How to bypass AI content detectors in traffic arbitrage?

Why arbitrageurs use AI content: what are the benefits?

If you are engaged in full-time traffic arbitrage, you need to create new ad creatives on a regular basis. Creating them and other content manually requires a lot of time and skills. To develop even a simple ad, you need several hours and skills in commercial copywriting and graphic design. And, in addition to basic design theory, you will also need professional knowledge of at least one graphic editor — for example, Photoshop.

Of course, the use of specific skills and software depends on the objectives. In some cases, it is enough to write a catchy headline and take a stock image from the Internet. But still, to get a high conversion rate, you almost always need professional skills in content creation. It takes a long time and is difficult to develop them. And arbitrageurs already perform dozens of tasks every day.

However, it is possible to optimize time costs and automate the development of high-quality advertising content. For this purpose, arbitrageurs use neural networks.

Generating AI content is not difficult. Using artificial intelligence comes down to writing detailed prompts — queries (requests). Of course, to do this, you need to know how to use special commands and their parameters. But the developers of each AI provide users with detailed instructions on existing functions.

Using Midjourney neural network as an example, AI content creation is performed in 2 main stages. First you need to register and subscribe:

1. Go to the official Midjourney website;
2. Click the “Sign In” button in the bottom right corner of the homepage;
3. Enter the email address or phone number that is linked to your existing Discord account;
4. Enter your Discord account password. If you don’t have an account, you will need to create one. To do this, in the current window you will need to click on the “Register” link;
5. Click the “Login” button;
6. Click the “Authorize” button in the opened window;
7. Authorize the email or phone number for the Discord account if the neural network asks for it;
8. Download the Discord app to your computer and log into your account;
9. Return to the Midjourney website;
10. Buy a subscription on the automatically loaded page.

What you should do next:

1. Return to Discord on PC;
2. Click the “Add Server” button in the bottom left corner of the interface;
3. Click the “Join Server” button;
4. Enter the link to the Midjourney server;
5. Click “Join Server”;
6. Pass the captcha.

The second step is directly creating the AI content itself. First, you need to select any Discord room on the AI server. Then you need to enter the command “/imagine” and assign a request to it with the help of variables in English. The main variables are:

• Task. Does not have a designator variable. The task is simply written at the very beginning of the request — right after the word “prompt”. The parameter defines what exactly the artificial intelligence will generate. If you specify the task, for example, “a cat with troll ears”, the neural network will generate a picture of a cat with troll ears;
• Aspect ratio. It is denoted by the variable “–aspect” or “–ar”. The parameter specifies the aspect ratio of the future image, for example, 2:1. You can specify any ratio;
• Entropy. It is denoted by the variable “–chaos”. Entropy has a parameter “number”. It specifies an integer from 0 to 100. The value affects how diverse the results of the neural network will be. The higher the number in the parameter, the more unusual will be the results;
• Negation. It is denoted by the variable “–no”. The negation parameter can be almost any word, for example, “plants”. If you specify it, the neural network will generate images without plants. The negation in the prompt makes it so that the artificial intelligence does not use any objects when generating pictures;
• Rendering quality. Denoted by the variable “–quality” or “–q”. The variable specifies the duration of rendering of the future image. The value can be integer or fractional to hundredths, i.e. 1 or .25. The longer the time, the higher will be the quality of the final image;
• Seed. It is denoted by the variable “–seed”. Seed has an integer parameter. It specifies values from 0 to 4,294,967,295. With seed, Midjourney creates a visual noise field and uses it as a starting point to generate a network of the future image when no other visual input is available;
• Stop. Denoted by the variable “–stop”. Stop has an integer parameter. It specifies numbers from 10 to 100. The stop function in the query instructs the neural network to terminate before the image generation is finished. This can be used to generate more blurry and less detailed images.

Here’s a quick checklist for drafting a quality prompt from scratch:

• Visualize the future outcome and clearly set the goal;
• Examine existing teams and variables. A fundamental understanding of what they do is required;
• Select only those teams and variables that may be needed for the task;
• Compose a query. It should be short and clearly convey to the artificial intelligence the essence of the task. This is the only way to get the desired result.

Let’s look at some sample requests to Midjourney. The first prompt is “/imagine prompt a white cat with a black spot on its right ear lying on the floor –aspect 16:9 –no carpets” –stop 95>. At this prompt, the artificial intelligence should generate a 16:9 aspect ratio picture that shows a white cat with a black spot on its right ear lying on the floor and no carpets. Midjourney only has to complete 95% of the work. The rest of the variables will be used with default settings.

The second prompt is “/imagine prompt a robot sitting at a computer desk –no keyboards –seed 672154”. According to this request, the neural network should create an image of a robot sitting at a computer desk and no keyboard. In doing so, the AI will use seed 672154. The other variables will be with default parameters.

The third prompt is “/imagine prompt a man with a hat”. At this prompt, the neural network will create a picture of a man with a hat. All variables will be used with default parameters.

As of January 2024, there are hundreds of artificial intelligences that somehow prove useful in internet marketing. But most of them are not suitable for regular usage. Therefore, our editorial team has gathered the top-5 most useful neural networks for traffic arbitrage. So let’s quickly check them out!

Why ad networks don’t approve AI content

At the end of 2023, many neural networks are trained so well that they can create quality content for most queries. But there’s a catch. For example, when generating texts, artificial intelligences do not take into account all the interests of readers. Therefore, the materials are hardly useful for Internet users, which worsens behavioral factors. For search engines, this is a red flag. As a result, they lower the position of pages with AI content in the output.

The same logic is used by advertising networks. Unedited AI content in advertising attracts fewer users than high-quality materials created by specialists. At the same time, behavioral factors are also worse. As a result, the profit of ad networks is decreasing. But platforms are not ready to lose profit. That’s why some platforms are against AI content.

Do networks ban for using content from neural networks?

As of the beginning of January 2024, no advertising platform blocks arbitrageurs’ accounts for using AI-generated content. At the same time, all networks allow neural network-generated content to be shown. But then what is the problem?

There is talk on the Internet that search engines and advertising platforms will eventually begin to introduce algorithms to reduce the quality of AI content ranking. If search engines and networks do start integrating such algorithms, moneymakers will no longer be able to effectively utilize generated content in its original form.

Also, by December 2023, arbitrageurs have already experienced refusals to connect sites to the Google AdSense. Of course, with insignificant filling of the landing page with generated materials there are no problems yet. But if at least a third of the site is made up of AI content, the probability of rejection becomes high.

The only thing required is to humanize the materials. The task can be solved by using the most detailed prompts. Then you won’t need to edit the received materials manually. But it is important to remember: queries should still remain short.

However, the considered method does not always work. In most cases, you still have to edit materials manually. For example, if we are talking about AI texts, the following is needed to humanize them:

Reorganize sentences. By default, neural networks generate texts from sentences of approximately the same length. They need to be edited: add more information, shorten some sentences, and lengthen others. The text should look varied. Break the materials into logical blocks. Neural networks create texts with continuous “sheets”, unless otherwise specified in the prompts. Therefore, the received materials should be divided intoblocks: headings, subheadings, lists, tables, and so on.

After these 2 points are fulfilled, texts become more human and have a better chance to pass detector checks. But you should keep in mind that programs sometimes still detect slight AI “interference” in materials. This means that the texts are not humanized enough. Then you need to do both items on the list again — work on the materials more thoroughly.

As of January 2024, the use of AI content in advertising is legal. Networks don’t ban arbitrageurs for it. Also, the ranking of content from neural networks is still on par with the ranking of human content. The current problem is only that some money makers expect the ranking of AI content to deteriorate in search engines and ad networks in the future.

Special detectors are used to identify content that is generated by language models. They are very accurate in identifying AI materials. But if you humanize the generated content, it will pass the detector’s check. But it is better to humanize the material manually. Then the probability of passing the check will be higher.