Regular and Substantive Interaction (RSI) and AI in Online Courses
Does AI count as Regular and Substantive Interaction (RSI) under 34 CFR 600.2? Learn why AI bots fail distance education audits and how faculty stay central.
As colleges add AI tutors, automated discussion replies, and AI-drafted grading comments to online courses, instructional designers and distance education directors keep running into the same compliance question: do those interactions count toward Regular and Substantive Interaction (RSI)?
Under federal distance education rules (34 CFR § 600.2), interactions with AI chatbots, adaptive software, or automated feedback tools do not count as Regular and Substantive Interaction. Federal regulations require regular and substantive interaction between students and a qualified human instructor. Generative AI can support student practice and help faculty work through large grading queues, provided a human instructor initiates the schedule, evaluates student work, and delivers substantive feedback.
Getting that boundary right matters both for federal compliance and for teaching quality. When online programs rely on unreviewed automation to simulate instructor presence, students notice immediately, and institutions expose themselves to audit risk.
What 34 CFR § 600.2 Requires for Regular and Substantive Interaction
In U.S. higher education, Regular and Substantive Interaction is the legal line separating distance education from correspondence education. Under rules finalized by the U.S. Department of Education in September 2020 (effective July 1, 2021) and detailed in the Federal Student Aid Handbook, institutions face strict enrollment and aid limits on correspondence programs. Courses must qualify as distance education to preserve Title IV federal student aid eligibility.
As summarized in WCET's guiding principles on RSI, the regulation breaks the standard into two parts.
1. The "Substantive" Standard
Substantive interaction means engaging students in teaching, learning, and assessment consistent with course content. Under 34 CFR § 600.2, it must include at least two of five activities carried out by an instructor who meets institutional accrediting standards:
- Providing direct instruction.
- Assessing or providing feedback on a student's coursework.
- Providing information or responding to questions about the content of a course or competency.
- Facilitating a group discussion regarding the content of a course or competency.
- Conducting other instructional activities approved by the institution's or program's accrediting agency.
2. The "Regular" Standard
An institution ensures regular interaction prior to course completion in two ways:
- Providing opportunities for substantive interactions with the student on a predictable and scheduled basis matching the course length and content load.
- Monitoring academic engagement and success, with an instructor responsible for initiating substantive interaction when needed based on that monitoring or on a student's request.
Passive availability ("email me if you have questions") falls short, and grading only at the end of a term falls short too.
Why AI Chatbots and Automated Grading Fail the RSI Test
Many campuses are tempted to treat an interactive AI tutor or automated feedback script as a substitute for instructor touchpoints in large online classes. Federal rules separate academic engagement from RSI.
Under 34 CFR § 600.2, academic engagement covers active student participation in an instructional activity, including interactive tutorials or computer-assisted instruction. An AI study assistant can count toward academic engagement for attendance tracking, yet that interaction still fails the RSI standard because no human instructor participates in the exchange.
Three shortcuts put online programs at risk during federal or accreditation reviews:
- Unreviewed AI grading sent straight to the gradebook: When software scores open-ended work and posts machine comments without faculty review, the instructor has delegated evaluation instead of performing it. As examined in our guide on faculty-controlled AI assessment, human judgment must remain central to grading.
- Automated bots replying inside discussion forums: Posting automated AI replies in weekly discussion boards creates visible activity while failing the requirement for human facilitation.
- Passive AI Q&A mistaken for scheduled instruction: Even a course-grounded chatbot waits for a student to type a prompt, failing both the human-instructor rule and the predictable, instructor-initiated schedule.
Our comparison of guided AI conversations vs. open-ended chatbots examines why unconstrained bots drift instructionally. On compliance, the rule is just as direct: software can prepare the ground, while instructors must own the teaching relationship.
Four Online Course Patterns That Keep RSI Intact While Using AI
Online programs do not need to ban AI tools to stay compliant. The practical goal is structuring courses so AI absorbs repetitive preparation or practice overhead while the instructor delivers timely, human feedback.
Consider four illustrative course designs across common online disciplines:
1. Pre-Seminar AI Tutoring Paired With Weekly Instructor Synthesis
Suppose a department runs an online introductory microeconomics course with 90 students. During the week, students work through supply-and-demand problem sets alongside a syllabus-grounded AI tutor configured to give Socratic hints without printing final answers.
On Thursday evening, the instructor reviews a summary of the concepts where students stalled most often. On Friday morning, the instructor posts a four-minute video addressing those two misconceptions (direct instruction) and enters the scheduled Q&A thread to answer student follow-ups (responding to content questions). AI handled round-the-clock practice, while the instructor delivered scheduled teaching aimed at real cohort gaps.
2. Asynchronous Spoken Assessments With Human Rubric Calibration
Traditional online essays and post-and-reply discussion boards often produce formulaic AI prose that leaves instructors grading surface phrasing. Moving a portion of those touchpoints to short, asynchronous oral check-ins restores authentic evidence of learning.
In an online RN-to-BSN clinical pharmacology module, for example, students record a five-minute spoken explanation of contraindications and dosing adjustments for a patient case. An AI pipeline transcribes the audio and aligns verbatim student quotes with the faculty rubric. The nursing instructor listens to flagged moments, adjusts or confirms criterion scores, adds personalized clinical guidance, and releases the grade. Reviewing a five-minute defense takes less time than parsing a five-page essay, and feedback responds directly to what the student said.
3. Ungraded Simulation Rehearsals Before Instructor Debriefs
In a graduate public policy seminar, students practice a budget hearing inside an AI role-play simulation, testing arguments against an AI city council persona across multiple ungraded attempts. Afterward, each student records a brief spoken reflection identifying where their policy logic held up and where they had to revise their numbers.
The simulation itself serves as academic engagement. The faculty member satisfies substantive interaction by evaluating the reflection and responding with specific critique before the student's final memo is due.
4. Engagement Monitoring That Triggers Real Human Outreach
Regular interaction requires proactive monitoring. Learning analytics inside an LMS or assessment tool can flag online students who missed a midpoint reading check or struggled repeatedly on a practice scenario during week three.
Instead of letting an automated bot send generic encouragement emails, the instructor or teaching team uses those alerts to send a personal note or invite the student to a ten-minute office-hour check-in. Software surfaces who needs attention; an instructor initiates the contact.
For broader strategies on structuring online modules, see our guide to assessment design for online courses in the age of generative AI and our comparison of synchronous vs. asynchronous oral assessment.
How to Document RSI for Accreditation and Financial Aid Audits
During regional accreditation reviews or federal program audits, reviewers look for verifiable evidence in the syllabus, LMS shell, and gradebook to confirm that a course operates as described. As WCET's artificial intelligence practice guidance notes, institutions updating their AI policies should state plainly which instructional tasks may be AI-assisted and which require direct human faculty action.
Instructional designers and online program directors can make compliance visible in four ways:
- Publish the interaction cadence in the syllabus: State clearly when and how the instructor engages students each week (for example, Tuesday concept debriefs, Friday rubric feedback releases, and scheduled office hours).
- Preserve an audit trail of human review: Ensure your assessment tools record when an instructor reviewed student work, adjusted or confirmed scores, and released individualized feedback, keeping those touchpoints separate from auto-graded quizzes.
- Label automated practice tools accurately: Describe AI study assistants and practice simulations in the syllabus as supplemental academic engagement tools, distinct from instructor-led grading.
- Build accessibility into every required touchpoint: Every required oral or interactive check-in should offer equivalent text-based or extended-time options for students who need accommodations, following established practices for accessibility in oral assessment.
Where Rocketproof Fits in RSI-Compliant Online Courses
Rocketproof is built around a clear governance boundary: AI prepares evidence and drafts rubric suggestions, while faculty retain full authority over evaluation and grading. That architecture aligns directly with how online programs preserve Regular and Substantive Interaction.
Across authentic assessment workflows such as Paper Defense, Reading Assessment, Case Scenarios, and Role-Play Interviews, Rocketproof captures spoken or typed student responses, builds a verbatim transcript linked to audio timestamps, and drafts preliminary rubric notes citing exact student quotes. AI never issues or releases final grades on its own. An instructor reviews the evidence in one screen, adjusts rubric points, writes personalized feedback (or adapts snippets from a reusable Review Bank), and releases grades manually.
When connected through LTI 1.3 LMS integration, Canvas courses (verified end to end, with Blackboard Learn, Moodle, and D2L Brightspace supported on the same LTI 1.3 standard) receive scores and qualitative faculty feedback in the gradebook only after the instructor triggers release.
Rocketproof's Course Assistant also supports online courses without blurring the RSI line. Instructors configure the assistant per course, choose the approved syllabus and reading sources, select Direct, Balanced, or Socratic mode, and keep the tool as a private draft until they publish it. Between scheduled faculty touchpoints, students get cited help from course materials while instructors stay in charge of teaching and evaluation.
To inspect the review screen and student workflow firsthand, explore the interactive Rocketproof demo or review institutional options on our pricing page.
Frequently Asked Questions
Does an AI chatbot or AI tutor count as Regular and Substantive Interaction (RSI)?
No. Under U.S. Department of Education rules in 34 CFR § 600.2, Regular and Substantive Interaction must occur between students and a qualified human instructor. Interactions with AI chatbots or automated tutoring software can count as student academic engagement, yet they do not satisfy federal RSI requirements.
Can faculty use AI to help draft grading feedback and still meet RSI requirements?
Yes, provided a human instructor stays in control of the evaluation. When software transcribes student work or drafts preliminary rubric notes tied to evidence, and the instructor personally reviews the submission, calibrates the score, edits or writes feedback, and releases the grade, the feedback remains an instructor-led interaction. Unreviewed automated grading sent straight to students does not qualify.
How do online courses maintain RSI if they stop using traditional discussion boards?
Faculty can replace repetitive discussion boards with scheduled weekly concept debriefs answering student questions, asynchronous spoken check-ins where instructors deliver targeted audio or written feedback, and small-group facilitation led directly by the teaching team. Moving away from crowded text forums often improves instructor feedback because students explain concepts in their own words.
What happens if an online course fails to meet federal RSI standards?
If federal or accreditation reviewers determine that an online course lacks regular and substantive interaction initiated by human instructors, the program risks being reclassified as a correspondence course. Exceeding statutory limits on correspondence courses or enrollment can jeopardize an institution's eligibility to administer Title IV federal financial aid.
Next Steps for Online Program Leaders and Instructional Designers
Audit one high-enrollment online course before next term begins against the five substantive activities in 34 CFR § 600.2. Look for any module where automated quizzes, AI tutors, or auto-graded rubrics carry the whole interaction load without a scheduled, human-reviewed instructor touchpoint.
Replacing one stale discussion board with a short, faculty-reviewed spoken check-in or weekly synthesis debrief strengthens both student learning and compliance documentation. Walk through the Rocketproof demo or review LTI 1.3 LMS integration to see how faculty-controlled assessment fits into your online course architecture.
See Rocketproof in action
Explore the interactive demo as a student and an instructor, or request a pilot for your department.
