I used ChatGPT 5.6 to help me investigate the “Make money with AI world” — is it a scam?

One of the paradoxes I find at the moment is the disconnect between how easy things are to make with AI and how hard it is to make money.

Let me rephrase, it feels impossible to make money with AI.

Before I get into it, I want to lay out who this article is for

• Developers: I’m going to teach you how /goal mode works with a real use case. You probably should be using goal mode more than you are.
• Students: any time you need to do research, this is a good first step
• Office workers: maybe you are preparing a business case, maybe you are wanting to get information on competitor pricing. This article will show you a blueprint for that.

Not a Medium member? Keep reading for free by clicking **here**.

Back to how hard making money is.

But the thing that I find annoying is the culture of Twitter and YouTube and having a bunch of videos that promise that making money with AI is easy. Or that you can make $5,000+ a month easily, by making a vibe coded app.

Let’s be real for a second.

Making money is really really hard. AI has been good for over a year now and I’ve tried a few ideas but none of them have made money.

I’m sure most people are in my boat.

I don’t actually have a problem with people saying that they have made money, I find I don’t like the word easy.

I’m distrustful of anything that is easy and even though AI has made things easier, maybe it is actually comforting that making money is still hard?

But, with the release of ChatGPT 5.6, I thought what better way to put it through it’s paces and to show you the power of goal mode than to help me investigate the Make Money with AI world like an investigative journalist.

[embed]

So step 1, I opened the ChatGPT desktop app as Codex now lives inside there.

Up the top left, I changed the drop down from Work to Codex.

next, I started my prompt by writing /goal. Note for Claude Code users, you can do the same thing by typing /goal.

This will set ChatGPT on a longer task than a normal prompt.

Here is the prompt that I wrote, feel free to try this experiment yourself and see the results.

i want you to do a lot of research for me, there are a lot of videos, with this
one being a prime example:
[video that I pasted here]
as an example of "how to make money with ai"

i want you to find 50 examples of these, using youtube, blog posts or things like that
your job is to catalogue this, like you are an investigate journalist
you are then going to assess the source, the claim they are making, 
then assess against hamilton helmer and the 7 powers and just general businesses theories
the hypothesis i have, there is only just large scale companies making money with AI, foundational models, large VC backed firms like elevenlabs and then a very strange fake lower level, that is basically just a meta marketing scam or pyramid scheme without actual legitimate depth
conversely if you can find evidence to the opposite then that is great, let me know as id be super interested, you have a blank repo

In plain English what I asked to do was

• Analyse 50 videos from different creators across different business archetypes
• Test the claims of the video and the evidence like an investigative journalist
• Test the business models that they were discussing and whether they indicated any business fundamentals like Hamilton Helmers 7 laws of Power.
• Give the video a grade

ChatGPT ran for 22 minutes before coming back with this.

That’s right from that one prompt I got

• An investigative report
• A full 50-source catalogue of the videos with a detailed breakdown
• An excel spreadsheet ranking the videos
• A methodology document

That is simply amazing.

Now for the purposes of this article, I’m not going to reveal any specific video that was graded negatively as I don’t want to call out any single creator and I’d rather talk about the trends in general.

First of all it assessed each video and gave it a grade of how strong the evidence was.

A being the best, D being the worst.

In the 50 videos there was 1 A video which meant the evidence was independently anchored.

5 B videos which had strong primary documentation or they were transparent experiments.

That left 44 C and D. C being self-reported with no substantial evidence and D being weak or a hypothetical.

So far, not so good for the make money with AI YouTube crowd.

In terms of the different archetypes they covered a broad range.

This coverage aligns with my expectations. Because AI is good at generating, everything is in the generation camp, from music to YouTube videos, to books to voice-agent and digital products.

There was a couple of automation agency and AI freelancing vidoes in there as well.

Now I don’t mind sharing the videos that got a grade of B and A as I feel that is positive for those creators.

Interestingly these videos were all experiments and speaking from the first person.

• “I tried selling AI art prints on Etsy for 30 days”
• “I tried the easiest AI side hustle for 30 days”
• “Can I sell an AI-generated book on Amazon kdp”
• “I tried selling AI images on Adobe Stock for 1 year”

The cool thing about these videos as well? The AI prints on etsy and trying the AI side hustles were also the most viewed videos of the sample! People love hearing about actual experiments and results.

What side hustle made the most money?

Of the videos that had verifiable proof of their earnings, the amounts earned were tiny.

• $45.98 for selling AI prints
• $83 from making AI music
• $593 from selling AI images on Adobe stock.

That’s it. Basically not even covering the cost of the subscription.

The low amounts also seem to confirm a theory that I had in another post I did.

[embed]Why AI May Hollow Out the Middle of the Economy AI may favour giant context-rich firms and tiny fast-moving teams, leaving medium-sized companies trapped between both…chrisdunlop.substack.com

Basically I find it strange that you have billion dollar businesses (ChatGPT, Anthropic, Cursor, ElevenLabs) and very few examples of AI companies beneath that.

It makes this weird world where AI is hollowing out the middle of the economy.

Clearly it’s also a case that because generation is so easy, everything is trying it and when everyone tries that, it reduces the value of what everyone is doing.

But as much as this is an article warning you to be weary about what people claim in a get rich quick video, it is also a video showing you the power of /goal mode and ChatGPT 5.6 sol.

I was very impressed to be honest, the report was so fast, it was thorough, it was balanced and if you ever have a task that requires research like this, then using goal mode is a perfect way to get started.

One last note,

Before believing the next AI income video, here is what Sol recommended that we ask.

1. Who is the arm’s-length customer, and what painful job are they paying to solve?
2. Is the number revenue, GMV, bookings, ARR, or actual profit?
3. What period does it cover, and how many failed attempts preceded it?
4. Are ads, refunds, fees, model costs, and founder labour included?
5. What percentage of students achieved the result — not just the best example?
6. Does the source earn from operating the method, or from teaching/tools/referrals?
7. What stops a thousand competitors using the same models and prompts?
8. Which Seven Power belongs to the operator rather than the platform?

Before you go

Subscribe to my Substack newsletter because you get the following:

• A brand-new article for executives on Sunday that’s only posted on Substack.
• Links to every Medium post I’ve written in the past week
• Book recommendations every week for you to spend your Audible credits on

What do you think? Do you find it weird that we have billion dollar AI companies and basically not that many smaller AI businesses? Does anyone use anything other than large scale models? Am I just living under a rock?