The Guerilla Content Pipeline: GSC to Publish
A traditional blog production pipeline takes 6–10 hours per post: research, outline, draft, edit, fact-check, format, publish, distribute. At a $50/hour writer rate, that's $300–$500 per post. At a $150/hour content marketing rate, $900–$1,500.
The guerilla content pipeline produces the same output in 90 minutes. Brass-SEO uses it for blog posts on this site. The 24-post Q3 sprint that shipped in May 2026 used it for every post. Quality is verifiable in the work itself, not in a claim.
This post breaks the pipeline into five steps, with the specific tools and AI prompts that make each step run in 5–25 minutes instead of 60–120.
Quick Navigation
- The Five-Step Pipeline
- Step 1: Mine GSC for the Right Topic
- Step 2: Audit a Competing Page for AI Citability
- Step 3: Draft With AI Using the Audit Data
- Step 4: Human Edit and Fact-Check
- Step 5: Publish, IndexNow, Social
- The Time Budget
- What the Pipeline Doesn't Replace
- Frequently Asked Questions
The Five-Step Pipeline
The Brass-SEO guerilla content pipeline runs in five steps, each with a target time budget. Total: 90 minutes for a 2,000–2,500 word post optimized for both traditional search and AI citation.
| Step | What you do | Tool | Time |
|---|---|---|---|
| 1 | Mine GSC for a topic with real impression demand | Brass-SEO + GSC | 10 min |
| 2 | Audit the ranking competitor's page for AI citability gaps | Claude/ChatGPT | 15 min |
| 3 | Draft using the audit data as the brief | Claude/ChatGPT | 25 min |
| 4 | Human edit and fact-check | You | 30 min |
| 5 | Publish, IndexNow, social drafts | Git push + npm scripts | 10 min |
The compression isn't from doing less work. It's from skipping the steps where you used to do work for nothing: keyword research without data backing, drafting without competitor structure, polishing without measuring.
For the broader Brass-SEO philosophy that justifies this approach, see Guerilla SEO: The 2026 Playbook.
Step 1: Mine GSC for the Right Topic
Topic selection is where 70% of bad blog posts go wrong. The query exists or it doesn't. Brass-SEO's recommended starting move is to pull queries from your own GSC where you're getting impressions but no clicks, or where you're ranking positions 11–30 and could move into top 10 with a dedicated page.
The 10-minute process:
- Open Brass-SEO and ask: "Which queries am I getting 100+ impressions on where I rank positions 11–30 and don't have a dedicated page?"
- The AI returns a list with query, current ranking page (if any), impressions, and position
- Pick one where the intent is content-worthy (not just a navigational query)
If you don't have Brass-SEO, the manual equivalent is GSC Performance → Queries → filter Position 11–30, then cross-reference your sitemap to find queries without dedicated pages.
The output of step 1 is a single sentence: "I'm writing a 2,000-word post targeting the query [X]." Everything downstream flows from that sentence.
For the broader content gap workflow, see Finding Content Gaps Without a $200/Month Tool and I Have Keywords but No Content Ideas. Now What?.
Step 2: Audit a Competing Page for AI Citability Gaps
The fastest brief is a competitor's page with their gaps marked. Pull the top-ranking organic page for your target query. Run it through Claude or ChatGPT with this audit prompt:
Read this page: [paste the URL or content]
For each H2 section, score 1-5 on these traits:
- Self-contained answer capsule in the first 1-2 sentences (does it
pass the substance test — contains a citable fact, number, or insight?)
- Link-free capsule (any embedded links in the opening?)
- Named expert quote or cited statistic anywhere in the section
- Original data (proprietary research, case study numbers)
Output: H2 heading, score 1-5 per trait, specific gap to exploit.
The output is your content brief. Each H2 the competitor scored 1–2 on is a section where your version can win citations they're losing. The Princeton GEO study and Gnuse 2025 research are clear about what wins (see Why ChatGPT Cites Some Pages and Skips Others for the full framework).
Brass-SEO automates this with the Brass-SEO AI Citability button, which runs the same audit on any URL and outputs a scored report. The manual prompt version takes 15 minutes; the automated version takes 90 seconds.
Step 3: Draft With AI Using the Audit Data
Draft generation is where untrained AI use goes worst. The fix is constraint. Don't ask the AI to "write a blog post about X." Hand it the audit, the target query, and the structural rules. Then ask for the draft.
A working draft prompt:
Write a 2,000-2,500 word blog post targeting the query [TARGET QUERY].
Structure requirements:
- Opening: 3-paragraph intro that leads with a specific insight, not
a generic statement
- Quick Navigation section
- 6-8 H2 sections
- After each H2: a 1-2 sentence answer capsule that:
* Makes sense standalone (passes structure test)
* Contains a specific fact/number/insight (passes substance test)
* Contains no links
- Each section uses the brand name [BRAND] in the capsule (not "we" or "our")
- At least one section cites a named expert or specific study
- FAQ section with exact heading "## Frequently Asked Questions"
- 5-6 FAQ questions
Sections to cover (from audit gaps):
[paste the audit gaps from Step 2]
Brand voice: [paste 2-3 sentences from an existing post that captures voice]
Output the full draft.
Claude or ChatGPT generates the draft in about 60 seconds. The draft will be 70–85% of final quality — substantively correct, structurally aligned, but lacking some specificity and rhythm. That's exactly what step 4 fixes.
For the underlying citation rules the prompt enforces, see Part 1: Answer Capsules: The Content Trait LLMs Cite Most and Why Expert Quotes Boost AI Citations by 41%.
Step 4: Human Edit and Fact-Check
The 30 minutes of human editing is where guerilla SEO either works or fails. AI drafts have three predictable weaknesses: generic substitutions, plausible-but-unverified facts, and a slightly-off brand voice. All three are fixable in one pass.
A working edit checklist:
- Read each capsule out loud. Capsules that sound generic ("Page speed matters for SEO") get rewritten with specifics ("Google's LCP threshold is 2.5 seconds").
- Verify every statistic. AI drafts confidently cite "studies that show 73%..." that don't exist. Either find the real source or remove the claim.
- Check the brand voice. Read three paragraphs from your strongest existing post. Does the new draft match the rhythm and word choices? Rewrite sentences that drift.
- Trim banned words. Brass-SEO's standards ban "comprehensive," "ensure," "leverage," "robust," and a long list of AI-prose tells. A find-and-replace pass catches 90% of them.
- Cut binary contrast constructions. "It's not X — it's Y" reads as AI. Pick one direct statement.
Time: 30 minutes for a 2,500-word post. The draft becomes recognizably human-edited content. The reader can't tell which sentences came from the AI.
For the writing rules Brass-SEO enforces internally, the writing standards document is required reading before any blog post (docs/blog-system/WRITING_STANDARDS.md).
Step 5: Publish, IndexNow, Social
The last 10 minutes is mechanical. Three commands:
# Commit and push (triggers Vercel deploy + social draft hook)
git add _posts/your-new-post.md
git commit -m "Add: Your Post Title"
git push origin main
# Submit to Bing/Yandex for fast indexing
npm run indexnow:blog
For Brass-SEO's blog, the git push triggers a pre-push hook that auto-generates Bluesky and LinkedIn drafts via Postiz. The drafts queue up at https://dispatch.brass-seo.com for review. The whole publishing step is push-once.
If you don't have IndexNow or a pre-push hook set up, the manual equivalents are: submit the URL to Google Search Console for indexing, and post a social link manually. Adds 5 minutes.
For the IndexNow setup details, see Brass-SEO's INDEXNOW_INTEGRATION_GUIDE in the docs (referenced in BLOG_SYSTEM_USER_GUIDE.md). For the social automation, the Bluesky/LinkedIn pipeline pattern is documented in docs/marketing/SOCIAL_PUBLISHER_USER_GUIDE.md in the Brass-SEO repo.
The Time Budget
The before-and-after for a 2,500-word post:
| Stage | Traditional pipeline | Guerilla pipeline |
|---|---|---|
| Topic selection | 30–60 min | 10 min |
| Outline + brief | 60 min | 15 min |
| Draft | 180–240 min | 25 min |
| Edit + fact-check | 90–120 min | 30 min |
| Publish + distribute | 30 min | 10 min |
| Total | 6–10 hours | 90 min |
The compression is real, but it isn't magic. The pipeline depends on:
- Existing GSC data (you can't mine queries from a site with no impressions)
- A working AI subscription (ChatGPT Plus or Claude Pro at $20/month)
- Discipline on the edit pass (skip it and the output is recognizably AI-generated)
- Standards documented somewhere (without rules, the AI defaults to generic prose)
For brand-new sites with no GSC data, the pipeline reduces to a 3-step version (topic from manual research, audit a competitor, draft + edit). Still under 3 hours per post.
What the Pipeline Doesn't Replace
The pipeline replaces drafting and structural work. It doesn't replace four other things:
- First-party data. Original research, customer interviews, your own case studies. AI can't fabricate these. Posts with genuine first-party data win citations the pipeline-only posts don't.
- Strategic content choices. What topics to cover, what angles to take, what to ignore — that's the operator's judgment. The pipeline executes a decision, not makes one.
- Editorial voice. AI matches a voice you give it. Without an established brand voice document or 5+ existing posts to reference, the output is generic.
- Distribution beyond automation. Social drafts get queued, but actual community engagement (Reddit, niche forums, podcast appearances) still needs a human.
The pipeline lifts the floor. It doesn't raise the ceiling. Operators who treat it as a content-quantity machine get diminishing returns. Operators who treat it as a force multiplier (to ship more of the work that matters) see the math compound.
For the broader strategy framing, see Guerilla SEO: The 2026 Playbook and the tactical follow-ups in this series.
Frequently Asked Questions
Will Google penalize posts produced this way?
Google's published guidance (March 2024 and later) is that AI-assisted content is fine when it's high-quality and useful. Penalties target low-substance content regardless of whether a human or AI wrote it. The pipeline above keeps human editorial control on every post, which is the practical line that matters. For the broader question, see Is Your AI Content Helping or Hurting SEO?.
Can I run this pipeline without ChatGPT Plus or Claude Pro?
Partially. The free tiers of both work for short drafts but have output limits that constrain longer posts. For 2,000+ word work, a paid tier ($20/month either way) pays for itself within the first 2–3 posts in time saved.
How do I keep the AI drafts from sounding generic?
Three moves. First, build a brand voice document with specific phrases and patterns the AI should match (use 3–5 of your strongest existing posts as input). Second, feed that document into every draft prompt as context. Third, do the 30-minute human edit — there's no skipping it. Posts that go up without the edit pass read as AI to anyone paying attention.
How does this compare to a hired ghostwriter?
A ghostwriter at $300–$500 per post produces content with their voice, on their schedule, with their fact-checking discipline (good ones are excellent; mediocre ones leave you re-editing). The pipeline produces content with your voice, on your schedule, with your fact-checking. The cost difference funds a year of subscription tools. Both can produce excellent work; the pipeline is faster and gives you the option to run it again next week.
What if I want longer posts (3,000+ words)?
The pipeline scales linearly: a 3,500-word post takes about 120 minutes instead of 90. The draft step grows most, since AI output past 3,000 words usually requires breaking into sections and reassembling. For posts over 4,000 words, plan on 150–180 minutes. Still cheaper than the alternative.