Insights ยท September 2026

Are assignments still built for a world without AI?

We read a cross-section of recent public syllabi, and compared courses before and after ChatGPT, to see whether graded work has changed since generative AI arrived. On paper, it mostly has not.

What we found

None

of the recent syllabi we sampled described a graded component redesigned for AI

Source: RichFeedback study of public syllabi, September 2026
Half

of the grade, in the median course we sampled, rested on take-home work with no design for AI

Source: RichFeedback study of public syllabi, September 2026
None

of the courses we compared before and after ChatGPT tied a change to AI

Source: RichFeedback study of public syllabi, September 2026

Policies changed; assignments did not

Most recent syllabi now say something about AI, usually whether students may use it. Very few change the graded work itself. A much larger study found the same.

Courses with an AI policy in the syllabus, Fall 202555%
Courses with a new AI-based task11%
Source: Chirikov, How Instructors Regulate AI in College (2026) (over 31,000 syllabi at one large public research university)

What faculty say they are doing

47% of faculty are modifying assessment design because of AI

0%50%100% of faculty
Redesigning around AI: 24%Reverting to in-class or proctored work: 23%
Source: Tyton Partners, Time for Class 2026

How we did it

We coded recent syllabi, most dated Fall 2025 to Fall 2026, from research universities, regional publics, private colleges and community colleges across a wide range of disciplines. We found them through searches that used no AI-related terms.

We also compared the same courses at the same institutions before ChatGPT (Fall 2022 or earlier) and after (Fall 2024 to Spring 2026).

A graded component counted as redesigned only if the syllabus said it had been changed because of AI. Take-home work means work completed outside class and graded on what was submitted.

Limits

A syllabus is not the assignment sheet, and some redesign lives in course sites the syllabus never shows.

The sample is modest and drawn from syllabi that institutions publish, so some states and institutions are over-represented.

Treat these findings as a signal, not a census.

See where your own assignments stand