skimming. Keyboard operation helps anybody on a slow connection or an unusual device.

The check that catches most problems takes an afternoon: operate the whole course using only a keyboard, run it through an automated checker, and view it at high zoom on a phone. None of that requires specialist expertise and it finds the majority of practical barriers.

Obligations differ by jurisdiction and by the standards a provider operates under, and where a course is delivered to the public or under a government arrangement there may be specific requirements. Those should be established from their own sources rather than inferred from general practice.

It is worth adding that accessibility work frequently improves the course structurally. Writing meaningful text alternatives forces somebody to articulate what each image is for, and a surprising number turn out to be decorative and removable. The same applies to heading structure, which exposes sections that are not actually about anything.