YouTube to Blog Post Automation for Pennies with Make.com

Most creators pay $50 a month for a tool that turns YouTube videos into blog posts, and still end up with generic output they have to rewrite. I built a system that does the same job for 1–5 cents per post, gives me full control over structure and affiliate links, and publishes a formatted WordPress draft in under two minutes.

The whole thing runs on Make.com, which handles the orchestration, and Apify for transcript and thumbnail extraction. Airtable acts as the trigger layer, one status change kicks off the entire pipeline.

How the pipeline works#

The trigger is an Airtable row. When you paste a YouTube URL and set the status to "Generate Blog Post," Make.com wakes up and starts pulling data.

The first step uses Apify's YouTube Transcript Ninja actor to scrape the full video transcript. A second Apify actor pulls the thumbnail and video description. Together, those three things, transcript, description, and thumbnail, become the raw material that goes into Claude.

Claude Sonnet 3.7 handles the writing. I fund my Anthropic developer account with $10 at a time and have barely spent $4 in a month of heavy use. The model receives everything Apify scraped, plus a detailed system prompt that specifies the exact structure, tone, SEO formatting, and affiliate link placement I want. It outputs the full post in HTML, ready to push directly to WordPress.

The Make.com WordPress module creates the draft, sets the featured image from the scraped thumbnail, and writes the SEO slug and meta description. The last module updates the Airtable row to "Blog Post Generated" so you don't accidentally re-run it.

That's four core modules. That's the whole system.

Where the affiliate links actually live#

The affiliate link injection isn't a post-processing step, it happens at the prompt level. In Airtable, there's a dedicated "Affiliate Links" field on each row. When you paste in your links before triggering the automation, Make.com passes that field as a variable directly into the Claude module.

The system prompt instructs Claude to include those links naturally in the body copy using HTML anchor tags, with anchor text kept to one to five words. When the WordPress draft appears, the links are already there and working. I tested this live, clicked the affiliate link in the generated post, and it redirected correctly.

If you want to understand how I think about the broader automation stack this lives inside, my full YouTube workflow breakdown covers the other pieces.

The three mistakes that kill these automations#

Building the Make.com workflow is the easy part. The quality of what comes out lives or dies on decisions most people get wrong before they even run the automation once.

Not giving the AI enough context. A bare transcript fed into a generic prompt produces a generic post. Claude needs to know the structure you want, the tone, whether to use H2s or H3s, how to handle links, what SEO signals to prioritize. The more specific the system prompt, the less cleanup you do on the output.

Not iterating on the system prompt. If you run the automation and get a mediocre article, hitting regenerate with the same prompt gives you the same mediocre article. Every time output misses the mark, that's a signal about something missing or wrong in the prompt. Fix it, save the updated version, and that improvement compounds across every future post you generate.

Over-engineering the workflow. "It is not about having as many nodes as possible. It's really about the output." Four modules is enough. If you're adding nodes to feel like you've built something sophisticated, you're solving the wrong problem. More operations means more points of failure and more time spent debugging.

What the system prompt actually does#

I won't paste the full prompt here, it's long and worth taking a screenshot from the video, but a few things it handles that are easy to miss:

• Instructs Claude to embed the YouTube video URL as a playable embed in the post body
• Specifies HTML formatting so the output drops cleanly into WordPress without conversion
• Tells Claude to include any links from the video description as natural anchor tags
• Explicitly says not to fabricate facts, only use what's in the transcript and description
• Adds NLP-friendly writing instructions to help the post read as human-written

The "no fabrication" instruction matters more than it sounds. Without it, Claude will sometimes fill gaps in the transcript with plausible-sounding details that aren't accurate. Grounding it to only the provided context keeps the post honest.

If you want your repurposed posts to actually rank, pair this system with a keyword targeting step before you write. I use a free SEO Keyword Generator GPT to find keywords your repurposed posts can actually rank for before the automation runs.

What this actually costs#

At roughly one cent per Apify run and a fraction of a cent per Claude generation at 5,000 max tokens, a single post costs between one and five cents depending on video length. Airtable's free plan covers the trigger layer. Make.com's free plan covers the operations if you're running a modest volume. WordPress is whatever you're already paying for hosting.

The $50/month SaaS tools aren't doing anything this pipeline can't do. What they're selling is convenience and a UI. What you give up is control, over the prompt, the formatting, the affiliate placement, the output structure. This system gives all of that back.

Watch the full video on YouTube: https://youtu.be/lj5yVCnnQR0

This post contains affiliate links. I only recommend tools I actually use.