Building a High Quality AI Content Pipeline

Jonathan Boshoff
June 2, 2024
min read

This app demonstrates a content automation workflow, which turns a target keyword and three example URLS into engaging journalistic-style content by chaining several smaller apps together. It utilizes Moonlit's "Apps as Functions" feature to stitch multiple content processing layers into one workflow.

Simply feed it information about what kind of page you’re trying to make and give it examples, and the app does the rest.

This app creates a content brief. Then turns that content brief into a first draft. Then It revises the first draft with EEAT insights.

The result is high-quality content with a journalist writing style. With sections where EEAT recommendations can be added like incorporating a quote from an expert or linking to a statistic.

It’s currently using Claude-3 Opus as a writing engine. I’m finding that this model is outperforming GPT-4 in content generation lately.

Try it here: Claude Journalist Content Workflow App

How the SEO content workflow app works:

  1. Generates an in-depth 20-point content brief
  2. Writes the first draft
  3. Performs an EEAT analysis
  4. Re-writes the first draft with EEAT recommendations
  5. Provides tasks for a research assistant to help validate and substantiate the content

What Is An AI Content Workflow?

An AI content workflow is an automation designed to use Artificial Intelligence tools like chat models (OpenAI, Claude, Gemini.) in a multi-step process to create content.

Super app

For example, this AI content workflow has four stages. A content brief, a first draft, EEAT research, and a final output.

Each step in this process contains an AI task that passes one output to the next.

By chaining together multiple prompts and apps, we achieve much higher quality outputs because each step has a separate token limit.

This multi-stage content-writing app can process over 240k tokens. I’m able to create much more complex workflows than what’s possible with a single prompt.

In contrast, OpenAI’s GPT-4 context window (token limit) is currently 32k tokens. This app can process 15x more tokens!

1. Generating an in-depth 20-point content brief

The first step in this workflow is to create an informative content brief.

It does this by using a nested app that I previously created, which produces a 20-point content brief for an SEO-friendly page to be created.

Check out the 20-point content brief app and try it for yourself.

2. Writing the first draft

SEO content maker app

The content brief is automatically passed to the next app, which writes the first drafts of content.

This is a separate app that I recently built, which uses GPT-4 to create the initial skeleton of content based on a content brief.

Try the simple SEO content writer here

This content is far from perfect. Which is why it’s fed to the next step, the EEAT analysis.

3. Performing an EEAT analysis

EEAT app cooking up recommendations

The EEAT analysis reviews three top-ranking pages to learn how they have used experience, expertise, authoritativeness, and trustworthiness in their content.

Then it provides specific recommendations on improving the first draft of content to incorporate similar insights into the content you’re working on.

The EEAT analysis and first draft are fed to the final app, a journalist-style writer powered by Claude-3 Opus.

4. Creating a finalized draft

Finished Content With EEAT

The final step receives the EEAT analysis and first draft and works on creating the finished piece of content.

It uses a markup to indicate which parts of the text are fictional. This way, it’s able to incorporate examples of expert quotes and insights, customer stories, statistics, and more.

Chat models currently aren’t great at doing research and writing simultaneously, so we’ll leave that task to a research assistant.

Example content:

Want to see a real example of a finished piece of content this app produces?

Here’s a Google doc with a full example page on living off dividends in retirement written by this app.

Take a look! It’s decent enough to impress a skeptic.

5. Research Assistant Tasks

Research tasks

This app leaves you with a to-do list of research tasks.

Like finding expert quotes, linking to statistics, or getting media to add to the page.

These steps are for your editorial team. I haven’t automated this part yet.

One great way to get expert insights for your article is with Connectively (formerly HARO)

The Importance Of Humanizing Content, Editing, And Fact-Checking

This app produces fictional content to be swapped out with real editorial content.

It’s important to note that the insights it recommends are crucial to creating journalistic content with first-hand experience and expertise.

Search engines and readers need to see real experiences and expert insights being shared on content to trust it.

Especially if you’re making claims about the reader’s money or life (YMYL).

Ensure the final content is trustworthy and filled with up-to-date and accurate information.

You can use tools like to help speed up the process of finding these unique insights to add to your content. Here are 9 effective Perplexity research prompts to get you started.

Note: Not all content needs to be written by a traditional expert. For example, a video game collector might be an expert on Diablo lore. A mom with four dogs may be an expert at raising puppies!

Troubleshooting the app:

This app may still hit token limits in its stages. It may also experience difficulties scraping some pages and throw an error message.

If you get an error message, read it to see if it was a token limit or page scraping error.

For token limits, try slightly shorter pages.

For page scraping errors, try scraping different URLs.

Customize This Workflow And Build Your Own

This app was built by chaining 4 apps together into a super app. It produces impressively high-quality content considering it’s created purely with AI. You can tweak all the steps within this app, add more steps, change the output language, or even clone this app and build your own.

About the Author
Jonathan Boshoff

Jonathan Boshoff is an SEO consultant with an expertise in artificial intelligence. Jonathan is a leading expert in AI-driven SEO. He founded the AI SEO Academy, offering comprehensive online courses that meld SEO knowledge with cutting-edge AI applications. Jonathan develops AI-powered applications, tools, and Python scripts that optimize and automate SEO tasks.