Yes, and the market split in two while the question was being asked. Courses whose core value is information transfer, explanations, frameworks, how-things-work, are compressing, because their product is now available free, instantly, personalized, from tools a third of adults already use. Courses that sell structure, application, accountability, and a finished outcome are holding, and in some niches strengthening.
The split is unforgiving about which side a course sits on. Buyers did not stop wanting to learn; they stopped needing to pay for access to knowing. What survives the free alternative is everything a chat window does not provide: a sequence someone committed to, deadlines with teeth, feedback on their actual work, peers in the same fight, and a curator whose judgment about what matters is the real syllabus.
- The course market split, not died: information-transfer courses compress against free AI, while application-and-outcome courses hold.
- The pressure is measurable at the buyer level: a third of U.S. adults use ChatGPT, and your prospective students check any syllabus against it.
- What survives is what chat can't do: sequence, deadlines, feedback on real work, peers, and a curator's judgment about what matters.
- The graduate test sorts course ideas: if completing your course leaves someone with knowledge rather than a finished thing, it competes with free.
- AI also rebuilt the economics: production costs collapsed and content became answerable, which favors small, sharp, outcome-priced programs.
Join the Collective Wisdom™ Newsletter
Written by us from real client builds and our own operation, not recycled AI news. What worked, what didn’t, and what to do about it. Read in a few minutes, useful the same day.
Join the newsletterReady to talk? Book a Rapid Transformation Call.
What actually happened to course sales when AI arrived?
A sorting, visible in every course-adjacent business that shares numbers honestly. The informational tier felt it first: the broad how-to course, the framework library, the recorded-lecture catalog, offers whose pitch was 'learn what I know', started meeting a buyer who had already asked the free tools and arrived either satisfied or skeptical. With 34% of U.S. adults using ChatGPT, roughly double the share of two years prior, the comparison shop is now the default first step, not an edge case.
The application tier tells a different story. Programs with real cohorts, deadlines, reviewed work, and a defined finish line kept selling, because their buyers were never purchasing information: they were purchasing the probability of actually doing the thing. That probability was always the scarce good; AI just unbundled it from the content that used to carry it.
The honest generalization: nobody's course died because AI knows things. Courses die when knowing things was the whole offer, and the era finally put a price on the difference, which is zero for one and rising for the other.
What do buyers still pay for in a course?
Five things, none of which a chat window provides, and every durable course is built on some blend of them:
- Sequence. Someone with judgment decided what comes first, what to skip, and what order survives contact with reality. Infinite free information makes curation more valuable, not less; the syllabus is the product.
- Deadlines with witnesses. Cohorts, due dates, and visible progress convert intention into behavior. Buyers know their own follow-through statistics, and they are paying to beat them.
- Feedback on their actual work. Not the general case: their draft, their numbers, their attempt, reviewed against a standard. This is the layer closest to advisory value, and it prices accordingly.
- Peers in the same fight: the cohort effect, part accountability, part morale, part network, entirely unavailable at 11pm alone with a chatbot.
- The curator's name: whose judgment shaped this path, and what their endorsement of your completion signals.
The pattern underneath: buyers pay for changed odds, not transferred files. Courses that price and design around the odds keep working. Courses that price around the files are competing with free.
Which types of courses are most exposed, and which are fine?
Run any course, existing or planned, through the exposure gradient:
Most exposed:
- The explainer course: what X is, why X matters, how X generally works. This is precisely what free tools deliver, personalized, on demand.
- The recorded lecture library: passive video whose engagement model was always hope, now competing with an infinitely patient tutor.
- The generic-skills survey: broad, shallow coverage of a field's basics, which is the statistical center AI serves best.
Resilient:
- The outcome program: participants finish with a built thing, a launched offer, a working system, and the course is the scaffolding that got it built.
- The reviewed practicum: skill built through attempts and expert feedback, where the feedback loop is the product.
- The judgment-dense masterclass: an expert's specific calls, cases, and contrarian positions, material the consensus engine does not hold.
- The credentialed path, where completion signals something a market trusts.
The gradient compresses to one variable: how much of the course's value survives if every participant also has a free, tireless expert-explainer? That share is what buyers will keep paying for.
Should I still build the course I have been planning?
Run it through three filters before committing the season it will cost:
- The graduate test. Describe what a finisher has that a diligent free-tool user lacks. If the answer is 'understanding,' stop: that inventory is competing with free. If the answer is a finished artifact, a reviewed skill, or a changed number, keep going.
- The burnout test, honestly: is the course solving your buyers' problem or your calendar's? Courses built primarily as escape hatches from 1:1 work inherit the format's real demands, production, launches, audience-feeding, and often trade a delivery burden for a marketing one. Other formats exit 1:1 more directly.
- The volume test: courses are volume products, and volume needs audience or acquisition. Price the marketing honestly alongside the production.
If the idea survives all three, build it lean: presell a pilot cohort to existing contacts before recording anything, deliver it live and slightly ugly, and let the second cohort fund the polish. The graveyard of course projects is filled with beautifully produced curricula that skipped the presale, and the era's compression makes that skip more expensive than ever.
How does AI change how courses get built and delivered?
It collapsed the production tax and raised the delivery bar, which together reshape what a smart course looks like.
On the build side: the documentation quarter became a documentation fortnight. Your captured method drafts the curriculum, the worksheets, the session guides, and the assessments, in your voice, from your material. The expensive part of course creation was never the expertise; it was the rendering, and rendering is now cheap. This is also why 'I don't want to make a course' objections about production slog are aging out; the slog went.
On the delivery side, expectations moved: participants live with tools that answer instantly, so a course that is only scheduled content feels dated on arrival. The current bar:
- Answerable content: participants ask questions of the course material and get answers drawn from it, in the creator's voice, between sessions.
- Personalized artifacts: worksheets and plans generated from each participant's context rather than one-size templates.
- Tracked momentum: commitments and progress carried by the program's own systems.
The net effect favors exactly what the market split favors: small, sharp, outcome-priced programs, cheap to build and rich to deliver. Watching how the format keeps evolving is part of what the Collective Wisdom newsletter is for.
The PLB Perspective
I built and sold courses through the entire golden era, over seventy programs across my career, so hear this as a report from inside: the course was always two products wearing one price tag, information and probability, and buyers could never separate them until now. AI ripped the tag in half. The information half floats toward free, and the probability half, the odds that a real human finishes a real thing, prices higher than ever, because the noise made finishing rarer.
The mistake I watch established experts make is reading the split as a verdict on the format instead of a verdict on the offer. 'Courses are dead' and 'courses still work' are both lazy; the operative question is what your course's active ingredient was. The experts whose programs sold sequence, feedback, and finished outcomes barely felt the shift, and several are raising prices, because their competition, the info-course flood, is compressing out from under them and taking the market's noise with it.
And for the owner relieved to hear she does not have to build a course after all: notice that the same split runs through every format you might choose instead. Workshops, memberships, group programs, all of them divide into the information kind and the probability kind, and the era grades them identically. The format was never the decision. The active ingredient is, and yours, judgment applied until something is finished, works in every container you might pour it into.
Cindy Anne Molchany · Founder
The honest answer is that completion was always the format's weak point, and AI pushes in both directions: participants can now get unstuck instantly, which helps momentum, while the informational portions of a course feel more skippable than ever, which hurts engagement with passive content. Programs report the gains landing where AI supports application, answerable content, personalized artifacts, and the losses landing on lecture libraries.
Reprice the structure, not the number: strip or give away the informational layer, concentrate the paid offer on sequence, feedback, and outcomes, and price against the result rather than the content volume. Application-heavy courses are holding and often raising prices, because their comparison set is the cost of not finishing, not a chatbot subscription. A price cut on an information course just accelerates the compression.
The same split governs both: an information membership, content library plus loose community, compresses just like an information course, while a membership built on ongoing accountability, live judgment, and working sessions holds like an outcome course. Memberships add a retention burden courses avoid, and reward it with recurring revenue. Choose by the active ingredient you can sustain monthly, not by the billing model.
In the information tier, they already do, which is precisely the tier to exit: generic curricula are now effectively free to produce, and the flood makes that shelf unsellable for everyone. What AI cannot generate is the resilient tier's ingredients: an expert's earned judgment about sequence, real feedback on real work, and a name whose endorsement means something. The flood is an argument for moving up, not for competing down.
Good enough at what it was built for, probably. But AI moved the goalposts: the information layer of every program is now free at 11pm, and what clients pay for is what your program delivers beyond it.
By moving up the stack your clients just climbed: let AI have the informational layer, welcome their tool use into the program, and concentrate your delivery on application, accountability, and the calls only you can make.
If you have to ask, parts of it probably do, but dated is two different problems: surfaces that look old, and architecture that behaves old. Clients forgive the first far longer than the second.