Creating unique articles from imported RSS feeds using AI - CyberSEO

Did you know that the [gpt_article] shortcode is perfect for creating unique articles with CyberSEO Pro and RSS Retriever plugins based on imported content? The AI model can be inspired by a variety of sources, such as news articles or scientific publications. This isn’t just a simple rewrite; you’ll get new, comprehensive articles that provide additional information on the topic you specify. So let’s explore some strategies that might be useful.

First, let’s consider a simple example where we want to use AI to create a new article based on information from an external source, such as an RSS feed. Import the desired content source using CyberSEO Pro and RSS Retriever, then navigate to the “Post templates” settings tab.

Let’s say you import a news channel and want to generate articles based on current world news in light of global warming, waste management, or any other topic your site is dedicated to. Just add the appropriate prompt to the text of the structure and directives. For example, if your site focuses on artificial intelligence and you import an IT news feed, your shortcode [gpt_article] for the “Post content” template might look like this:

[gpt_article topic="%post_title%" engine="anthropic/claude-3-sonnet" sections="3" structure="Before generating the section names, analyze the article/story below and use it as a reference for the new article section structure to ensure all the essential information is included. Explain the benefits of AI technology by 2024.

THE ARTICLE FOR YOUR INSPIRATION:
%post_content_notags%" directives="Write a new, exciting article/story in a blog post format. Format the text for embedding in a WordPress post using only the <p>, <ul>, <li>, <strong>, <table>, <td>, <tr>, and <em> tags. Exclude any other HTML tags. Explain the benefits of AI technology by 2024. Use the reference article below for inspiration.

THE ARTICLE FOR YOUR INSPIRATION:
%post_content_notags%" language="English" heading="h3" max_tokens="2048" temperature="1"]

In the example above, we generate a three-section article that highlights the content of the original article in light of the benefits of AI technology by 2024. The generated articles may have more or fewer text sections, and the size in tokens may also vary. Note that we use the %post_content_notags% placeholder, which is replaced by the imported article content, stripped of HTML tags to save on token usage. If you select a language other than the source language, the article will be translated into the selected language.

Additionally, the example uses the Anthropic Claude 3 Sonnet model provided by OpenRouter.ai. You can use any other model, such as Claude 3 Opus, GPT-4 Turbo, GPT-4o, or Google Gemini Pro 1.5.

Note that the topic of our post (“Explain the benefits of AI technology by 2024”) and the content of the original article are included twice: once to create headings for the text sections of our post, and again to generate the post content. Of course, you can use the section1, section2, etc. parameters to fine-tune the content of each generated post section and specify the HTML elements you want to include.

The shortcode provided above will generate complete articles on the benefits of AI technology by 2024 in the context of the imported articles. Each generated text will include the specified HTML elements, such as bold fonts, lists, tables, etc. However, it will not include the original illustrations from the imported articles, as we did not ask the AI model to include them, and we provide it with the original text without any HTML elements, including images (note %post_content_notags%).

Accordingly, you can instruct the plugin to generate unique images for your articles using DALL·E 3, Stable Diffusion, or Midjourney in the “Media enrichment” tab of your feed settings (this method is detailed in this article). However, if you want the generated WordPress post to include images from the original article, you should explicitly ask the AI model to do so with additional prompts. You can do this both in the main prompt contained in the directives parameter and in the specific prompts for each section of the generated article where you want to retain the original images. Here is an example of a modified shortcode for this:

[gpt_article topic="%post_title%" engine="anthropic/claude-3-sonnet" sections="3" structure="Before generating the section names, analyze the article/story below and use it as a reference for the new article section structure to ensure all the essential information is included. Explain the benefits of AI technology by 2024.

THE ARTICLE FOR YOUR INSPIRATION:
%post_content_notags%" directives="Write a new, exciting article/story in a blog post format. Format the text for embedding in a WordPress post using only the <p>, <ul>, <li>, <strong>, <table>, <td>, <tr>, and <em> tags. Exclude any other HTML tags. Explain the benefits of AI technology by 2024. Use the reference article below for inspiration.

THE ARTICLE FOR YOUR INSPIRATION:
%post_content%" language="English" heading="h3" max_tokens="2048" temperature="1" section1="If there is an image in the source article, insert it at the top of the generated text."]

Note two important points:

  1. In the directives parameter, we now use the placeholder %post_content%, which includes all the HTML code of the imported article, thus containing links to the images it includes.
  2. In the section1 parameter, we explicitly ask the AI model to insert the found image from the imported article (if any) at the very beginning of each generated post, i.e., above the text of the first section.

Now let’s focus on the titles of the generated posts, since we’ve only discussed the nuances of creating post content. If our goal was to simply rewrite the content, the shortcode for the “Post title” template might look like this

[or_text model="anthropic/claude-3-sonnet" prompt="Reword this: %post_title%"]

However, since we are generating new posts not only based on the content of the imported articles, but also based on a specified topic, a more appropriate approach is to generate a new title based on the generated content. So your shortcode might look like this:

[or_text model="anthropic/claude-3-sonnet" prompt="Analyze the following article and create a title for it. Don't use colons, double quotes, quotation marks, etc. Keep it short.

The article:
%post_content_notags%" max_tokens="20" temperature="1"]

Since we are generating the post title based on the already created post content, we need to set the correct order of generating each post element. This will ensure that the main WordPress post content is generated first, followed by its excerpt and title. The Template Processing Order settings should look like this:

CyberSEO Pro - Template processing order

Now let’s consider generating meta for SEO titles and descriptions of your posts. The feed settings include templates for “Meta title” and “Meta description”. Here are examples of what they might look like:

Meta title

[or_text model="anthropic/claude-3-sonnet" prompt="Generate a search engine optimized meta title for an article based on the following information. No more than 70 characters! Include 1-2 keywords.

Keywords: %categories%
Original title: %post_title%

The article:
%post_content_notags%" max_tokens="15" temperature="1"]

Meta description

[or_text model="anthropic/claude-3-sonnet" prompt="Generate a search engine optimized meta description for an article based on the following information. No more than 160 characters! Include 1-3 keywords.

Keywords: %categories%
Original title: %post_title%

The article:
%post_content_notags%" max_tokens="40" temperature="1"]

In these examples, the shortcodes use the list of categories from the imported article as keywords. You can specify your list of keywords separated by commas. Note that all shortcodes support Spintax, so you can use randomly changing keyword lists. Spintax can be used in any shortcode parameter, such as directives, structure, sectionN, the number of generating sections, or even as the name of the AI model, which can also be randomly selected when generating each new post.

Of course, this is just an example demonstrating the capabilities of the [gpt_article] shortcode for generating new unique content based on the information obtained by the AI model when analyzing articles from your imported content feeds. While RSS Retriever can only work with RSS and Atom feeds, CyberSEO Pro can handle almost any public content sources on the web, including RSS, Atom, XML, JSON, and CSV of any structure. Moreover, this method works well, for instance, when importing product descriptions from Amazon and other marketplaces.

In this article, we used the [or_text] shortcode, which allows you to select from more than a hundred language AI models provided by OpenRouter.ai from a single account. You can use your chosen AI models directly through their official APIs. Use the [openai_gpt], [claude], and [gemini] shortcodes, respectively.

Remember that all the shortcodes provided above are simplified examples of using the [gpt_article] shortcode in CyberSEO Pro and RSS Retriever plugins. Instead of copying them directly into your feed settings, analyze how they work and create your own optimized shortcodes tailored to your specific autoblogging strategy.

Controlling the content of text sections in the [gpt_article] shortcode

Source: https://www.cyberseo.net/blog/creating-unique-articles-from-imported-rss-feeds-using-ai/


Leave a Reply