Making Money with AI: Automated Affiliate Articles and Videos

The AI that writes articles is already worrying Fiverr writers

The promise sounds like a hair-growth sales pitch: generating content with intelligence and charging for it almost without effort. The person who started the thread told the story plainly. After producing articles for a small niche and placing affiliate links, the texts «work quite well», and some people are ranking those same articles in search results. The kicker: Fiverr writers «are already shaking».

The interesting part is not the tool, but the estimulante ilegal: text generators, image creators like DALL-E, and a growing list of programs that receive improvements every few weeks coexist. The difference with other get-rich-quick fads is that the bottleneck has moved: it is no longer about producing, but producing well. The trick has become cheaper. The criteria, no.

From affiliate articles to automated YouTube channels

The first front is writing. The technique consists of creating texts for micro-niches and placing affiliate links: the reader looks for a product recommendation, reads a reasonable article, and clicks. No prior sector experience is needed. What is needed is volume, because most of those articles do not rank; a few do and sustain the entire operation.

The second front, more ambitious, is video. Channels are being documented that explain step by step how to set up a complete YouTube channel with AI tools, some free and others paid. The chain includes script, generated images, and synthetic voice. Those who have tried it summarize it thus: «with a couple of tools and very few hours, you can make videos on these topics for YouTube like hotcakes». Another tried it with a video about facade insulation systems and was surprised by the result. The barrier to entry, in the end, is measured in afternoons, not months.

The PAA+AI method: hundreds of articles and the token bill

The industrial version of this has a name and origin. The approach is known as PAA+AI and has been documented for over a year in web positioning circles. The idea is to build sites with hundreds of automatically generated articles, targeting the related questions Google shows in each search. The content stops being written one by one: it is produced in batches.

The technical part is resolved by connecting an AI tool via API. The most cited is OpenAI, with several models depending on price —the names circulating range from the cheapest, Ada and Babbage, to a Davinci described as the most expensive and best rated—. Payment is made per generated tokens, and as a reference, one thousand tokens are equivalent to about 750 words. There are third-party alternatives, such as AI21 Labs, and those who want to fine-tune the style can use their own Python scripts to make fine adjustments.

Here appears the first sustancia ilegal in the story. Automating costs little; hitting the mark, no. The optimistic calculation gives almost free articles. The realistic scenario includes hours of testing, format adjustments, and a considerable failure rate that no one puts in the headline.

Is AI-generated content indexed? What the first tests say

Yes, at least for now. This is the conclusion of those who reviewed their own articles after hearing that content generated with ChatGPT does not rank well. They reviewed the keywords of each text. Traffic had increased, especially on two of their websites. Several of the searches they considered difficult were on the first page of Google. All articles were indexed.

It is advisable to take it with caution. These are individual observations, not a study, and Google has not clarified its position with figures. But it dispels, for the moment, the idea that it is enough to smell the text of a machine to sink it. What sinks, historically, is bad content.

The legal problem that almost no one looks at: who owns what the machine generates?

A recurring doubt among those who want to monetize this is whether images generated by programs like Canva or Midjourney can be used freely. The typical case: someone writes a children's book and wants the AI to make the 40 illustrations without paying an illustrator. The answer given in the thread is that there should be no problem as long as one does not resort to characters with registered trademarks —nothing of Disney or Star Wars—, because copyright protects equally even if the image was painted by a machine.

But the issue has a second, more uncomfortable side, pointed out by another forum user: if the text was birthed by ChatGPT and the illustration painted by Midjourney, the author is not exactly the owner of anything, and someone with fewer scruples can copy the book, change its name, and sell it. The absence of clear authorship is, at the same time, the advantage and the Achilles' heel of the business.

The fine print: thirty hours to learn what the tutorial promises in ten minutes

The image of AI as a button that spits out money breaks against the learning curve. A program like Stable Diffusion does not work by typing a phrase and waiting. One must understand what each button does, how an image is composed, what prompts and embeddings are, and above all training, «by far» the most complicated part of the process. The calculation that is handled is that it takes between 30 and 40 hours to do something halfway decent.

From this arises a business avenue that almost no one mentions: offering courses. If the software is complex, someone will have to teach it. There are already Photoshop courses; Stable Diffusion courses are much scarcer. And it is not just training: hardware also filters. Some have put aside learning because they doubt that their graphics card can support the program.

---

Money, in the end, is not where they promised. It does not flow from a magic button, but from knowing what to ask the machine, how much each word costs, and what can be signed without someone coming to claim authorship. With this panorama, the question is no longer whether AI allows making money, but how long it takes for the rest to realize that the shortcut does not exist.