See Your Whole Site As One Map. Spot the Holes. Fix Them. Build Real Topical Authority.

A topic cluster is a group of articles that all cover the same broad subject, organised around a single comprehensive “pillar page” with supporting “spoke” articles linking back to it. Clusters tell Google (and AI tools like ChatGPT, Claude and Perplexity) that your site has real depth on a topic — which is one of the strongest authority signals in modern search. Sites that cluster their content rank significantly better than sites that publish the same articles as disconnected posts. Linkilo builds these clusters for you automatically and shows you exactly where to strengthen them.

The Insight Dashboard is the storytelling view — treemap, density heatmap, funnel network, automated insights and AI gap analysis. The Interactive List View is your workspace — drag posts between clusters, set pillars, split clusters, and see all your unclustered posts in a sticky sidebar. The Smart Pillar Suggestions view ranks your strongest non-pillar candidates with their full Pillar Score breakdown so you know which posts to grow into pillars next.

Every post gets a 0–100 Pillar Score, calculated from six signals: word count (longer wins, with bonuses at 500, 1000 and 2000 words), content format (“What is”, “Resource Collection” and “How-to Guide” formats score highest), entity type (Concepts and Processes outscore products and people), funnel stage (TOFU/awareness gets a boost since pillars are usually entry points), user intent (Informational scores extra), and the number of inbound internal links the post already has. The post with the highest score in each cluster becomes the suggested pillar. You can override this with one click on the star icon — we never lock you in.

Yes. Linkilo has a foundational keyword-based clustering engine that runs without any AI calls or API keys. It uses TF-IDF analysis, n-gram extraction, primary and secondary keywords, GSC keywords, tags and categories with weighted scoring to group posts. It’s free to run, takes seconds, and gives you a perfectly serviceable cluster structure. If you want better editorial cluster names like “Buying Guides” or “Humidors & Storage” instead of keyword-based labels, that’s when you click Regenerate with AI Themes and the AI takes over.

About a tenth of a cent per regeneration on a typical site. The main AI clustering call uses gpt-4.1-mini with batched post titles plus your top primary keywords, costing roughly $0.001 (yes, one tenth of one cent) and taking 5–15 seconds. The orphan-rescue pass that catches anything the model dropped costs about $0.0002. Per-orphan AI suggestions are around 500 tokens. Even on sites with thousands of posts, you’re looking at fractions of a cent per cluster regeneration. You pay OpenAI directly — we don’t mark up API usage.

It shows up in the Unclustered Posts sidebar in the Interactive List View, with a count badge so you can see at a glance how many orphans you have. From there you can drag any post into a cluster manually, click Suggest next to it for an inline AI recommendation of the best fit, or click Auto-Cluster All to let the AI assign every orphan in one bulk operation. There’s also a final safety-net step: if anything still doesn’t fit after the rescue pass, it lands in a “Miscellaneous Content” cluster so nothing is ever silently lost.

Completely. Drag posts between clusters with full undo. Set any post as pillar with one click on the star icon. Use the Split Cluster tool to break a big cluster into two by checkbox-selecting which posts move where. Create New Cluster to start one from scratch. Delete Cluster sends posts back to the orphan pool. Manually-created clusters are tagged separately and never touched by automation — even when you regenerate AI clusters, your manual edits are preserved. Pillar pages are also drag-locked to prevent accidental destruction.

Two ratios per cluster: Pillar → Spokes (how many of your supporting articles the pillar actually links to) and Spokes → Pillar (how many spokes link back to the pillar). It uses real link parsing — we scrape the actual <a href> tags in your post content and resolve every URL (handling absolute, relative, query-stripped, and trailing-slash variations) to make sure the link count is real, not just based on metadata. Anywhere a link is missing, Linkilo’s internal-linking engine queues it as a one-click suggestion inside the relevant post.

Click the lightbulb icon on any cluster card. The AI takes the pillar topic plus the titles of the existing supporting articles, compares it to what a fully-developed cluster on that topic looks like, and returns 5–7 specific sub-topics, long-tail keywords, or user questions that are clear gaps. You get the prompts as plain English headlines you can hand to a writer or use yourself. Re-run anytime — results aren’t cached aggressively, so as you publish new posts the gap list updates to reflect what’s actually still missing.

Yes. The Insight Dashboard has live filters for User Intent (Informational, Commercial, Transactional, Navigational) and Funnel Stage (TOFU/Awareness, MOFU/Consideration, BOFU/Decision). Apply a filter and the whole visualisation reshapes — you can see, for example, just your commercial-intent BOFU clusters, which is essentially a map of your money pages and their support content. Intent and funnel stage are determined by Linkilo’s upstream AI content analysis, so they’re consistent across every post type.

Yes. Topic clustering is hooked into Linkilo’s AI content processor — once a session of new posts finishes content analysis, clustering automatically re-runs in the background to assign them to the right clusters (and create new ones if needed). You can also trigger a regeneration manually anytime from the toolbar. There’s a verification pass at the end of every run that catches any post still left orphaned and either re-attempts assignment or drops it into a “Miscellaneous Content” cluster as a last resort, so nothing is ever silently dropped.

Yes — this is what makes Topic Clusters more than a visualisation. Cluster membership feeds directly into Linkilo’s link suggestion engine in three ways. First, posts in the same cluster are surfaced as high-relevance link targets with the justification “Same topic cluster.” Second, the link suggestion AI uses cluster membership as a 0.15-weighted component in its overall composite score. Third, after clustering completes, top AI-recommended keywords are auto-assigned to posts so anchor text stays consistent across the cluster. The Cluster Health check also surfaces missing pillar↔spoke links as ready-to-add suggestions inside the relevant posts.

Yes. AI clustering batches posts at 200 per call, so a site with 5,000 posts costs roughly the same per regeneration as one with 500 (still pennies). When the AI groups too many posts into one mega-cluster, the engine automatically splits them into 2–3 sub-clusters of 70–100 posts each, naming each with a meaningful topic extracted from the post titles. The visualisations are D3.js-based and stay snappy on hundreds of clusters. The hard limit is really only your OpenAI usage tier — we’ve tested smoothly on sites with several thousand posts.

No — manually-created clusters are tagged in the database and never touched by automated runs. When you regenerate, only auto-generated clusters get rebuilt; anything you made by hand stays exactly as you left it. Old auto-clusters aren’t deleted either, just archived as “legacy_archive” in the database, so if a regeneration produces a theme set you don’t love, it’s technically possible to roll back. There’s also a separate Remove All AI Clusters button if you want to wipe automated work without touching your manual clusters.

You control which post types Linkilo touches via the autopilot post-type config. By default, only post is clustered — pages, attachments, navigation menu items, and revisions are intentionally excluded. If you want product post types or custom post types included, add them to the config and they’ll be picked up on the next regeneration. If you want to exclude something currently being clustered, remove it from the config and re-run.

Foundational keyword-based clustering is instant — click Regenerate Clusters and you see results in seconds. AI themed clustering takes 5 to 15 seconds for a typical site. The Pillar Score, gap analysis, and intent/funnel insights need posts to have completed Linkilo’s AI content processing first, so the very richest cluster experience kicks in once that’s run on your back catalogue. On a brand-new install with no AI processing done yet, you’ll still get useful keyword-based clusters and pillar suggestions; the AI-powered insights light up as content analysis catches up.