Field Notes

Courses about making money with AI outsell the actual work 10 to 1 — we counted

Published

While researching verified AI income cases, something kept showing up: the top of the search results was filled not with people doing the work, but with people selling the method for doing it.

Rather than leave that as an impression, we counted.

How we counted

Written out so you can repeat it.

  • Date 16 August 2026
  • Where Kmong search, logged out
  • Two search terms
    • A: AI 수익화 ("AI monetisation") — what someone types looking for a method
    • B: AI 챗봇 개발 ("AI chatbot development") — what someone types looking to hire
  • Sample top 20 listings each, 40 total
  • Recorded title, listed price, review count
  • Not recorded seller names. We do not single out individuals

What came out

A. "AI monetisation"B. "AI chatbot development"
Sample2020
Median reviews545
Total reviews2,063623
Highest456143
Listings with zero reviews01
Median listed price₩105,500₩125,000

The median listing has 10.9× the reviews.

Read the median, not the total. The 3.3× ratio in the totals is dragged upward by a few outliers; the median describes a typical listing.

Median reviews

AI monetisation (information products)54 reviews
AI chatbot development (services)5 reviews

Median review count of the top 20 listings for each term. Review counts are not revenue.

The part that contradicted our guess

Before counting, we assumed information products sell more because they are cheaper. That was wrong.

The median service is the more expensive one — ₩125,000 against ₩105,500. Price is not what separates them. Two markets at roughly the same price sell at completely different speeds.

What the titles said

Nine of the twenty listings in sample A put a specific earnings figure in the title — amounts like "₩54M in four months" or "₩34M in five months". One more named no figure but implied money arriving without work.

In sample B, the count was zero.

Number of titles containing an earnings figure (out of 20 each)

AI monetisation9 listings
AI chatbot development0 listings

We classified these by reading the titles. A judgement was made.

By our own grading rules those earnings claims are grade D — unsubstantiated or promotional, so we do not list them as cases. Yet those are the listings carrying the most evidence of having sold.

What we are not saying

  • Not that those products are scams. We verified two things only: the titles carry unsubstantiated earnings claims, and the listings have many reviews. We did not buy any of them
  • Not naming anyone. Titles are quoted to show a pattern; no links, no seller names
  • Not that information products are bad and services are good. Good teaching is worth paying for

⚠️ Limits of these numbers

Read it this way

  • Review counts are evidence of sales, displayed by the platform rather than self-reported
  • The sample is 40 listings across two search terms
  • Search results shift with time, personalisation and sort order

Reading it this way is wrong

  • Reviews equal revenue — we know neither the price paid nor the period
  • 10.9× reviews means 10.9× income — we do not multiply
  • This represents the whole marketplace

Note especially that different search terms summon different demand. "AI monetisation" naturally surfaces method-selling; "AI chatbot development" naturally surfaces services. That is less a flaw in the measurement than part of the observation — it describes what someone actually sees when they search for money and AI.

The classification was ours, made by reading titles. Among the top 20 for "AI monetisation", not one listing was a service that does work on a client's behalf.

Why this matters to us

This table is the reason this site exists.

Money moves faster toward selling the method than toward doing the work. So search puts the method first, and someone starting out buys that first.

We chose the other side — publish only what we verified, print no figure we could not check, and write down that we ourselves have earned nothing yet. These numbers show how unglamorous that side is. It is a market with a median of five.

Next time we will widen it: more search terms, and the same terms counted again a month later to see whether the ratio holds. One count is a photograph, not a trend.

Back to Field Notes