What learning AI actually costs in 2026: we checked the prices against the API

In short: on 17 July 2026 we pulled prices straight from Udemy's internal API and from Coursera's pages. The headline finding: there is no Udemy discount — their own server returns zero savings and a zero discount percentage. A course costs €19.99. Coursera Plus is €50/month or €343/year. And Vanderbilt's prompt engineering specialization additionally requires a paid ChatGPT subscription, which puts a student at roughly €70 a month to watch video and receive an automatic 100%. Below: every number, the raw API response as proof, and an honest look at our own price tag — which comes out worse than we'd like.

Why price is worth checking at all

Price is the only property of a course you can know with total precision before you buy. Everything else — quality, freshness, whether it will actually help — becomes knowable only after payment clears. So money is the sensible place to start when you're trying to understand this market: it's the one part where you don't have to take anyone's word.

Except that here, too, you shouldn't. A course marketplace's storefront is built to present a price as an event rather than as a price list: the number arrives dressed as a vanishing opportunity. The trouble with that kind of dressing is that it can't be checked afterwards — it lives on the page for one moment and leaves no trace.

So we went and asked the platform's server what it thinks, rather than its shop window.

There is no Udemy discount. Here is their raw API response

The price block on a Udemy course page is loaded by a separate request to an internal endpoint, /api/course-landing-components/.../price_text. It's the same response the page uses to render everything you see. Verbatim:

"price": {"amount": 24.99, "currency": "EUR"},
"list_price": {"amount": 24.99, "currency": "EUR"},
"saving_price": {"amount": 0.0},
"has_discount_saving": false,
"discount_percent": 0

Read it slowly. price and list_price are identical: the price and the "full price" are the same number. saving_price is zero — there are no savings. has_discount_saving is false — no discount saving exists. discount_percent is zero — the discount percentage is zero.

This isn't our interpretation or our opinion. These are four fields from a server response, each stating the same thing in a different way. €24.99 is simply what the course costs. Not "a discounted price". Not "an offer about to expire". The price.

One practical consequence follows, and it's a solid one: waiting for a sale is pointless. You cannot wait for a discount on something already selling at its own price. And if a page is counting down to a price rise, the server feeding that page knows nothing about any rise: its response contains no list price above the current one, no percentage, and no deadline. If you want the course, the moment of purchase doesn't change the amount you pay; if you don't want it, a discount on something you don't want still leaves you with something you don't want.

What the courses actually cost

We walked the twelve most popular non-technical courses about ChatGPT and generative AI. The picture turned out to be strikingly uniform:

• Eleven of the twelve top courses: €19.99. One after another, no variation.
• One: €24.99. That's The Complete AI Guide (Julian Melanson and others), 376,845 students, 42 hours of video, 545 lectures.

Notice that price correlates with nothing — not length, not rating, not audience size. Generative AI for Beginners (Aakriti E-Learning) has 409,492 students, 121,079 reviews, a 4.53 rating, 4.5 hours and 29 lectures. Prompt and Context Engineering 101 (Mike Wheeler) has 84,942 students, a 4.31 rating (the worst in our sample), 2.5 hours, 34 lectures. ChatGPT: Complete Course For Work (Steve Ballinger) has 281,823 students, 16.5 hours, 158 lectures. All three: €19.99.

So 4.5 hours costs the same as 16.5 hours, and a 4.31-rated course costs the same as a 4.53-rated one. Price on this market carries no information about the product. It carries information about which number the platform considers acceptable for an impulse purchase.

Subscriptions: where money turns into a stream

One-off purchases aren't the only option, and the platforms push hard toward subscription. What we measured:

• Udemy Personal Plan — €20.00/month, with a promotional rate of €10.00/month.
• Coursera Plus — €50/month or €343/year. Refund window: 14 days.
• Google's own programmes on Coursera — $49/month after a 7-day trial.

Do the arithmetic and it gets uncomfortable. Udemy Personal at full price is €20/month — that is the price of an entire course, every month. One course bought for €19.99 is yours forever; a month of subscription at €20.00 is thirty days. The subscription wins in exactly one scenario: you genuinely complete more than one course a month. It's worth asking yourself honestly whether you completed even one in the past year.

Coursera Plus at €343/year works out to €28.58 a month if you spread it. The annual plan is nearly half the monthly rate, which is the standard move: the platform isn't selling access, it's selling your decision not to think about this for twelve months. The refund window is 14 days. Look at the asymmetry there: a year of commitment, two weeks to change your mind. You decide in July whether you'll be studying in March, and you get a fortnight to consider it.

The second bill nobody warns you about

Now the thing we found buried in Coursera's structure.

Vanderbilt's Prompt Engineering Specialization requires a paid ChatGPT+ subscription in order to complete the assignments. That specialization has 138,967 enrolments, a 4.8 rating and 39 hours of programme. The platform subscription does not include the tool the entire course is built around. You pay Coursera for the right to watch someone teach a tool, and you separately pay for the tool.

Add it up. Coursera Plus: €50/month. ChatGPT Plus: roughly $20/month on top.

⚠️ Here we have to stop and say the inconvenient part out loud: we could not confirm ChatGPT Plus's exact price against a primary source. Every OpenAI domain returned 403 to our crawler. The ~$20/month figure comes from a secondary source, and we're flagging it as unconfirmed rather than presenting it as fact. It's the only number in this article with that status, and we'd rather say so than bury it.

With that caveat: about €70 a month is what the platform-plus-model bundle costs a student who wants to watch the videos and do the assignments.

What €70 a month buys

The obvious question is what a person gets for that. Here is a verbatim review from a student of that very Vanderbilt specialization:

"you send your assignments and immediatly you got your results: 100% correct. I am still speechless. I put all that effort in and have no idea whether I my answer was correct or not."
— Shinysheep, 21.09.2023, 1★ (Vanderbilt Prompt Engineering)

A person submitted an assignment. The system instantly replied: 100% correct. No check happened — and the important thing is that this isn't negligence, and it isn't cutting costs on graders. There is no model inside the Coursera lesson. The platform physically cannot run your prompt and look at what came out of it. It can grade a multiple-choice quiz, and there its powers end.

Which produces a genuinely strange arrangement: a student buys a ChatGPT subscription in order to complete an assignment that the platform is unable to assess — and receives an automatic 100% for it. The money for the tool is paid; the tool never enters the grade.

Neighbouring complaints of the same species:

"Videos are too short and superficial so you end up memorizing sentence by sentence to pass quizzes. Not a learning experience."
— Laurie J Phillips, 12.04.2024, 3★ (IBM)

"quizzes give unhelpful feedback for incorrect answers and just say 'watch the video again'"
— Cory Covino, 04.05.2024, 2★ (IBM)

"Watch the video again" is what a system says when it has nothing to say about your specific answer. Not because the course author is lazy, but because diagnosing a particular mistake requires something that understood the mistake. Neither an autograder nor a multiple-choice quiz can do that.

The third bill: the tools themselves

The course isn't the only line item. The tools being taught cost money too, and that belongs in the budget from the start. What we measured on Google's side (prices as shown for Spain):

• Google AI Plus — €4.99/month.
• Google AI Pro — €21.99/month.
• Google AI Ultra — from €99.99/month.

That's a twentyfold spread, from €4.99 to €99.99. And here it's worth holding two numbers from different sections of this article side by side. A course costs €19.99 once. Google AI Pro costs €21.99 every month. A single month of the tool subscription costs more than a permanent course about that tool.

Which leads somewhere unexpected: in a beginner's AI budget, courses are not the main expense. Subscriptions are. Over a year Google AI Pro becomes €263.88, while the course stays €19.99 forever. If money is tight, the thing to optimise is not the thing people usually try to optimise. What the free tiers actually give you, and where their walls are, is covered in our piece on free AI tools.

Regional pricing: the same course at €4.27 and at €19.99

We looked at Udemy's regional price grid (Price Tier Matrix V3, converted at ECB rates on 16 July 2026). It turns out "the price of the course" isn't a property of the course at all. It's a property of where you're standing:

• Brazil: price floor R$24.90 ≈ €4.27.
• Mexico: MXN 129 ≈ €6.46.
• Colombia: COP 34,900 ≈ €9.45.
• Argentina: ARS is supported by neither Udemy nor Coursera — people there pay rich-country prices.

The gap between the Brazilian floor and the European price is close to fivefold for byte-identical content. That's standard geographic price discrimination, and by itself it isn't a scandal: the platform is adapting to purchasing power. The scandal is the last line. Argentina got no local currency, which means an Argentinian pays €19.99 — exactly what someone in a country with incomparably higher income pays. Not because anyone decided that on the merits, but because a currency isn't wired up.

This is also why comparing course prices "in general" is meaningless. The number in the shop window depends on where you're looking from, and any argument about "expensive or cheap" that doesn't name a country is an argument about nothing.

Now, honestly, about us

We'd rather not write this section. We're writing it because without it, everything above is advertising dressed as research.

Our course costs €9. A Udemy course costs €19.99. The temptation to write "we're a fraction of the price" is enormous, and it would be false. €9 against €19.99 is half, not "a fraction". We are talking about eleven euros — a sum that changes nobody's life and cannot reasonably decide anything.

It gets worse. Our all-courses subscription is €14.99/month. Udemy Personal Plan is €20.00/month at full price and €10.00/month on promo. So against Udemy's promotional rate we are fifty percent more expensive, and against the full rate we're a fifth cheaper. The honest phrasing is: parity or worse. We have no price advantage whatsoever.

And the genuinely awkward part: regions. Udemy's price floor in Brazil is €4.27. Our €9 there is twice the market floor. In Colombia the floor is €9.45, so there we're competitive — but by accident: our price simply happens to land nearby. We never built a regional grid, and in Brazil you can see that with the naked eye.

Why publish this? Because the conclusion we drew from our own measurements is a conclusion against our own marketing: price cannot be our differentiator. If you are choosing a course on price, don't choose us — take Udemy Personal on promo at €10/month, or find a free one. We mean that seriously. Eleven euros is a bad reason to spend twenty hours of your life.

There is only one thing that can differentiate: what is actually taught, and whether there's any way to check that you learned it. That's the rest of this article.

What no amount of money buys

We collected the negative reviews across all of these courses, and the complaints clustered into a remarkably narrow set of themes. Not one of them is about price.

One: reading off the slide. This is the dominant complaint about the video format, and it sounds identical across different authors and different platforms:

"why read straight from the slide? I can do that. This was not a helpful course at all"
— Janie I., 02.07.2026, 1★ (Justin Barnett)

"50% of this course is reading script like a robot from the slides. …they are just reading text from the slides which you can also you from any good website"
— Shashank T., 13.04.2026, 1★ (The Complete AI Guide)

"Written by AI, delivered by AI. The slides are way too crowded to be useful and it really doesn't help to have the bot read them out word-for-word"
— Bradley M., 16.04.2026, 1★ (RPATech, 118k students)

"I can do that" is not a verdict on the instructor; it's a verdict on the format. If the video is just narrated text, then text beats video on every axis: your pace, search on the page, going back, and a transcript by definition.

Two: padding and repetition. A direct consequence of selling duration as value:

"I finished week 2. And all the lessons could be in 1 lesson. I jumped to week 3, 4, 5 and 6, and I see some concept reapeted. Very bad time spending for me."
— Shai Mizrachi, 22.07.2023, 1★ (Vanderbilt)

"The course is far too drawn out… padding the course with filler and making it longer than it needs to be."
— Victoria X., 13.02.2026, 2★ (The Complete AI Guide)

"The majority of the class he just rambles around unimportant subject… There's really no 'meat' in this course. A major waste of time and money!"
— Kevin K., 20.02.2026, 1★ (ChatGPT for Work)

Notice how this connects back to price. Because €19.99 buys both 4.5 hours and 42 hours, the author has every incentive to inflate the hours: "42 hours for €24.99" looks like a better deal than "4.5 hours for €19.99". The platform's pricing policy manufactures filler. It isn't authorial malice; it's the architecture of the shop window.

Three: the content goes stale. And this one is money burned outright:

"Most content is from 2024. This course is not bad for its time, but just too dated now."
— Martin F., 27.05.2026, 2★ (Generative AI for Beginners — on a course advertised as updated 04.2026)

"The content is mostly from 2023. …I invested my 41 hours and Im learning content which is from 2023. Very disappointed."
— Harsh A., 09.04.2026, 1.5★ (The Complete AI Guide)

Forty-one hours. That's a working week spent on three-year-old material — and that week cost €24.99, which is to say a small fraction of what the time was worth. It also shows the ceiling on ratings as a signal: that course sits at 4.52 with 376,845 students.

Four: promised prompting, didn't teach prompting. On courses with prompt engineering in the title, no less:

"There is nothing teached about creating a good prompt. It is just an overview of types of prompts."
— Geralt O., 01.07.2026, 2★ (Mike Wheeler)

"No specific guidance on prompt engineering… what to avoid while asking, how to organize your thoughts, how to give feedback to AI based on its answers etc."
— Bharat Ram A., 05.06.2026, 1.5★ (Mike Wheeler)

The share of negative ratings on Udemy is far from trivial: Generative AI for Beginners has 9.61% of ratings at ≤3.5★; The Complete AI Guide has 10.36%. On Coursera the negative share is lower — 1.5–3.5%. Meanwhile the prerequisites across every top course are identical: no code needed anywhere. Verbatim: "No prerequisites as ChatGPT is a tool anyone can access and use immediately" (Ballinger); "No coding skills needed" (Wheeler); "A desire to learn" (Aakriti); "No prior experience with AI or programming is needed" (The Complete AI Guide).

Work out your own number

Below is a prompt with all of our measurements already baked in. It doesn't ask you for figures and it doesn't need the internet: it computes and compares three scenarios. Hit Run and see where the lines cross.

You are a meticulous financial analyst. Compute the cost of one year of AI
learning across three scenarios and show the arithmetic step by step.

Data (measured 17 July 2026, all prices in EUR, tax included):
A. One-off purchases: 3 Udemy courses at €19.99 + 1 course at €24.99.
   Lifetime access. No discounts exist: the API returns discount_percent = 0.
B. Coursera Plus: €50/month rolling OR €343/year paid up front.
   Plus a mandatory model subscription to complete assignments: €20/month
   (estimate; exact price unconfirmed). Refund window: 14 days.
C. Udemy Personal Plan: €20.00/month at full price, €10.00/month on promo.

Do exactly this:
1. Annual cost of each scenario. Report A, B-monthly, B-annual, C-full,
   C-promo separately.
2. For B, compute what share of the money goes to the platform and what share
   pays for the model. Give the percentage.
3. Find the break-even of C against A: how many courses per year must a person
   actually finish for a €20/month subscription to beat buying at €19.99 each?
   Compute the same for the €10/month promo rate.
4. Compare B's annual cost to the price of a single course from A. What multiple?
5. Name exactly one assumption in these calculations that most strongly drives
   the result, and explain why it is that one.
6. Close with one line: under what behaviour is scenario C better than A, and
   under what behaviour is it worse. No padding.

Pay attention to point 5. The most sensitive assumption is nearly always the same one: how many courses you will actually finish. Every subscription is built on that number being smaller than you believe it is at the moment you pay.

How to read a pricing page

The second prompt is a drill. It contains the same raw API response we showed above. The exercise is learning to separate the data from its presentation.

Below is the raw response from an online course platform's internal API,
which the page uses to render its price block.

--- API RESPONSE ---
"price": {"amount": 24.99, "currency": "EUR"},
"list_price": {"amount": 24.99, "currency": "EUR"},
"saving_price": {"amount": 0.0},
"has_discount_saving": false,
"discount_percent": 0
--- END OF RESPONSE ---

Answer strictly by the numbers:
1. List each field and translate it into plain language in one phrase.
2. Which single number is a buyer obliged to take away from this response,
   and which conclusion may NOT be drawn from it?
3. List the claims a storefront could display on top of this response that
   would contradict it. For each, name the field that refutes it.
4. State in one sentence what a buyer learns from the API response that they
   cannot learn from the page.
5. Name one thing this API response does NOT prove — i.e. where the conclusion
   "discounts never happen" would be a stretch.
6. Give me three questions to ask any page with a countdown timer before
   I press Buy.

Point 5 is the important one, and it's aimed squarely at us. Our measurement was taken on 17 July 2026 across twelve courses. It proves that at that moment, on those courses, there was no discount. It does not prove discounts never happen anywhere. If the model's answer to point 5 fails to notice this, the model is flattering you — and that's worth seeing too.

Vet a course before you pay

The third prompt handles the practical case: you have a course page open and five minutes to decide. It doesn't evaluate the course — it can't, it hasn't seen it. It does something else: it builds you a checklist where every item is verifiable on the page in front of you.

You are a sceptical learning advisor. Help me decide whether to buy a €19.99
AI course on a large platform. You have not seen this course — do not try to
evaluate it and do not invent anything about it.

Instead, build me a 5-minute vetting protocol.

Context (real market measurements as of 17 July 2026 — use them):
— Price barely correlates with length: 4.5 hours and 16.5 hours both cost €19.99.
— Top course lengths range from 2.5 to 42 hours. Lecture counts: 29 to 545.
— Frequent buyer complaints, verbatim: "50% of this course is reading script
  like a robot from the slides"; "all the lessons could be in 1 lesson";
  "The content is mostly from 2023… I invested my 41 hours";
  "There is nothing teached about creating a good prompt".
— Share of ratings at 3.5 stars or below on top courses: 9.61% and 10.36%,
  against average ratings above 4.5.

Produce:
1. Seven checks I can run on the course page itself within 5 minutes.
   Each with an explicit stop criterion (what exactly, if I see it, closes the tab).
2. For each check, which complaint from the list above it maps to.
3. Explain why a 4.5 average with 10% of ratings at 3.5 or below is NOT the same
   thing as "this course works for 9 people out of 10".
4. One question whose answer cannot be obtained before purchase by any means.
5. Bottom line: three signals that mean close the tab immediately.

Point 4 is the most honest thing in that prompt. There is one thing you cannot find out before paying: whether you will learn. Everything else is a proxy of varying crookedness.

What we don't know

The section that articles about pricing never include. We're including it because without it the previous ten sections are worth nothing.

• We could not confirm ChatGPT Plus's exact price. Every OpenAI domain returned 403 to our crawler. The ~$20/month figure is secondary-source. Hence "about €70/month" rather than "€70.00/month".
• We took the Udemy Personal Plan price from the landing page, not the API. udemy.com returns 403 on a direct request. The €20.00 and €10.00 figures come off the pricing page, which is one notch weaker evidence than the discount JSON.
• We did not filter Coursera reviews by star rating. Their filtering is client-side: the server HTML always returns the first page. Which means those quotes are the default listing any visitor sees. The good news: we weren't digging for the worst. The bad news: we don't have a systematic sample of the negatives.
• Our measurement is a single-day snapshot. 17 July 2026, twelve Udemy courses and six Coursera ones. Not "the market" — a slice of it. Prices can change tomorrow, and if they do, tomorrow's measurement is right and we aren't.

The whole thing in one paragraph

There is no Udemy discount — the raw JSON says so, and waiting for a sale is pointless. A course costs €19.99 whether it contains 4.5 hours or 42, so the price tells you nothing about the product. The real money isn't in courses, it's in subscriptions: Coursera Plus at €50/month plus a mandatory ChatGPT+ on top comes to roughly €70/month for video and an automatic 100%, while Google AI Pro at €21.99 costs more per month than a permanent course about it. Our €9 against €19.99 is half, not "a fraction", and in Brazil our price sits at twice the market floor: price cannot be our differentiator. Choose a course not on price but on one question — is there anything inside it that can check whether you learned? Next in this topic: free AI tools and their limits, the prompt engineer profession, AI agents in plain words and ready-made prompts for work.