Content sites that grow predictably do not publish 100 disconnected posts. They publish 8-15 pillar topics, each surrounded by 10-20 supporting cluster articles, all internally linked. The hard part is picking pillars that have real demand and not building a 15-article cluster around a keyword nobody searches for. This workflow uses AI as a structured brainstorming partner, then forces every candidate through a demand check in a real keyword tool before it earns a slot on your calendar.
TL;DR
- Brief an AI model (GPT-5.5, Claude Opus 4.7, or Gemini 3.1 Pro) on your niche, current pillars, and top-10 traffic articles, and ask for 5 adjacent pillar candidates with 6-8 cluster topics each.
- AI is excellent at adjacency and idea volume; it is unreliable on current search volume. Verify the top 2-3 cluster topics per candidate in Google Keyword Planner (free, shows ranges), Ahrefs, or Semrush before committing.
- Pick 1-2 pillars per quarter, not 3+. Each pillar needs the pillar page plus 6-10 cluster articles at quality, which is 8-22 articles per pillar over 90 days.
- Ship a few cluster articles first, land the pillar page in week 3-4, then ship the rest. The pillar needs clusters to link into it.
- Cross-link every cluster page back to its pillar with descriptive anchor text. In 2026 these internal links double as signals AI answer engines use to map which page is authoritative on a topic.
What this covers
A repeatable process to identify pillar topics adjacent to your existing strengths, generate 6-8 cluster ideas per pillar, verify search demand, and produce a 90-day publishing calendar with an internal-linking map. The output is a one-page plan you can hand to a writer, or to future-you, and start shipping against immediately.
Who this is for
Content site owners planning the next 3 months, SEO managers running editorial calendars, niche bloggers expanding into adjacent topics, and indie creators trying to compound traffic instead of chasing one-off viral hits. Less useful for news sites or single-author personal blogs where the brand IS the topic and clustering is incidental.
When to reach for it
Quarterly planning, entering a new sub-niche, after a Google update that exposed weak topical authority, or when an existing cluster has matured and you need to decide what comes next. The hub-and-spoke model itself is well documented if you want the SEO theory behind it; Search Engine Land’s topic clusters guide is a solid reference. Once a cluster ships, its pillar category page becomes the center of gravity, so run the AI category-page audit tutorial on it before duplicate intros and weak internal linking quietly cap the cluster’s ranking.
Before you start
- List your existing pillars and your top 10 highest-traffic articles. The plan extends from strength, not a blank slate.
- Have Google Search Console open and a keyword tool ready. AI demand estimates are directional only and must be checked against real numbers.
- Decide your capacity honestly: how many articles per week can you ship at quality? Plan for 70% of that number to leave room for editing and unforeseen depth.
- Define what “winning” looks like per pillar: ranking the pillar page, ranking 5 cluster pages, or hitting a traffic target.
Which keyword tool, and what each one costs
You do not need a paid tool to run this workflow, but you do need at least one source of real volume data. The free option is enough to start.
| Tool | Cost (as of June 2026) | What it gives you | Best for |
|---|---|---|---|
| Google Search Console | Free | Actual impressions, clicks, position for queries you already rank for | Finding clusters where you have early traction |
| Google Keyword Planner | Free (Google Ads account) | Search volume as ranges (e.g. 1K-10K) unless you run ads | Prospecting brand-new topics on a budget |
| Ahrefs | Starter $29, Lite $129, Standard $249 | Keywords Explorer (170+ countries, clickstream volume), Content Explorer on Standard | Backlink-aware research, precise volume |
| Semrush | Pro $139.95, Guru $259.95, Business $499.95 | Keyword Magic Tool (25B+ keyword DB), built-in clustering | Clustering and the Content Marketing toolkit |
Both Ahrefs and Semrush bill roughly 17% cheaper annually. Use Search Console for topics you already touch and Keyword Planner (or a paid tool) for prospecting new ones.
Step by step
Use any of the three frontier models for the brainstorming steps. As of June 2026 GPT-5.5, Claude Opus 4.7, and Gemini 3.1 Pro all carry roughly a 1M-token context window, so you can paste your full top-10 list, current sitemap, and category descriptions into one prompt instead of summarizing.
- Brief the AI on your site. Paste real data: “Niche: X. Current pillars: A, B, C. Top 10 articles by traffic with URLs: [list]. Audience: [who]. Goal: identify 5 candidate new pillars adjacent to existing strengths.” The more concrete the brief, the less the model invents.
- Get pillar candidates. Ask for 5 candidates. For each, require: why it is adjacent, what audience overlap exists, what the pillar page would cover, and a confidence rating 1-5.
- Generate cluster topics per candidate. “For each of the 5 candidate pillars, list 6-8 cluster article topics. Mix high-search-volume, long-tail, and supporting how-to angles.”
- Verify demand. Take the strongest 2-3 cluster topics per candidate to your keyword tool. If a candidate pillar has zero topics with measurable search volume, drop it. AI is great at adjacency and terrible at demand, so this step is non-negotiable.
- Pick 1-2 pillars for the quarter. Resist 3+. Each pillar needs the pillar page plus 6-10 supporting articles at quality, which is 8-22 articles per pillar over 90 days.
- Map internal linking. For each chosen pillar, ask: “Map the internal links: which cluster articles link to the pillar, which link to each other, and where do existing articles outside this cluster link in?” Draw it as a hub-and-spoke with cross-cluster bridges. Every cluster page links back to the pillar with descriptive anchor text containing the pillar’s target term.
- Generate the calendar. “Sequence the articles over 12 weeks: the pillar page lands in week 3-4 (not week 1, you need cluster context first), supporting articles cluster around it, dependencies first.”
- Validate with a writer or yourself. Read the calendar end to end. Anything you cannot honestly write at a high quality level comes off the list. Replace it with something you can.
Pillar candidate prompt template
Site niche: [niche]
Current pillars: [3-5 pillar names]
Top 10 articles by traffic: [titles + URLs]
Audience: [reader description]
Task: Suggest 5 candidate new pillars adjacent to my existing
strengths. For each, give:
- Pillar name and one-sentence definition
- Adjacency rationale (which current pillar it bridges from)
- 6-8 cluster article topics with one-line angles
- 3 highest-likely-search-volume cluster topics
- Confidence 1-5 that this is worth buildingFirst-run exercise
- Pick your single highest-traffic existing pillar.
- Run the prompt above scoped to find 3 candidate adjacent pillars (not 5) so you can validate faster.
- Pick the most promising candidate and validate its 3 highest-volume cluster topics in a real keyword tool.
- If at least two of three have meaningful demand, build the cluster. If none do, the candidate fails — pick another.
Quality check
- Did the AI invent topics that sound plausible but nobody searches? Verify the top 3 per cluster in a keyword tool.
- Are the cluster topics actually clusters around the pillar, or scattered topics that happen to share a tag? They must share search intent or topical relationship.
- Did the plan include any topics you would not personally publish? Cut them — your bar matters more than the AI’s enthusiasm.
- Does the internal-linking map have any orphan clusters (clusters that don’t link to or from existing content)? Add bridges.
How to reuse this workflow
- Save the prompt template with placeholders for niche, current pillars, and goals. Reuse quarterly.
- Keep an “AI suggested but failed demand check” log. Patterns emerge: certain AI suggestions consistently overstate demand in your niche.
- Re-run the workflow after each Google algorithm update — adjacent topics that were dead before may suddenly have demand.
Recommended workflow
Niche + current pillars -> AI suggests new pillars + clusters -> verify demand in keyword tool -> pick 1-2 candidates -> cluster map with internal links -> 12-week calendar -> ship. When a cluster needs more sub-topic candidates than the demand check surfaces, sweep wider with our 30 content angles in 10 minutes brainstorm and then validate the survivors.
Common mistakes
- Adding pillars without verifying search demand. AI will happily invent topics nobody searches for, and you’ll publish 15 articles into a void.
- Skipping internal-link strategy. Clusters need cross-linking to rank; standalone articles compete with each other for authority.
- Trying to launch 5 pillars in one quarter. You will produce 15 mediocre articles instead of 8 strong ones. Pick 1-2.
- Planning clusters without a corresponding distribution stack. Pair this with AI content marketing stack guide so each cluster has a clear repurpose pipeline before it ships.
- Building a cluster around your interest instead of audience demand. The right test: would your existing top-10 readers care about this?
- Letting the pillar page ship first. Pillar pages need cluster articles to link in; ship a few clusters first, then the pillar, then more clusters.
FAQ
- How many pillars should a site have?: Most healthy content sites run 5-15. Below 5 you under-rank for any single topic; above 15 you spread effort thin. A site with 20 interconnected articles on one subject routinely outranks a site with a single 5,000-word guide, so depth beats breadth.
- How many cluster articles per pillar?: 10-20 is the sweet spot. Below 6 the cluster looks weak; above 25 the pillar page cannot link all of them prominently.
- Should pillar pages be longer than cluster pages?: Yes. Pillar pages typically run 3,000-5,000 words of high-level coverage that links to every cluster, while clusters run 800-2,000 words on one long-tail query. The pillar is the canonical authority; the clusters carry the depth.
- Can AI replace keyword research tools?: No. AI is strong at idea generation and adjacency and weak at current real search volume. Pair it with Search Console (queries you already rank for) plus Keyword Planner or a paid tool for prospecting.
- What if my niche has no obvious adjacency?: Look at audience adjacency, not topic adjacency. “Who else is my reader and what else do they care about?” produces better candidates than drilling down a topic tree.
- How often should I add new pillars?: One new pillar per quarter for an established site; one every six months once you have 10+ pillars.
- Do internal links between clusters still matter in 2026?: More than ever. Beyond passing PageRank, AI answer engines read your internal links as a semantic map to decide which page is the authority on a topic, so a cluster with no inbound links is invisible to both Google and the LLMs summarizing your niche.
Related
- Pillar / cluster page structure
- Plan long-tail keyword site
- How to Pick a Niche That Has Real Search Demand
- AI category page audit tutorial
- AI content marketing stack guide
Tags: #Tutorial #SEO #AI coding #Content cluster #Pillar / Cluster