SEO content automation

What this does

You give it a topic, a primary keyword you want to rank for, and the product you want to mention. The workflow scrapes the top 10 articles ranking for that keyword. It analyzes what they all do well, builds an article structure that beats them, writes each section in your voice, and drops the finished article into a Google Doc. About 12 minutes from topic to draft.

This is the bigger sibling of "Content brief automation." Briefs stop at the structure; this goes all the way to a written article you can edit and ship.

The problem this solves

Publishing one SEO article is fine. Publishing 10 a month is brutal.

The bottleneck is rarely the keyword research. You can pull 100 keywords from Ahrefs in an afternoon. The bottleneck is everything between "I have a keyword" and "the post is live." For each article, someone has to:

• Read the top 10 search results to understand what readers actually want
• Pull out the patterns – which sections everyone covers, which ones are weak, what your article should do differently
• Write the outline
• Write the article, in your voice, with the right keyword density, with the brand mention woven in naturally
• Edit it three times so it doesn't read as AI

Done well, that's 4-6 hours per article. Done at scale, it's a full-time writer at $5-8K a month plus an editor.

The other options are the cheap AI content tools that produce articles that read like AI from sentence one. Or hiring an agency at $300-500/article that produces decent copy on a 2-week turnaround per piece.

This workflow does the heavy lifting. The output isn't a final article – it's a strong first draft that needs your editor's eye for fact-checks, opinion injection, and a brand-voice pass. End-to-end, your team spends 30-60 minutes per article on the edit instead of 4-6 hours on the writing. The math on 10 articles a month gets very different.

What you put in

A row in a Google Sheet:

• Topic – e.g. "how to build a content calendar"
• Primary keyword – the phrase you want to rank for, exactly
• Product to mention – the company / tool / service you're writing this for (gives the workflow context so the brand mention is on-tone)
• Optional: a styling guide reference (paste the URL of your best existing blog post, the workflow uses it as a voice anchor)

Set the row to "Planned" and the workflow takes it.

What you get out

A Google Doc with the full article:

• Title – keyword-natural, not robotic
• Intro that hooks on the reader's actual problem
• Section-by-section content, structured around what the top-ranking articles do well plus what they all miss
• The brand mention woven in where it's earned, not bolted on
• A conclusion that doesn't read like every other AI conclusion
• A clean, edit-ready format with H2s, H3s, bullet lists, and proper spacing

Sheet row flips to "Drafted" with a link to the Doc.

How long per article

Workflow time: about 10-12 minutes from "Planned" to finished Doc. Most of that is the SERP scraping waiting on response times.

Your team's edit time: 30-60 minutes per article. Fact-checks, opinion, a brand-voice pass, image selection.

End-to-end per article: about 1-2 hours of human time. Down from 4-6 hours. At 10 articles a month, that's 30-40 hours of writer time saved every month.

When this is a good fit

• You publish SEO content at volume (more than 5 articles a month)
• You already do keyword research – the workflow doesn't pick topics, you hand it the topic + keyword
• You have a styling guide or you're willing to write one as part of setup. Without that, the articles come out in a generic SEO-AI voice that reads obviously machine-written.
• You're willing to have your editor do a real edit pass on every article. The output is a strong first draft, not a publish-ready piece.

When this isn't a good fit

• You publish 1-2 articles a month. The setup time isn't worth it at that volume.
• You want zero human review on output. That doesn't exist. The articles always benefit from fact-checks and your team's actual opinion injected.
• You're targeting YMYL topics (medical, legal, financial advice) where an AI draft can't safely be the foundation. Those need a subject-matter expert from sentence one.
• You don't have a styling guide and don't want to invest in one. The article quality is bounded by how clearly your voice is defined.

What's actually under the hood

The workflow runs on n8n. About 12 GPT calls in sequence, plus SERP scraping and Jina for content extraction. Here's the rough shape:

1. Read the topic + keyword + product from your sheet
2. SERP scrape: fetch the top 10 organic results for the keyword
3. For each of the top 10, scrape the article content (via Jina) and summarize it (GPT call per article – 10 summaries)
4. Synthesis call: given the 10 summaries, what's the best possible article structure to beat them?
5. Outline expansion: turn the structure into a detailed section-by-section outline, factoring in the primary keyword and the product context
6. For each section, a separate GPT call writes the content. Section-by-section instead of one mega-call because the model loses focus on the article's overall thread otherwise.
7. Intro + conclusion get their own dedicated calls (they have specific structural rules that differ from body sections)
8. Compile everything into a Google Doc with proper formatting
9. Update the sheet row to "Drafted"

The reason this works (and most "AI article writers" don't) is the section-by-section approach. Asking GPT to write a full 2000-word article in one call produces flat, repetitive prose. Each section gets its own prompt that knows:

• What this specific section needs to cover (from the outline)
• What goes before and after (so transitions feel earned)
• The brand voice rules
• The reader's stage of attention at that point in the article

The prompts took 30+ rounds against real customer blogs to get the output to a "writers actually want to edit this" quality. Generic prompts produce generic articles. The output quality is in the prompt chain.

What you own at handover

• The full n8n workflow file
• Every GPT prompt in plain text, documented
• The Google Sheet templates (input + tracking)
• A styling guide doc – either yours, or one we'll write together during setup based on your existing top posts
• Optional: a wired-in WordPress / Webflow push from approved articles, so the Google Doc flows straight to CMS draft after edit
• A Loom showing the end-to-end loop and where to tweak when your voice or funnel shifts
• A runbook covering the common edge cases: what to do when SERP scraping returns a paywall page, how to handle topics where the top 10 are all video results, how to rerun individual sections without redoing the whole article

Why I can help

The wiring of an "AI article generator" is not the moat. n8n + GPT + a SERP API is something anyone can build in a weekend. The result is the kind of article that gets 200 organic visits a month and looks like every other AI-generated SEO article.

The moat is what makes the output not read as AI:

• Section-by-section generation with section-specific prompts, not one giant call
• A SERP analysis step that informs structure, not a "generate an outline" prompt that produces the same outline for every keyword
• Brand voice rules baked into every section call, not bolted on at the end as a "rewrite in our tone" pass
• Editorial structure rules that match how readers actually skim long-form SEO content
• A separate process for the intro and conclusion, because they need different rules than body sections

I've tuned this prompt chain across 5 different B2B SaaS verticals in 2025. Each one needed adjustment for their specific voice, funnel, and target audience. The prompts that ship to you are the ones that came out of that work, customized to your blog during setup.

What it costs to run

Per article: about $0.20-$0.50 in OpenAI tokens (GPT-4.1-mini for most steps, GPT-4 for the synthesis call). SERP scraping API: $0.05-$0.10 per article. Jina scraping: usually free tier covers it. Total: under $1 per article.

Build cost: 2+ weeks of my time to wire the workflow, tune the prompts to your voice + funnel, set up the sheets, write the styling guide if you don't have one, and train your editor on the loop.

After build, you're running it forever on your own infrastructure.

How to start

Book a call. Bring 3 topics + keywords from your backlog that you've been wanting to publish. We'll run the workflow on one of them during the call. You decide on the spot whether the draft passes the "I'd hand this to my editor" test.