I created an automation system that streamlines the process of generating news articles for a specific audience. This system utilizes Make.com, a no-code platform, combined with AI tools like GPT-4 and DALL-E 3. By automating the content creation process, I can focus on curating and editing high-quality articles rather than spending time on repetitive tasks.
The automation begins with identifying relevant websites that cater to the interests of my audience. Once I have this list, I can generate RSS feeds from these sites, allowing the automation to pull in new articles as they are published. The system then filters out irrelevant content, ensuring that I only receive news that matters to my readers.
This automation not only saves time but also enhances the relevance of the content I publish. With AI handling the initial drafts, I can concentrate on refining the articles to ensure they meet the needs of my audience.
Step 1: Identifying Relevant Websites
The first step in my automation process involves compiling a list of websites that publish content relevant to my audience. For instance, in the context of a home security blog, I focus on major brands and their product releases. I gather URLs that link directly to sections of these websites where newsworthy content is published.
This approach goes beyond simply linking to the homepage of a website. I pinpoint specific pages, such as product innovation sections or press release archives, that provide timely updates about new offerings in the market.
By creating this targeted list, I ensure that the automation has a solid foundation from which to draw relevant articles. This step is crucial because it sets the stage for the subsequent steps in the automation process.
Step 2: Creating RSS Feeds
After identifying the relevant websites, the next step involves creating RSS feeds from them. I use a user-friendly tool called RSS App, which simplifies the process of generating feeds without requiring any coding skills. This tool allows me to quickly convert the selected URLs into RSS feeds, making them accessible for my automation.
To create a feed, I enter the URL into RSS App and let the system generate it automatically. However, I always double-check the output to ensure it captures the necessary news items. If the automatic feed includes irrelevant content, I can use the RSS builder feature to refine it further.
Creating multiple feeds is straightforward, and I can easily manage them within the app. Once I have all the feeds set up, I bundle them together into a single aggregate feed. This bundled feed becomes the source for my automation, allowing it to pull in new articles efficiently.
Using RSS App for Feed Generation
RSS App provides an intuitive interface for generating feeds. After logging into the platform, I select the “New Feed” option and input the relevant URLs. The app will attempt to automatically create the feed, but I can also manually configure it if needed.
Once the feed is generated, I review it to ensure it displays the correct number of news items. This verification step is essential to confirm that the feed is functioning as intended.
In cases where the automatic feed captures unwanted items, I utilize the RSS builder feature. This allows me to select only the news items I want, ensuring that the feed remains focused on relevant content.
Filtering Non-Relevant News Items
An important aspect of my automation is filtering out non-relevant news items. Once the RSS feeds are set up and aggregated, the next step involves implementing a filtering mechanism using AI. I leverage OpenAI’s GPT-4 to assess the relevance of each article pulled from the feeds.
The filtering process works by analyzing the content of each article and assigning a relevance score. I define specific criteria for what constitutes relevant content based on my audience’s interests. If an article scores below a predetermined threshold, it is excluded from further processing.
This filtering mechanism helps ensure that only the most pertinent articles make it through to the drafting stage. It allows me to maintain a high standard of quality in the content I publish, focusing on what truly matters to my readers.
Step 3: Setting Up Automation on Make.com
With the RSS feeds created and the filtering criteria established, the final step is to set up the automation on Make.com. This no-code platform enables me to create workflows that integrate the various components of my automation seamlessly.
I start by adding an RSS module to watch for new items in the bundled feeds. When a new article is published, the automation triggers a series of actions that include scraping the article text and filtering it based on relevance.
By connecting different modules within Make.com, I can streamline the entire process from feed monitoring to article generation. This setup allows me to efficiently create drafts based on the most relevant news items, which I can then edit and publish on my website.
Fetching News Item Text
I created an automation that fetches the text of news items from the RSS feeds I’ve set up. This is done using a web scraping platform, which is efficient for extracting content from web pages. I prefer using ScrapeTO for this task due to its simplicity and effectiveness.
After setting up my ScrapeTO account and connecting it to Make.com, I can easily scrape the text from the URLs provided in the RSS feed. The scraping process allows me to capture only the relevant text from each article, which is crucial for maintaining quality in the news items I generate.
Using the API key from ScrapeTO, I input the URL from the RSS item to get the article’s content. The tool provides options for response formats, and I choose a joined format for simplicity. This lets me get a clean output of the article text, ready for the next steps in the automation process.
Implementing AI Filtering with GPT-4
Once I have the scraped text, the next step is to implement AI filtering using OpenAI’s GPT-4. This filtering mechanism assesses the relevance of each article based on predefined criteria that align with my audience’s interests.
I set up a module in Make.com that sends the scraped content to GPT-4, along with a system message that instructs the AI to evaluate the news item’s relevance. The AI outputs a relevance score, a rationale for that score, and a proposed new title for the article.
This structured output from GPT-4 is essential for determining which articles will proceed to the drafting stage. I assign a relevance threshold, ensuring that only articles with a score above this threshold make it through. This step is crucial for maintaining high-quality content that resonates with my audience.
Generating the Draft Article
After filtering the articles, I move on to generating the draft articles using GPT-4. I provide the AI with the relevant news item text, the new title, and the URL for context. The system message instructs GPT-4 to write a short, objective news article based on the content provided.
This part of the automation is designed to produce engaging articles that are factual and free of marketing jargon. I want the content to remain straightforward and informative, focusing on the key points that matter to my readers.
Once GPT-4 completes the article, I review the output for coherence and quality. Although the AI generates a solid draft, I find that it sometimes needs a bit of human touch to enhance readability and engagement.
Humanizing the Article Content
To improve the quality of the draft articles, I implement another step where I humanize the content. This involves refining the language used in the articles to make them more relatable and accessible to a broader audience.
Using a second chat completion module with GPT-4, I provide instructions to edit the generated article while preserving its structure. I emphasize that the content should be suitable for both novice and experienced readers, blending technical details with practical advice.
This step often results in articles that read more naturally and engage the audience effectively. The humanized content maintains the essential facts while ensuring that it flows well and is enjoyable to read.
Generating Images with DALL-E 3
After finalizing the article content, I move on to generating images using DALL-E 3. Images can enhance the appeal of articles, making them visually attractive to readers. I start by crafting a prompt that summarizes the main topic of the article.
I pass this prompt to DALL-E 3, instructing it to create a realistic image that reflects the article’s content. This image adds visual context to the article, making it more engaging for the audience.
Once the image is created, I convert it to a suitable format for uploading to WordPress. This ensures that the article not only has compelling text but also eye-catching visuals that draw readers in.
Formatting Articles for HTML
The final step in the automation process involves formatting the articles for HTML. Since the articles are to be published on a website, they need to adhere to proper HTML syntax.
I use GPT-4 again to format the humanized article into valid HTML. This includes wrapping paragraphs in <p> tags, using <h2> for headings, and incorporating <strong> tags for emphasis.
Additionally, I ensure that a hyperlink to the original source is included in the article. This not only provides credit but also allows readers to explore more about the news item if they wish. After formatting, the article is ready for publication, ensuring that all elements are in place for a seamless upload to WordPress.
Creating Media Items in WordPress
I created an automation that generates media items in WordPress for the images produced by DALL-E 3. This process is essential for ensuring that the articles I publish not only contain engaging text but also visually appealing images.
To start, I need to add a module in Make.com that connects to my WordPress site. I select the option to create a media item. This requires the API key from the Make.com extension I installed on WordPress. Once connected, I can easily upload images generated by DALL-E 3.
For the file input, I use the binary data from the image conversion process. This ensures that the image is uploaded correctly. I also set a unique file name for the image to avoid overwriting existing files. Including a timestamp in the file name helps maintain uniqueness.
Publishing the Post on WordPress
After creating the media item, the next step involves publishing the post on WordPress. I add another module that allows me to create a post. This module requires the title, content, and type of the post. The type is set to ‘post’.
The title of the post is the new title generated earlier, ensuring it’s relevant and engaging. For the content, I provide the HTML formatted article that includes the main text and the image link. This step is crucial because it prepares the article for publication on my site.
Once all necessary fields are filled in, I set the status of the post to ‘draft’. This allows me to review and edit the content before it goes live. After confirming everything is in place, I click to publish the post, and it appears in my WordPress dashboard as a draft.
Testing the Full Automation Process
Testing the full automation process is essential to ensure everything works seamlessly. I run the automation to see how well it pulls in new articles, generates media, and creates posts. Selecting a news item from the RSS feed triggers the entire sequence.
The automation starts by fetching the news item text, then filtering it for relevance. If the article meets the relevance criteria, it proceeds to generate a draft article and an accompanying image. After generating the media item, it publishes the post as a draft on WordPress.
During testing, I monitor each step to confirm that the content is correctly formatted and that the images are uploaded. It’s crucial to check the final output in WordPress to ensure the articles appear as expected. Any issues encountered during testing can be addressed by adjusting the automation modules or prompts used.
Reviewing Generated Articles
Reviewing the generated articles is a vital part of the process. Once the articles are published as drafts, I take the time to read through each one carefully. This allows me to assess the quality of the content and make any necessary edits.
During the review, I check for coherence, factual accuracy, and overall readability. While the AI-generated drafts provide a solid foundation, I often find that they need a human touch to enhance engagement. I focus on ensuring the language is relatable and that the information is presented clearly.
Additionally, I look at the images included in the articles. Sometimes, the generated images may not fit perfectly with the content, so I consider replacing them with more relevant visuals. This step helps to ensure that my posts not only inform but also resonate with my audience.
