Why Courses Are Dying, and Why AI Is Not the Reason

Bottom line up front. Course selling did not die because AI made information free. It died because claims outran substantiation, buyers stopped believing sellers, and the technology arrived afterward to finish a market that had already lost its credibility. If you diagnose this as an AI problem you will buy the wrong fix.

Everyone in the online education space is having the same conversation right now. Courses are dying. AI made the information layer worthless. The big platforms are pivoting to personalized learning while independent creators use AI to produce more static content faster.

Most of that is correct. The diagnosis is wrong.

AI did not kill course selling. Trust did, and it happened first.

Look at the timeline. The Federal Trade Commission sent its Notice of Penalty Offenses Concerning Money-Making Opportunities to more than 1,100 companies in October 2021, warning them about deceptive earnings claims and putting them on notice for civil penalties up to $43,792 per violation. That was four years before an AI agent could set up a Substack for you.

The resale course collapse ran through 2023 and 2024. Buyers who had been sold an income opportunity started talking. Refund demands went up. The words themselves took damage, to the point that 'digital marketing' and 'affiliate marketing' now carry a smell for people who never bought anything from either one.

Then AI arrived and made the how-to layer free.

So the sequence is trust, then technology. Not the other way around.

This is not a semantic argument. The diagnosis determines the treatment.

If you believe AI killed courses, the fix is technology. You buy a personalization tool, build adaptive paths, add an AI coach, and hope that closes the gap.

If you believe trust broke first, the fix is your claims. You substantiate what you say. You stop implying results you cannot document. You show people what typical looks like instead of showing them your best case and hoping nobody asks.

A creator who buys the technology and skips the claims problem ends up with a better looking offer that nobody believes. The personalization is real and the credibility is still gone.

The first is education sold on skill and outcome. Completion is the metric. This category has real problems, but it is not collapsing. Research published in Science in 2019 by Justin Reich and José Ruipérez-Valiente analyzed 565 MIT and Harvard courses across 12.67 million registrations and found that 3.13 percent of all participants completed in 2017 to 2018, and that 52 percent of registrants never started at all.

Worth knowing the other half of that finding, because it gets left out. Among students who paid for the verified track, 46 percent completed. Paying changes behavior.

The second is the income opportunity product sold on earnings potential. This is the category that broke, and no platform publishes audited completion data for it. That absence is not an accident.

The split shows up in the revenue. Business of Apps reports the online course app sector earned $2.16 billion in 2025, a 6.4 percent decline year over year, with user numbers down as well. Corporate and enterprise training kept growing over the same period. Consumer courses shrinking while corporate training expands is a cleaner signal than any total market projection.

In a master resell rights product, the buyer becomes the seller.

That single design choice means the creator loses control of every claim made downstream. You can write a compliant sales page, publish a fair earnings disclaimer, and instruct your community not to use hype language, and it will not matter. Thousands of resellers are now marketing your product with their own income screenshots, and not one of them is substantiated.

This is not a bad actor problem. Bad actors are a subset. The model manufactures unsubstantiated earnings claims at scale by default, and the person who built it cannot stop it from the inside.

That is why the largest resale products in this space have moved off resell rights entirely. The people running them ran out of ways to control it.

The Business Opportunity Rule has never covered most course and coaching sellers, which forced the FTC to pursue these cases one at a time under Section 5 while deceptive earnings claims spread faster than enforcement could follow.

In January 2025 the Commission proposed expanding that rule to cover business coaching and other money-making opportunities, and renaming it the Business and Money-Making Opportunity Rule. That proposal is not final and its outcome is uncertain. It is a signal of direction, not a settled requirement.

One point is worth being precise about, because it gets said wrong constantly. Unsubstantiated income claims were never legal. Section 5 of the FTC Act has always required earnings claims to be truthful and substantiated, and the FTC removed the old 'results not typical' disclaimer safe harbor when it revised the Endorsement Guides in 2009. A disclaimer has not cured a misleading claim for over fifteen years.

What changed is not the law. It is the cost of ignoring it.

If you sell education of any kind, the question is not whether AI will replace your content. It probably will replace the how-to layer, and that shift is already visible in what the major platforms are building.

The question is whether anyone will believe you when you tell them what your product does.

Credibility is the asset that survives a format change. It is also the one that takes longest to rebuild, which is why the operators who protected it through the last three years are in a better position now than the ones who monetized faster.

I have been saying a version of this for three years and mostly got called the compliance guy for it. That was never the compliment it sounded like. But the sequence held. Trust broke, then the regulators moved, and now the technology is finishing what the trust failure started.

Structure first. Claims you can stand behind. Then the tools.

• FTC Notice of Penalty Offenses Concerning Money-Making Opportunities, October 26, 2021. More than 1,100 companies notified, civil penalties up to $43,792 per violation. Federal Trade Commission, ftc.gov.
• FTC Notice of Proposed Rulemaking, Business Opportunity Rule, January 13, 2025. Proposal to expand coverage to business coaching and money-making opportunities. Status: proposed, not final. Federal Trade Commission, ftc.gov.
• FTC Endorsement Guides revision, 2009. Removed the safe harbor for atypical testimonials carrying a disclaimer. 16 CFR Part 255.
• Reich, J. and Ruipérez-Valiente, J.A. The MOOC Pivot. Science, Volume 363, Issue 6423, January 2019.
• Business of Apps, online course app revenue data, 2025.
• Market sizing note: total online education market estimates range from roughly $200 billion to over $350 billion with projected growth from 8 to nearly 20 percent depending on the research firm. No single figure is cited here because the spread is too wide to treat any of them as authoritative.