just a punk

I decided to write this post inspired by a question from someone who made the following points:

This is the information that should be considered and conclusions drawn. However, most webmasters reacted to this news as if it were the end of the world, with sentiments such as "all is lost", "the sky is falling", "life has no meaning", "we're gonna die".

In reality, of course, this is not the case. Google has once again repeated that it will ban spam. But hasn't it always done that? Now, this applies not only to "articles" written by third-world copywriters, but also to content generated by AI. Are you surprised?

Google does not claim to be against AI-generated content. It clearly states that machine-generated spam will be de-indexed. However, this definitely does not apply to quality content that provides useful information. Remember that Google itself is the developer of the Gemini AI model, which it says was created to improve the quality of articles on the web. The same is being done by Bing's owner, Microsoft, which is investing billions in the development of OpenAI's GPT.

Isn't it curious that the companies that own the largest search engines profit from selling AI models created for content generation? Quite a coincidence, isn't it? And will they really penalize their own content? Does anyone really think that the business operates on such a strange model?

In fact, it's worth taking a moment to understand what's really happening and how you can use it to your advantage. The reality is that probably 99% of people who decide to use AI for content generation have no idea what they're getting into. They think of AI as a magic wand that will write content for them personally, not taking into account the millions of other "geniuses" who have been struck by the same idea and expect to instantly hit the top of Google's SERP.

But they won't If you ask GPT to write an article about the health benefits of cold showers, what you'll get is essentially doorway page spam. It will be a short, generic text with no formatting and, most importantly, no meaning. It's basically a copy-paste of standard advice and facts. Google will certainly penalize this.

To generate quality content, i.e., full-fledged articles that make sense, you need to write appropriate prompts. OpenAI's GPT-4, for example, has "learned" the entire internet, and its "brain" definitely contains enough useful information to write an article on any topic. The thing is how to extract it, which requires leaving the right, detailed prompts. Modern text models like GPT-4 Turbo, Anthropic Claude 3, and soon Google Gemini 1.5, have a huge context window for incoming data - your prompts. Thus, you can provide them with a whole book of very detailed instructions.

Your prompt should specify the overall structure of your articles, the information to be included in the final text, the style of presentation, or even an authorial voice - maybe you want the article to be written by Stephen King, or maybe you prefer the conspiratorial trash talk of Tucker Carlson or Joe Rogan? Modern language models can do this too - they are familiar with these personal styles.

Be sure to include in your prompt the HTML structure of your article. Tell the language model which HTML tags it should use. Give it examples of different <div> blocks in your article, used by your WordPress theme to style various Q&A, FAQ, Pros and Cons sections, etc. Provide examples of HTML templates for styling tables, inserting images, and video embeds. Create a comprehensive specification for formatting HTML documents and submit it to the language model as part of your prompt.

Creating large articles with coherent content and exceeding the 4K token limit is something that can be done with the CyberSEO Pro plugin, which can generate articles up to 128K tokens, the size of an average book.

Regarding the indexing of AI-generated articles by search engines like Google, check the articles on cyberseo.net. The fact that articles generated with GPT-4 are indexed and even rank well for common queries such as "where to get images for an autoblog?" proves my statement. This is a compelling example that counters the common belief that Google categorically penalizes machine-generated content. Instead, it suggests that Google's algorithms are sophisticated enough to distinguish between spammy, low-effort AI content and high-quality, AI-generated content that provides real value to readers.

To members of this forum I would recommend using the Mistral Large, since its capabilities are very similar to those of GPT-4, with the added advantage of handling (analyzing and generating) texts up to 32K tokens in length. Most importantly, Mistral Large is uncensored. Yes, you've got it right.

Regarding concerns about content being flagged as AI-generated, the suggestion of using a synonymizer is intriguing. By creating a custom synonym table - because a list of typical words and phrases for your niche is rather limited - you can effectively mask the AI origin of your content. This method disrupts the typical patterns that GPT detectors look for and breaks the GPT flow, potentially avoiding detection as machine-generated content. While this approach doesn't inherently improve the quality of the text, it does provide a workaround for those who want to minimize the likelihood of their content being identified as AI-generated.

__________________
Obey the Cowgod