How people brief AI to build a course: 37,284 prompts

Go from storyboard to an interactive course in minutes

Start Free

Half of the people who asked Coursebox to build a course this year described the whole thing in 21 words or fewer. That is about one sentence for a course that will run to several modules, a quiz and a certificate.

I looked at the anonymised first prompts of 37,284 people who created a course with AI on Coursebox between May and November 2026, and classified a random sample of 10,000 of them with AI, checking a sample of the results by hand. Nearly half of the briefs (47%) left out every design decision an instructional designer would ask about first: who the course is for, their level, how long it should be, what learners should be able to do afterwards and how learning will be checked.

How long is a typical AI course prompt?

The typical prompt to create a course with AI is 21 words (134 characters) long, according to Coursebox's analysis of 37,284 people's first prompts (May to November 2026). 44% of people write fewer than 100 characters, roughly "create a course on fire safety for warehouse staff". 20% write more than 500 characters, usually because they paste in their own outline or notes.

A one-sentence brief still produces a full course. The cost shows up later, in the number of rounds it takes to get the course people wanted.

What do AI course prompts leave out?

Nearly half of AI course prompts leave out every key design decision: 47% mention none of the audience, the level, the length, the learning outcomes or the assessment, and only 13% mention three or more. I checked each learning-related prompt in the sample for those five decisions.

The audience is the detail people include most often, in 31% of prompts. Learning outcomes (17%) and assessment (16%) appear in only about one prompt in six, even though those two decide what every module is for and how you will know the course worked.

More than a quarter of prompts (28%) point the AI at the person's own material: an uploaded file, pasted text or a link. Those courses come out bigger. Outlines built from a document have a median of 8 modules, against 6 for prompt-only outlines.

Who writes the most complete course briefs?

Corporate learning teams and training providers write the most complete course briefs: only about one in five of their prompts leaves out every key design decision, against nearly half overall. Corporate teams name the audience in 70% of prompts, and training providers (45%) and corporate teams (44%) give the AI their own source material most often.

Course creators state learning outcomes more often than any other group (44%), which suits a course sold on its promise. They are also the least likely to give the AI their own material, at 17%. For a paid course that is the expensive gap: your method, examples and stories are what a buyer is paying for, and a prompt alone cannot supply them.

What do people change after the first draft?

After the first draft, people mostly add to the course: new content (17% of edit requests), images (14%), a quiz (10%) and examples (8%). Very few ask for it to be shorter (1.5%) or simpler (2%). Before the draft, 76% have already sent more than one prompt to fill in what the first brief left out.

Much of what gets added afterwards, such as the quiz and the examples, can go into the first brief instead.

What should a prompt to create a course include?

A prompt to create a course should state six things: who it is for, what learners should be able to do afterwards, how long they have, how learning will be checked, what to build from and the tone. Those are the elements the most complete briefs in the data had in common. If you only add two, add the audience and the outcomes; they change the structure of the course as well as the wording.

This is the template I use. Copy it, replace the parts in square brackets and attach your document:

Create a course for [AUDIENCE].

By the end, learners should be able to:
- [OUTCOME_1]
- [OUTCOME_2]
- [OUTCOME_3]

Level: [LEVEL]
Length: [LENGTH], split into [MODULES] modules.
Assessment: [ASSESSMENT]
Build it from the attached [SOURCE] and keep to its procedures and terminology.
Tone: [TONE]

How I measured this

The data is anonymised and aggregated. It covers 137,126 course-creation prompts from 37,284 people on Coursebox between 14 May and 9 November 2026, with bots, staff and test accounts removed. No names, email addresses or company details were used in the analysis, no prompt text is published, and any group smaller than 30 is left out. Length, language and file use are measured on every person's first prompt. Categories come from an AI classification of a random sample of 10,000 first prompts written in people's own words, and 800 editor requests. A manual check of 40 classifications from each of the two AI models found about 92% correct, and 304 prompts labelled by both models were checked by hand to calibrate the design-decision figures. The examples above are illustrative, not real prompts. The full methodology is in the report.