Writing With AI for Instructional Design

Learning designers are increasingly using AI tools for a wide range of instructional design tasks— summarizing content, brainstorming scenarios, generating images, and creating assessment items. Currently, I’m focusing on the many ways to use AI tools as writing assistants, since we are often responsible for many different types of writing. Here are some of the most helpful strategies I use for writing with AI for instructional design tasks.

1. Use Different Prompt Types for Different Results

Voice Prompts: Voice prompts feel very natural when using a mobile device but also work well on a laptop. They make AI more accessible to a broader audience. Speaking is faster and more conducive to a natural flow of language than typing. Therefore, voice queries may provide a richer context than typed queries. Voice is ideal for stream-of-consciousness prompting, thinking aloud, and spontaneous requests. Be prepared for longer and more dynamic conversations than with structured and multistep prompts.

Structured Prompts: This is the type of prompt you may have first learned to use. A structured prompt provides all the necessary information at once. For example, when asking for a summary, you’d include details like the purpose, context, intended audience, level of detail, format, and tone. While they take time to construct, structured prompts often produce clear, usable responses on the first try.

Multistep Prompts: You may have discovered that chatbots seem to have small working memories like humans. If you provide too much information at once, part of your request may slip through the cracks. If that’s the case, then breaking your queries into smaller steps is a good solution. Break your request into smaller steps to guide the AI and give yourself time to reflect and iterate. Multistep queries avoid overwhelm, give you time to think things through, and allow you to easily iterate. You will end up with multiple responses, but chatbots seem to be able to pull a cohesive response together at the end.

2. Fill in the Gaps with AI Research Tools

You may get cryptic PowerPoint slides from a subject matter expert (SME) that require an advanced medical or engineering degree to understand. Or you may find large gaps in the content you’re working with that you need to close. AI research tools will summarize, clarify, and fill in the gaps. Remember to get permission to upload documents if you want summaries.

If you need to research academic journals, many AI apps will speed up your search, summarize the results, and provide a bibliography. Here are my two favorites so far:

• Elicit AI: Elicit is a top-notch AI research tool with a free plan. It will send you a report in response to your research question. It may help you refine your query for better results.
• Consensus: Consensus is another excellent AI research tool with a free plan that allows you to filter by date and topic. It provides a summary of the research results and a list of downloadable citations. Consensus offers follow-up prompts to help you go deeper into your research.

Some people are using NotebookLM for research. This AI tool from Google does a good job of summarizing papers. It will also create an audio summary in a conversational form. However, I’ve found errors in the summaries and citations, which I discuss here: Five Easy Ways Instructional Designers Can Use NotebookLM.

Can you trust the results? As with all AI tools that generate reports and summaries, double-check for accuracy. If you don’t have access to the journal, review the abstract to see if it aligns with the AI summary.

3. Get a Deeper Response

Have you noticed that chatbots will often default to bland and superficial answers? There are several approaches to getting a fuller and deeper response.

• If you are a paid subscriber to ChatGPT, select the “Deep research” button below the prompt box. Read more about “deep research.”  This feature accesses and summarizes research. You have a limited number of deep research requests depending on your subscription type.
• Perplexity and Gemini also have deep research capabilities. We can probably assume that all LLMs with conversational features will be heading in this direction.

Or try these prompt strategies:

• “That was a superficial Level 1 response. Now give me Level 2 and Level 3 responses.” You can continue asking for deeper versions. In my experience, the responses beyond Level 3 may take you down an unrelated path or seem nonsensical. Try it and see.
• You can also be more specific about what you’re seeking. “The audience is already familiar with these concepts. Can you provide an advanced version for an experienced audience?”
• Try this, “Explain [insert topic] at a graduate school level.”

I frequently ask chatbots for evidence and citations of the findings they present. Except for research-oriented AI tools, the citations are not reliable.

4. Engage Writing Assistance

I believe in writing from the heart, so I don’t use AI to write full articles or courses. But I do use it for proofreading, tightening awkward sentences, and trimming long paragraphs. I subscribe to Grammarly to catch writing errors and have also experimented with ChatGPT Canvas (the results are often corrected by Grammarly). Other tools to check are ProWritingAid and Hemingway Editor.

I never take all of their suggestions. In fact, I maintain a critical eye on their recommendations, as AI writing assistants can take away your personality and voice. For example, Grammarly frequently asks, “Want to sound more confident?” Actually, I do not. I want to sound like myself!

5. Rephrase for Accuracy

When writing a complex explanation, you need accurate and precise wording. Gaining clarity is another use of writing with AI for instructional design. A thesaurus is wonderful for finding a word that captures the connotation you want. But AI apps can help you revise phrases and sentences to express the nuance you’re seeking. The prompt strategies below are helpful.

• Get very specific about the audience, tone, and context of the phrase or sentence rather than asking, “Can you rephrase this?”
• Request many revisions when you can’t quite identify what you want. “List 10 conversational and informal versions of this sentence.” Then, select and rearrange parts of different responses to fit your needs.
• If you’re not allowed to simplify terminology for legal or other reasons, be specific about the words to keep. “Make this sentence more concise, but keep the legal terms.”
• Be specific about the reading level. “Revise this paragraph for an 8th-grade reading level without changing its meaning.”

6. Analyze for Reading Level

Assessing reading level is another powerful use case for writing with AI for instructional design. Depending on your audience, you may need to simplify sentence structure and vocabulary.

Prompt your AI writing assistant to use any of the following metrics below. Then, upload a document or paste the text and ask for an analysis.

• Flesch-Kincaid Grade Level – Calculates the approximate U.S. grade level needed to comprehend the text based on sentence length and syllables per word.
• Flesch Reading Ease – Scores text on a 100-point scale where higher numbers indicate easier reading (100 is very easy, 0 is very difficult).
• Gunning Fog Index – Estimates the years of formal education needed to understand the text.
• SMOG (Simple Measure of Gobbledygook) – Calculates reading level by counting polysyllabic words.
• Coleman-Liau Index – Uses characters per word instead of syllables.

I tested one of my articles and got a range of results for the Flesch-Kincaid Grade Level shown below. Perhaps using several metrics would yield more helpful results.

• ChatGPT: Grade level 10-12
• Claude: Grade level 11-12
• Gemini: Grade level 13.6
• Word: Grade level 13.4

However, the advantage of using an AI writing assistant for a grade level analysis is the feedback it provides to help you meet the desired level. I received suggestions to simplify sentence structure, reduce jargon, and break up long paragraphs.

Conclusion

I think most people would concur that using AI for writing assistance can potentially expedite your work, improve your writing, and catch grammar errors. However, we must remain aware that the recommendations and suggestions can transform your unique writing style into something bland and average.

To overcome this, some writers prompt their AI writing assistant to create a style guide based on their writings. Then, they request that the AI tool write in their voice. This approach may produce decent draft versions for workplace writing tasks. However, if you write to share with others, to organize your thoughts, and as a vehicle for your creative energy, I think you will lose the joy of the process by letting AI write for you.