SEO Agents
Paste a raw keyword list and get back a site structure: clusters, pillar pages and internal links.
Drop in your exported keyword list — from any research tool, any size — and the agent groups terms by topic and search intent, then turns the groups into a plan: which pillar pages to build, which supporting pages feed them, what URLs to use, and how the pages should link to each other.
Try an example
Keyword research tools are good at producing lists and bad at telling you what to do with them. A thousand exported terms is an inventory, not a strategy. The Keyword Cluster Agent closes that gap: it reads your raw list and reorganizes it into the structure a website actually needs — topics, subtopics, and the intent behind each group.
It's for SEOs planning a content program, founders deciding what pages their site needs, and agencies turning a research deliverable into an architecture a client can approve. You bring the list; the agent brings the organizing logic.
The input is as simple as paste: keywords one per line, with volumes if you have them. The agent clusters terms that belong on the same page — because splitting 'crm for freelancers' and 'best crm for freelancers' across two pages just makes them compete — labels each cluster with its dominant intent, and identifies which terms are questions, comparisons or purchase-stage searches.
What separates this from a grouping spreadsheet is that it doesn't stop at groups. Each cluster becomes a recommendation: a pillar page that owns the topic, the supporting pages that deepen it, suggested URLs, and the internal links that should connect them. You get the skeleton of a site section, not just sorted rows.
The plan doesn't have to live in a slide deck. Inside Dual7 you can act on it directly — draft the pillar page, brief the supporting articles, and build the section on a codebase you own. And because recommendations come with reasoning, you can adjust the groupings before committing a quarter of content budget to them.
Terms that answer the same search are grouped together, so your pages stop competing with each other for the same query.
Each group is marked informational, commercial or transactional, so you know whether it wants a guide, a comparison or a product page.
Clusters roll up into a hub-and-spoke plan: the pillar page that owns the topic and the supporting pages that feed it authority.
Every recommended page comes with a clean URL slug, so the plan translates directly into a sitemap.
The agent specifies which supporting pages should link where, so linking is designed into the architecture instead of retrofitted.
Paste output from whatever research tool you use — no integrations to configure, no format conversions to fight.
Groupings come with explanations, so a content roadmap review becomes a discussion about choices, not a leap of faith.
Drop in your export — one term per line, volumes optional. Any research tool's output works.
The agent groups terms by topic and intent, and shows its reasoning so you can merge or split groups as you see fit.
Each cluster becomes pillar and supporting page recommendations with suggested URLs.
Use the internal link map to structure the section, then draft pages directly in Dual7 or hand the plan to your writers.
Terms are clustered by meaning and the page that should answer them, not by shared words — 'crm for freelancers' and 'freelance client management software' belong together even though they share no vocabulary.
Every cluster is labeled by what the searcher wants: to learn, to compare, or to act. Intent determines the page type the agent recommends, so you don't brief a blog post for a buying query.
By deciding which terms share a page, the agent heads off the classic problem of two similar pages splitting each other's visibility before either page exists.
Broad, high-level clusters are flagged as pillar candidates — the pages that anchor a topic — with the narrower clusters assigned as their supporting content.
Recommended pages arrive with clean, consistent slugs that reflect the cluster's primary term, ready to drop into a sitemap or CMS.
The output names the links each supporting page should carry — up to the pillar, across to siblings — so topical structure is expressed in actual links, not just intention.
Your full list grouped into themes, each with a primary term and its supporting keywords.
Informational, commercial or transactional classification per cluster, with a note on the evidence.
The hub pages to build, the clusters they own, and suggested titles and URLs.
The articles and sub-pages that feed each pillar, in a suggested build order.
Which pages should link to which, with anchor text direction, ready to design into templates.
Example output
Input: an exported list mixing 'accounting software for landlords', 'how to reconcile rent payments', 'quickbooks alternatives' and 397 more. Output: five pillar pages (small business, landlords, freelancers, comparisons, pricing), twenty-three supporting articles with slugs, each cluster intent-labeled, and a link map connecting every supporting page to its pillar and two siblings.
An SEO lead pastes the quarter's research export and walks into the planning meeting with pillars, supporting pages and a build order instead of a raw spreadsheet.
A founder with a seed keyword list learns which pages the site actually needs before any design or copy work begins.
An account strategist converts a client's keyword export into a page architecture with reasoning attached, ready for approval.
An editor checks whether planned articles overlap with existing pages by clustering the backlog against the live sitemap's keywords.
A merchandising manager groups product-adjacent terms to decide which buying guides and comparison pages the catalog needs.
A content marketer organizes two years of scattered post ideas into topic hubs, revealing which themes are saturated and which are bare.
A consultant clusters the terms an existing blog ranks for and finds three posts competing for the same intent — a consolidation plan in one pass.
Software companies turn sprawling feature and comparison terms into structured hubs around use cases and personas.
Stores organize product-adjacent research terms into buying guides and category content that support the catalog.
Firms cluster service questions into practice-area hubs that demonstrate depth rather than scattered posts.
Editorial teams plan coverage by topic cluster instead of chasing individual keywords article by article.
Clinics group patient questions by condition and treatment, building resource sections that match how people search.
Strategy teams convert research deliverables into site architectures clients can understand and approve.
You provide
300 keywords about project management software — mix of 'what is', 'best', 'vs' and pricing terms
The agent
The agent separates informational definitions from commercial comparisons and transactional pricing terms, then clusters within each intent group.
You get
A plan with a 'project management software' pillar, comparison pages for the major 'vs' terms, a pricing page owning the cost cluster, and guides feeding the pillar — each with URLs and link targets.
You provide
Our export has 'wedding photographer prices', 'how much does a wedding photographer cost', 'wedding photography packages' — do these need three pages?
The agent
The agent groups all three into one commercial-investigation cluster, since searchers want the same answer.
You get
One recommended pricing-and-packages page targeting the cluster's primary term, with the variants folded in as sections — and a note that three separate pages would compete with each other.
You provide
Keyword list for a plumbing company serving three cities
The agent
The agent separates service clusters from location modifiers and identifies which combinations justify dedicated pages.
You get
Service pillars (boilers, drains, emergencies), location pages where search demand supports them, and a link structure connecting each location to the services offered there.
You provide
Two years of blog post ideas for a HR software company, 180 terms
The agent
The agent clusters the backlog, revealing which themes have depth and which are one-post wonders.
You get
Six topic hubs with existing ideas slotted in as supporting content, gaps flagged per hub, and a build order prioritizing hubs with commercial intent.
You provide
Cluster these 500 terms and tell me what to build first for a new CRM product
The agent
The agent clusters the list, then sequences recommendations by intent stage and likely business value.
You get
A phased plan: comparison and alternative pages first (closest to purchase), use-case pillars second, educational guides as ongoing support — all internally linked.
| Aspect | With the agent | Manual process |
|---|---|---|
| Input | A pasted export from any keyword tool | The same export, plus a spreadsheet and a free afternoon |
| Grouping logic | By meaning and the page that should answer the search | Usually by shared words, which splits synonyms apart |
| Intent handling | Every cluster labeled informational, commercial or transactional | Intent tagged by hand, row by row, if at all |
| Output | Pillar pages, supporting pages, URLs and link map | Sorted rows that still need someone to design the architecture |
| Cannibalization check | Overlapping terms are merged onto one page by design | Discovered after two pages already compete |
| Internal linking | Links specified as part of the plan | Retrofitted months later during an audit |
| Time for 500 keywords | One session from paste to page plan | Days of sorting, labeling and second-guessing |
Paste five hundred terms and they all get processed. A chat window truncates, forgets, or quietly drops rows from long lists.
A chatbot can suggest groupings; the agent carries them through to pillars, URLs and internal links — the decisions a site is actually built from.
Clustering rules apply evenly from row one to row five hundred, instead of drifting as a chat context fills up.
Inside Dual7, recommended pages can be drafted and published on your domain — the roadmap and the build happen in one place.
Volumes let the agent sequence recommendations sensibly. Without them you'll still get clusters, but prioritization leans on intent alone.
Remove branded terms for competitors you're not targeting, duplicate rows and obvious junk. Better input means tighter clusters.
If a cluster is commercial, don't brief it as a blog post. Intent is the single strongest signal for what page type will satisfy the search.
When the agent groups terms together, trust it. Splitting a cluster into multiple pages because the keywords look different recreates cannibalization.
A pillar page with no supporting content is a promise with no evidence. Plan each hub with at least a few supporting pages from the start.
Internal links work best when they're written into the content. Hand the map to whoever drafts each page so links appear naturally.
New products and new markets change which terms matter. Re-run the list rather than bolting new pages onto an old architecture.
It's grouping keywords that should be answered by the same page. Searchers using 'crm for freelancers' and 'best crm for freelancers' want the same thing, so one page should target both. Clustering turns a raw list into a set of page-level decisions.
Research finds the terms; clustering organizes them. This agent assumes you already have a list — from any research tool — and answers the question research tools leave open: which terms belong on which pages.
One keyword per line, pasted into the text area. Search volumes are optional but help with prioritization. Exports from any research tool work — just paste the keyword column.
By meaning and intent: terms a single page could satisfy belong together, even when they share no words. It also uses intent signals — 'vs' and 'best' suggest comparison pages, 'how to' suggests guides — so clusters map to real page types.
A pillar is the hub page that owns a broad topic, supported by narrower pages that go deeper on subtopics. The supporting pages link up to the pillar and across to each other, which concentrates the topical authority of the whole section.
No tool can guarantee rankings. What clustering does is prevent structural mistakes — pages competing with each other, topics with no depth, links that don't express your architecture — that hold otherwise good content back.
Yes, and you should where your knowledge says otherwise. The agent shows its reasoning precisely so you can merge, split or reassign groups before committing to a content plan.
The agent is built for full exports — hundreds of terms in one paste. Very large lists are processed as a whole so clustering logic stays consistent across the entire set, which is where chat-based approaches fall apart.
The map is a planning artifact: it tells you which of your planned pages should link to which, so links are written into new content as it's drafted. It's scoped to the architecture you're building — for auditing and fixing links across pages already live on your site, that's the internal linking agent's job.
Yes. The page plan can feed directly into drafting — brief the pillar page, then its supporting articles, and publish the section to your domain. The planning and the building happen in one place.
Both. For an existing site, cluster the terms you already rank for alongside your targets — overlaps reveal cannibalization to fix, and gaps reveal the supporting pages your pillars are missing.
Dual7 Agents
Paste a raw keyword list and get back a site structure: clusters, pillar pages and internal links. Start with your own input — the output is an editable project you own.
500+ users in production · Kanan.co moved off Salesforce onto Dual7 in four months · 100% code owned
Dual7 · a product of AlgorithmShift · Build once. Own forever.