Teaching · Practical guide
Use AI to Make Learning Materials More Accessible
Adapt a learning resource without losing its meaning. Use a worked instruction rewrite, a text alternative and a practical accessibility review.
Preserve the learning goal while removing barriers
Start by naming what the learner must do. If the outcome is interpreting a chart, an unnecessarily complicated instruction should not become a separate test of advanced English. If the outcome is explaining a technical term, however, removing that term may remove the learning itself. Write down the content that must survive every adaptation: quantities, units, conditions, sequence and required evidence.
CAST offers a framework for considering varied ways of engaging with, accessing and expressing learning. Use it to ask where a barrier occurs, not to classify people into fixed learning types. A clear written version may help many learners; it does not replace an individual adjustment agreed through the institution’s support process.
Prepare a controlled adaptation request
Supply a short piece of material that you are permitted to use. Tell the AI who the resource is for using general prerequisites, not disability records or personal histories. Ask it to propose a simpler version, a glossary and a list of changes. Keep the original next to the result so you can compare meaning rather than just deciding which version sounds friendlier.
Set a boundary: the tool must not add deadlines, change numbers, remove warnings or infer a learner’s needs. A request such as “make this suitable for everyone” is not testable. A request to split a long instruction into numbered actions while preserving its conditions is specific enough to review.
Worked example: simplify without changing the task
Original fictional instruction: “Having inspected the three observations, determine the proportion of unsuccessful transmissions relative to the entirety of transmissions, present the result as a percentage to one decimal place, and distinguish observed variation from a demonstrated causal effect.”
Revised version: “1. For each observation, divide unsuccessful transmissions by all transmissions. 2. Multiply by 100. 3. Write each answer as a percentage with one decimal place. 4. Compare the results. Explain why a difference alone does not show what caused it.” Glossary: proportion means a part compared with the whole; causal means showing that one factor produced a change.
Check with fictional values: 3 unsuccessful out of 80 gives 3/80 × 100 = 3.75%, which rounds to 3.8%. The rewrite retains the denominator, percentage conversion, rounding and causal limitation. A version saying “count failed transmissions” would lose the central calculation and should be rejected.
Make essential visual information available as text
For a chart showing 3.8%, 5.0% and 2.5% across three observations, provide those values in an accessible table as well as a short description of the comparison. Do not rely on red and green alone to identify high and low results. W3C distinguishes image alternatives by the image’s purpose; a decorative image and a chart carrying essential data need different treatment.
An AI-generated description must be checked against the actual chart. “Performance improved steadily” is wrong for this sequence, because the middle value is higher than the first. A suitable description is “The unsuccessful-transmission percentage is highest in observation 2 and lowest in observation 3.” Keep the underlying values available so learners can make their own comparison.
Review the delivery format
Headings should describe the section and follow a meaningful structure. Keep instructions as real text rather than text embedded only in a picture. When preparing video, review captions and technical vocabulary rather than accepting automatic transcription unchanged. If a learner must navigate controls, test those controls with a keyboard in the actual learning platform.
A document that looks attractive can still have a poor reading order. Review the exported format, not only the editing window. Offer an equivalent route to the same objective when a particular format is inaccessible, while following the institution’s adjustment procedures. AI can help draft alternatives, but it cannot certify that the final resource meets every learner’s requirements.
Practice: a 20-minute adaptation and answer check
Spend four minutes selecting an instruction and marking its non-negotiable meaning. Use five minutes to request a plain-language version and glossary. Spend six minutes checking every condition, number and technical term. Use the last five minutes to add a text alternative for any essential visual and test the exported resource.
Example answer for “Compare 6 failures out of 120 with 3 failures out of 50”: both rates are needed. They are 5% and 6%; the first has more failures by count but a lower failure percentage. An accessible rewrite must preserve the two different totals, otherwise it changes the comparison. A text table with total, failures and rate allows the calculation to be inspected.
Self-review: did the objective and factual meaning survive? Can the essential information be accessed without relying only on colour, sound or an image? Did you check the delivered format and identify any adjustment requiring specialist support? Save the original, revised version and one sentence explaining each substantive change.
Reusable prompt
Select and copy this template, or customise your own using the free builder.
Adapt this approved learning instruction for [learner prerequisites]. Preserve every quantity, unit, condition, warning and required learning action. Return numbered instructions, a brief glossary, a list of changes and checks for meaning. Suggest a text alternative for essential visual information, but do not invent unreadable values or claim accessibility certification. Material: [non-sensitive text].Open the prompt builder →
Pause · Think · Check
Test your understanding
Choose one answer per question, then check your reasoning. This is practice; scores are not saved.