The exact playbook we ran on our own site — founder's account, written by Hamish, who also runs the Income Stream Surfers YouTube channel. No fabricated numbers, no client NDAs to hide behind. This is the site you're reading.
When we launched Harbor, we had a product built to generate AI-written SEO content at scale. We also had the classic founder problem: the product was good, nobody knew it existed, and we couldn't credibly tell people to use AI-generated content for SEO if our own site relied on paid ads to acquire users.
The obvious answer: run Harbor on Harbor. Eat our own cooking. If the product couldn't rank our own site, we had no business selling it to anyone else.
Month zero: blank domain, zero organic traffic, zero content. Six months in: over 10,000 monthly organic visitors and a handful of citations across Google AI Overviews, Perplexity, and ChatGPT. No paid boosts, no link buying, no PR stunts. Just the workflow we built the product to enable.
What follows is the actual playbook — the strategy, the execution, what we'd do differently, and the Harbor features we leaned on. Specific hard numbers are reported as ranges or round abstractions where the exact figure would be noisy; everything directional is what we actually saw.
Everything that followed was downstream of these four decisions. Every time we were tempted to deviate, growth slowed.
We didn't split across microsites. Every article lived on harborseo.ai under a cluster that mapped to a primary topic we could credibly own.
We designed the clusters before writing anything. Each cluster had a pillar page and 20–40 supporting articles, all interlinked. Keywords slotted into the cluster map, not the other way round.
A sustainable cadence: two writers' worth of output, produced by Harbor in a few hours per day. Enough to compound; not so much that editorial quality slipped.
Every article got a 30-day check. If it didn't rank in the top 30 on its target query, it got a rework pass. Most of the rework was answer front-loading and FAQ expansion.
Briefs → drafts → editorial pass → publish → measure → rework. Every step lived inside Harbor, and the whole loop closed itself.
We started by mapping every query a Harbor user might have across their SEO journey. We grouped queries into clusters, assigned each cluster a pillar page, and drafted a spoke list of 20–40 supporting articles per pillar.
Harbor generated a full brief for each article: target query, search intent, top 10 SERP analysis, H2/H3 outline, internal link suggestions, and FAQ questions. A human editor approved or tweaked the brief in under 2 minutes.
Harbor wrote the full 3,500-word draft from the approved brief — with answer front-loading, FAQ schema, comparison tables, internal links already in place, and every stat cited. Average draft time: ~8 minutes.
A human read every article before publish. Tone adjustments, local context, adding anecdotes from our own product use. 10–15 minutes of human time per article. This step is non-negotiable — it's what keeps the site from feeling like a content farm.
Harbor pushed the final article to WordPress via the REST API — featured image, meta description, schema, category, tags, all populated. Zero copy-paste. Publishing was literally one click.
Every article hit a review queue 30 days after publish. If it was ranking top-30, we left it alone. If not, Harbor's rework agent analyzed the current top 10 SERP and regenerated the sections that needed strengthening.
In our testing, the first two months look slow and the last two look magical. Nothing changed between them except compounding.
A mid-tail query in our primary cluster started showing our article as a cited source in Google's AI Overview box. Referral traffic from AI Overviews appeared in GA4 as direct/organic the same week.
A prompt we tracked weekly started citing our pillar page as one of four sources. Small traffic volume on its own — but a leading indicator that the on-page structure (answer front-loading, FAQ schema, clean headings) was doing its job.
The site started growing ~20% month-over-month even in weeks where we published less. Internal linking density, backlink accumulation, and cluster maturity had become compounding inputs in their own right.
Month-by-month visitors, with the context of what was actually happening that month. Figures are rounded — directional, not audited.
| Month | Organic visitors | What was happening |
|---|---|---|
| Month 0 | 0 | Blank domain. No content. No rankings. |
| Month 1 | < 100 | First cluster published. Crawl starts, nothing ranks yet. |
| Month 2 | ~400 | First long-tail rankings appear. Pages 3–5 mostly. |
| Month 3 | ~1,400 | Cluster maturing. Internal linking starts compounding. |
| Month 4 | ~3,200 | First page rankings on mid-tail queries. First AI Overview citation. |
| Month 5 | ~6,500 | Compound growth kicks in. First ChatGPT citation tracked. |
| Month 6 | 10,000+ | Steady state. Traffic now grows ~20% MoM without new clusters. |
Four things we got wrong the first time. If you're starting now, skip our mistakes.
We wrote pillar pages and spokes in parallel. In hindsight, pillars first — let them mature for two weeks — then release the spokes underneath. Pillars anchor the cluster; spokes without a ranked pillar float.
We added Bing Webmaster Tools in month 3. That was ~8 weeks of SearchGPT/ChatGPT invisibility we didn't need to eat. If you care about AI search citation, Bing submission is a month-zero task.
Our most-cited pages are the ones with our own data — survey results, usage benchmarks from the product, before/after metrics. We added these in month 4. They should have been in from month 1.
We spent too long in the first month manually rewriting Harbor's drafts. By month 3 we had learned to trust the output and edit lightly. The output quality hadn't changed — our over-editing had.
If you want to run the same playbook, these are the tools — links take you to the feature page for each.
Harbor's keyword research surfaces clusters and maps intent to H2/H3 structure. The cluster becomes the content plan.
Full briefs with SERP analysis, outline, FAQ suggestions, and internal link candidates — generated in under a minute per article.
3,500-word drafts in ~8 minutes. Answer front-loading, FAQ schema, internal links, comparison tables — all baked in.
Direct REST API publish with featured images, meta, schema, and categories. No copy-paste, no 'staging in Google Docs'.
Articles that don't rank get an automated rework pass against the current top 10 SERP. The loop closes itself.
Harbor generates the briefs, writes the drafts, publishes to WordPress, and reworks the articles that don't rank — on a workflow we built by running it on ourselves.