Illustrated cover reading "AI in Instructional Design: What Changed in 2026 and What Didn't" on an aqua and coral background with a zigzag pattern

AI in Instructional Design: What Changed in 2026 and What Didn’t

Table of Contents

Share:
Easily share intel
Summarise this page with your favorite AI assistant

AI in instructional design moved from side experiment to everyday drafting help in 2026. Most designers now use it for outlines, objectives, quiz items and narration, and authoring tools now build it into the editor. What did not change is who answers for accuracy, learning value and fit. People still do.

This post sorts the evidence into three groups: what changed, what changed in part, and what did not. Each row names its source and rates the evidence. Most numbers come from self-reported surveys, so read them as signals, not measurements.

A note on dates. Both big surveys were taken in 2025. The product news below runs from April to September 2026. Together they show where things stand as of November 2026.

Key takeaways

  • Use is now common. About 80% of designers in an ATD survey use AI, and 57% of respondents in a Synthesia survey use it actively. Both samples lean toward early adopters.
  • The biggest product change in 2026 is that AI moved inside authoring tools, with course drafts, translation, avatars and learner chatbots built in.
  • Time savings look large but are self-reported. No source we read measured learner results.
  • Accuracy, security and legal worries did not fade, and SME review and real learner input are still needed.
  • Track your own review time. It shows whether AI saves hours on your content.

What changed in AI in instructional design in 2026?

Three things changed. Adoption became normal, AI moved inside authoring tools, and vendors shifted from chat prompts to agents that build whole drafts. The evidence for the first is survey data. The evidence for the other two is vendor announcements, which show direction, not results.

How many instructional designers use AI now?

In a 2025 survey of 232 talent development professionals, ATD Research reported that about 80% of instructional designers use AI tools. Nearly two-thirds began within the last year. The report was sponsored by Clarity Consultants.

The Synthesia 2026 report drew 421 responses from L&D professionals. It found 57% actively using AI and another 30% running early pilots. Synthesia says its survey circulated mainly in AI-friendly networks, and Synthesia sells AI video. So “most designers” is likely, not proven.

ATD also found that 20% of instructional designers do not use AI. Most of them cite a lack of trust in AI-generated content.

What did authoring tools add this year?

Articulate is one clear example. Its Q1 2026 release post, dated April 23, lists AI images inside Rise course drafts and AI help that syncs Storyline animation with narration. A July 16 post adds an AI Tutor in public beta, a chatbot learners can use inside a Rise course.

On September 10, Articulate announced Nova, an AI agent. Articulate says Nova is available in eligible accounts with 20,000 launch credits per user. It also lists translation into 80-plus languages and AI avatar videos. These are the vendor’s own statements as of November 2026, which we have not tested.

The pattern matters most. The first-draft step now lives inside the tool you build in. That makes AI easy to use, and unchecked drafts easy to publish.

Which instructional design tasks do people hand to AI first?

Drafting tasks come first: outlines, storyboards, learning objectives, quiz items and narration. Analysis tasks come last. ATD found AI used most for outlining, storyboarding and objectives, and least for learner personas and learner data. Synthesia’s survey shows the same split.

The surveys measure different things, so do not compare them directly. ATD counts people who “always or often” use AI for a task. Synthesia counts any use.

Task What the surveys found
Narration or voice 54% always or often use AI (ATD, 2025); 63% use text-to-speech (Synthesia, 2026)
Quiz generation 60% use AI (Synthesia, 2026)
Video creation 52% use AI (Synthesia, 2026); under 20% always or often use it for video, avatars or simulations (ATD, 2025)
Learner personas and data analysis 21% and 19% always or often use AI (ATD, 2025)

Our reading, which is judgment and not a survey result, is that people use AI where they can check the output at a glance. A quiz draft is quick to scan. A needs analysis is not.

Does AI in instructional design make courses better, or only faster?

Self-reports say faster, and often better. In ATD’s survey, 37% said AI greatly cut design time and 70% said it improved course quality. Those are opinions. The one controlled study we found was small, and none of the sources we read tracked what learners did afterward.

That controlled study came from Carnegie Mellon. In a 2024 course project, 27 master’s students designed micro-lessons with and without ChatGPT. Raters who did not know which was which scored the work. AI-assisted lessons averaged 12.69 on a 15-point rubric, against 11.72 without AI. The students were novices from one course, and the researchers found no sign that the benefit carried over to work done without AI.

A second source shows the other side. An Iowa State team published a design case in December 2025 about building an online course with generative AI. It is a reflection on one project, not an experiment. Still, the details are useful:

  • AI-drafted objectives often used verbs you cannot measure, or joined several actions in one line.
  • AI-drafted quiz items mostly tested recall, and in some the correct answer was always the longest option.
  • Scripts sounded unnatural, and fixing them sometimes took as long as writing them.
  • Generated images had misspellings, so the team hired graphic designers instead.
  • AI helped most with early ideas, structure and draft feedback text.

A fair summary follows. AI can lift a first draft, especially for newer designers. It does not remove the editing work, and a gain on a short lesson does not prove a gain on a full course.

Pro tip: Time The Review, Not Only The Draft. When you trial AI on a lesson, log two numbers: minutes to a first draft, and minutes until your SME signs off. A fast draft that needs a long fix is not a saving.

What changed, what changed in part, and what did not?

Of the eight areas below, two changed clearly, three changed in part, and three did not change. The verdicts are our reading of the sources, not a vote from the sources themselves. “Weak” evidence means the claim is self-reported or stated by a vendor, with no measurement behind it.

Area Verdict Evidence How strong
Who uses AI Changed About 80% of designers use AI (ATD, 2025). 57% active and 30% piloting (Synthesia, 2026). Moderate. Large samples, but self-selected.
Where AI lives Changed Articulate 2026 posts: AI course drafts, AI Tutor beta, Nova agent. Weak as proof of results. Vendor statements.
Design time Changed in part 37% say AI greatly reduced course design time (ATD, 2025). Weak. Self-reported, no measured hours.
Design quality Changed in part 70% say quality improved (ATD). Rubric scores 12.69 against 11.72 (Carnegie Mellon, 2024). Moderate for novices on short lessons. Unknown beyond that.
Team rules Changed in part 74% say their organization encourages trying AI, yet the report says governance lags adoption (Synthesia, 2026). Moderate. One vendor survey.
Accuracy and trust Did not change 52% name accuracy as a barrier (Synthesia). Most non-users cite trust (ATD). Weak objectives and recall-only quiz items (Iowa State, 2025). Moderate. Three sources agree, each with limits.
Security and legal limits Did not change 58% name security and 41% name legal limits as barriers. 59% avoid personal or sensitive learner data in AI (Synthesia). Moderate. We found no source showing these worries fell.
SME and learner input Did not change AI-made scenarios did not match users’ work, so the team used real user input (Iowa State). Weak. One design case.

The “did not change” rows share one cause. AI writes fluent text whether or not it is true, so checking still takes a person who knows the subject.

Which tasks does human instructional design judgment still decide?

Judgment still decides what the course should achieve, what is true, and what learners need. AI can draft around those calls, but it cannot make them for you. This list is our own judgment, not a survey result, so adjust it to your risk level.

These are the calls we would keep with a named person:

  • Deciding whether training is the right fix at all, before anyone drafts an outline.
  • Writing objectives that a learner can show and an assessor can observe.
  • Checking every fact against the SME and the source document.
  • Choosing practice activities that match the real job.
  • Setting pass marks and deciding what happens when someone fails, and what to cut.
  • Approving which files may go into which AI tool.

A strict version of this idea is the five-gate review in our guide to AI course creation from aviation SOPs. Each gate has an owner, a record and a stop rule.

Pro tip: Keep A Short Do-Not-Draft List. Name three outputs your team never accepts from AI without a second human check, such as answer keys, regulatory wording, and anything built from learner data. A short list beats a long policy nobody opens.

How should a small team use AI in the design process now?

Start with one drafting task, set data rules first, and name a human reviewer for every output. Measure review time and errors for a month, then expand one task at a time. This is our recommendation from the evidence above, not a method we have tested.

  1. Pick one task you can check by eye. An outline or a quiz draft is a good start. Our post on drafting an AI course outline from trainer notes shows one version, with an SME review step.
  2. Sort your files by sensitivity. Decide in writing which kinds of files may go into which tool. Our guide to AI data privacy for eLearning content shows one way to set up five classes.
  3. Name the reviewers. The SME checks facts. The designer checks objectives and practice.
  4. Log the numbers. Record draft minutes, review minutes and errors found for each lesson over four weeks.
  5. Check the cost model. Credits and your own provider key bill differently, and review labor can cost more than model use. See AI course creation cost with your own API key.

What should you do next?

Pick one lesson you are about to build and run the log from step four on it. One lesson gives you real numbers for your own content, your own SME and your own review rules.

If you want a tool that drafts from your own source files, SimpliAuthor’s AI course authoring page says it reads prompts, documents and links to pull out objectives and draft lessons. It also says you can connect your own OpenAI, Gemini or other provider key. The page has a start-free button, so you can start free on SimpliAuthor and run your first draft through the same review steps.

SimpliAuthor is made by SimpliTrain, the same company that publishes this site, so weigh this recommendation accordingly. We have not tested SimpliAuthor’s AI features against any other tool. The description comes from the vendor’s published documentation as of November 2026.

Frequently asked questions

What is AI in instructional design?

It means using AI tools, mostly generative AI, to help with design work such as outlines, objectives, scripts, quiz items, narration, images and translation. A person still decides what to teach and checks the result. In ATD’s survey, 96% of designers who use AI use generative AI.

Will AI replace instructional designers?

The sources we read describe AI taking over drafting tasks, not the whole role. ATD respondents used it most for outlines, storyboards and objectives, and the Iowa State team needed human review at every stage. None of the sources measured job losses, so this is a judgment, not a forecast.

Which AI tools do instructional designers use most?

In Synthesia’s survey of 421 respondents, 74% used ChatGPT, 54% used Copilot and 39% used Gemini. Synthesia sells AI video and its sample leans toward early adopters, so treat the order as a rough guide. Authoring vendors now add their own AI too.

Can AI write good learning objectives?

It can draft them, but check each one. The Iowa State team found that AI objectives often used verbs you cannot measure, or packed several actions into one line, so they rewrote them with SMEs. Use AI for a first list, then test each one: can you observe it and assess it?

How much time does AI save instructional designers?

Self-reports say a lot. In ATD’s 2025 survey, 37% said AI greatly reduced course design time. We found no source that measured hours on real projects. Review time can eat the savings, so see where AI course costs really sit.

Is it safe to put learner data into AI tools?

Many teams avoid it. In Synthesia’s report, 59% avoid personal or sensitive learner data in AI, and 58% name security as a barrier. Sort your files by sensitivity and approve each tool in writing before anyone uploads. Our AI data privacy guide shows how.

Want to learn more?

Reach out to us to learn more.

One Platform for
All Your Training Needs

Get a personalized demo.