The 10 Best Topic Cluster Tools in 2026

Key Takeaways

• The pillar and cluster model is sold as keyword grouping. What decides whether it works is the internal link graph and whether each page has a job no other page has.
• The expensive mistake is not a missing cluster page. It is two pages that should have been one, and every keyword-volume tool here will happily recommend both.
• Semrush ranks first because it holds demand data, a site crawl and rank tracking in one place, which is the minimum for spotting cannibalisation before it is a year old.
• Ahrefs has the better internal link opportunities report and the better backlink index. Semrush has the better tracking habit. That is the whole choice between them.
• Screaming Frog costs roughly £199 a year and shows orphan pages and inlink counts more clearly than platforms costing ten times more.
• Storyflow has no keyword volume, no keyword difficulty, no SERP analysis, no crawler, no rank tracking, no content scoring and no Search Console integration. It ranks third for arranging the cluster, not sourcing it.
• Topical authority is a working planning model, not a confirmed ranking factor. Google documents canonical selection and recommends consolidating similar pages. It has never documented a topical authority score.

A Topic Cluster Is an Information Architecture Problem in a Keyword Costume

The standard pitch is familiar. Pick a broad head term, write a long pillar page, add fifteen to thirty supporting pages on long tail variations, link them all to the pillar and back, and repeat until you have topical authority. That describes the output. It says nothing about the decision that determines whether the cluster works, which is not which keywords to group but which pages should exist at all.

A cluster is a graph. Nodes are pages, edges are internal links. Every property people attribute to a good cluster (the pillar accumulating relevance, the supporting pages being discoverable, the site reading as coherent on a subject) belongs to that graph, not to the spreadsheet the keywords were grouped in. So the useful question about any cluster tool is whether it can see the graph you already have.

The Distinct Job Test is what we run before any cluster page gets commissioned, and it is the spine of this ranking. It needs three inputs, and almost no tool supplies all three.

One: what pages already exist on this subject? Not what keywords exist. What URLs exist, including the four you forgot about from 2023.

Two: how do they link to each other? Which page has no inbound internal links, and which is linked from ninety others because it sits in a footer.

Three: what job does each page claim? One sentence describing what a reader gets here that they cannot get on any other page you own. If two pages need the same sentence, you do not have two pages. You have one page and a problem.

Question three has no tool. Questions one and two are the ranking axis, because a tool that cannot see your site can only ever tell you to add.

Cannibalisation is the failure the cluster model is supposed to prevent, and the one it most reliably causes. You commission fifteen supporting pages off a keyword grouping. Eight are genuinely distinct. Four overlap with each other. Three overlap with the pillar. Google now has to pick a canonical among near-identical pages, and the one it picks may not be the one you would have chosen. Google documents that behaviour in its guidance on duplicate content, and its advice is to consolidate.

The expensive mistake is not a missing cluster page. It is two pages that should have been one. A missing page costs you the traffic it would have earned. A duplicate page costs you an existing page's clarity, a redirect decision you will put off for a year, editing time nobody should have spent, and a link graph where the pillar's internal authority splits across three near-identical children.

We run over a thousand pages on this blog. Every painful piece of remedial work in the last two years has been a merge, never an addition.

The ten tools below are ranked by how much of the Distinct Job Test they can run.

At a Glance: The 10 Tools Compared

How I Ranked These

I come from documentary, where the same failure has a different name. You shoot two sequences covering the same beat, you cannot bear to cut either, and the film sags because neither earns its place. Over the last two years I have planned and pruned clusters across a blog of more than a thousand pages, in every tool listed here.

Five criteria, in order.

1. Can it show you the pages you already have? The test: ask it to list every URL on your site that no other page links to.

2. Can it show two of your own pages competing for one query? The test: pick a query, ask for every URL of yours that has ranked for it in the last ninety days. If the answer is only ever one URL, the tool is guessing.

3. Does it make each page's job explicit? The test: is there a place where a human sentence describing the page's unique job lives next to the keyword. In most tools there is not, and the keyword becomes the job by default.

4. Can you see the cluster as a shape? The test: look at a pillar and its children at once and identify which child has no link back and which two are the same page in different words.

5. What does it cost at the scale you operate? A forty page site and an agency with eleven clients are not solving the same problem, and the price gap here is roughly forty to one.

Pricing is as of August 2026 and changes frequently. Verify with each vendor.

Quick Picks by Job Type

• Best overall: Semrush. Demand, crawl and rank tracking in one account, exactly what cannibalisation detection requires.
• Best internal link report: Ahrefs. The internal link opportunities view is the single most useful cluster screen here.
• Best for arranging the cluster: Storyflow. Seeing the shape and questioning it before anything is commissioned.
• Best link graph on a budget: Screaming Frog. Orphan pages and inlink counts for the price of one month of a platform.
• Best grouping method: Keyword Insights. Clusters by live SERP overlap rather than string similarity.
• Best content inventory: MarketMuse. Tells you what you already published before you commission more.
• Best cheap brief tool: Frase. Most of the SERP research job at optimiser prices divided by ten.

1. Semrush

The verdict. The only tool here that answers questions one and two on its own, which is why it ranks first despite being the most expensive mainstream option.

Best for. Teams who need keyword demand, a crawl of their own site and rank tracking without stitching three subscriptions together.

Pricing. Pro from roughly $140 per month, Guru from roughly $250, Business from roughly $500, as of August 2026. Extra users cost extra on every tier.

Why it ranks here. Cannibalisation is a longitudinal problem. You cannot see it in a snapshot, because the symptom is a swap: page A ranks in March, page B in May, page A returns in July, and the position never settles. Detecting that requires rank tracking with history, which requires you to have been tracking the query already. Semrush is the tool most likely to have been, because Position Tracking sits in the same account as everything else you use.

Site Audit is the second reason. It crawls your site and reports internal link distribution, pages with a single inbound internal link, orphan pages when you connect Analytics or a sitemap, and crawl depth. That is the first two questions of the Distinct Job Test in one product. No optimiser on this list answers either.

The Keyword Strategy Builder handles the grouping half, producing pillar and subpage structures from a seed term. I trust it less than Keyword Insights, because its clusters lean on semantic similarity rather than on which URLs share a SERP. Its advantage is that the groups arrive attached to a system that also knows what you published last year. It will not connect the halves for you: noticing that suggested subtopic seven is a page you shipped in 2024 under a different title is still your job.

Strengths.

• Demand data, site crawl and rank tracking in one subscription.
• Position Tracking history makes the ranking swap visible rather than theoretical.
• Site Audit surfaces orphan pages and pages with one inbound internal link.
• Keyword Strategy Builder produces pillar and subpage structures directly.

Limitations.

• The most expensive mainstream option, with per-user costs on top.
• Keyword grouping leans semantic, so it groups terms that do not share a SERP.
• Nothing checks a proposed cluster against your published pages automatically.

The trade off. You pay platform prices for one thing you cannot get elsewhere: demand and your own site in the same account, over time.

2. Ahrefs

The verdict. Better internal link reporting than Semrush and the better backlink index. It ranks second only because cannibalisation is a tracking problem and Semrush's tracking is more likely to be running.

Best for. Finding the internal links you should have built and the pages nothing points at.

Pricing. Lite from roughly $129 per month, Standard from roughly $249, Advanced from roughly $449, as of August 2026. Credit limits apply per tier.

Why it ranks here. Ahrefs Site Audit contains the most useful cluster screen in any product on this page: internal link opportunities. It reads pages you already rank for, finds unlinked mentions of those terms elsewhere on your site, and hands you the specific paragraph where a link belongs. That is the cluster's edges, generated from your own content, with no keyword research involved.

It also does the orphan page job properly. Connect a sitemap or Search Console and it lists URLs receiving no internal links. On our own site the first honest orphan audit turned up pages that had been live for two years and had never been linked from anything. Site Explorer's Organic Keywords report shows the overlap symptom: filter your domain to a query and count how many of your URLs have held a position for it.

Where it falls behind Semrush is the tracking half by default. Rank Tracker is good and is a separate setup step that teams skip, so the historical record you need in eight months does not exist. That is a workflow gap rather than a capability gap, and it decides the order of the top two.

Strengths.

• Internal link opportunities is the best single cluster feature in any tool here.
• Orphan page detection is clear and needs no separate crawler.
• The strongest backlink index, which matters when deciding which page in an overlapping pair to keep.

Limitations.

• Credit limits on lower tiers bite quickly on large sites.
• Rank tracking is a deliberate setup step, so the history you need later is often missing.
• No content scoring or editor, so writers work elsewhere.
• Keyword grouping features are thinner than Semrush's.

The trade off. Ahrefs shows you the graph better. Semrush is more likely to have recorded the evidence when the overlap finally shows up.

3. Storyflow

The verdict. The place to arrange a cluster and see its shape before anything is commissioned. It sources none of the data and this post is not going to pretend otherwise.

Best for. Laying out a pillar and its children spatially, writing the one-sentence job for each, and catching the two that are the same page before either gets written.

Pricing. Paid only during early access. Plus is $7.99 per month billed annually or $9.99 monthly, adding 200 plus Story blueprints and unlimited file uploads. Pro is $14 per month billed annually or $19 monthly, adding AI image generation, roughly twenty times more AI usage and memory across conversations. Max is $39 per month billed annually or $49 monthly, adding forty times more AI and Team Workspace with permissions and roles. Pricing is flat per account rather than per seat, and anyone a paid member invites joins free. The Free plan launches before the end of 2026.

Why it ranks here. The Distinct Job Test has three questions. Storyflow answers exactly one, the third, and it is the one no SEO platform touches.

A cluster laid out on a canvas is a cluster you can read. The pillar sits in the middle, the children sit around it, and each child carries its one-sentence job on the card. Two cards with the same sentence are visible from across the room in a way two rows in a spreadsheet are not, because rows are read one at a time and a canvas is read all at once. That is the entire claim, and it is narrow.

The second use is arrangement. A pillar with twenty-two children is usually a pillar with four sub-pillars and a naming problem, and you find that out by moving cards into groups until the groups stop feeling arbitrary.

The AI adds one thing for this job. It reads your full active canvas board, plus up to one Tactic and up to three Documents you @-mention, so you can put twenty proposed cluster pages on a board and ask which two are the same page. It answers from what is on the board, and it has no idea what is on your website, what anything ranks for, or how many people search for it.

Which is the honest limit. The expensive mistake is not a missing cluster page. It is two pages that should have been one, and a canvas is unusually good at making that pair visible while both are still proposals. It is useless at telling you the pair already exists as live URLs. That is why Storyflow ranks third rather than first in a category where the data belongs to somebody else.

Strengths.

• The whole cluster is visible at once, so two overlapping proposals are visible at once.
• The one-sentence job lives on the card next to the keyword, which is where the Distinct Job Test gets run.
• Rearranging a hierarchy is dragging, so restructuring a twenty-page cluster takes minutes.
• Whole-board AI answers questions about the proposed cluster, including which entries duplicate each other.
• Invited collaborators join free and pricing is flat per account, so a freelance SEO and two writers cost nothing extra.

Limitations.

• No keyword volume and no keyword difficulty data of any kind. You bring the terms.
• No SERP analysis. It cannot tell you who ranks for a query or what those pages contain.
• No crawler. It never sees your website, so it cannot find an orphan page, an existing near-duplicate, or a broken internal link.
• No rank tracking, so the ranking swap that reveals cannibalisation is invisible here.
• No content scoring and no optimisation grading against a SERP.
• No Search Console or analytics integration, and no export of a cluster map into a CMS, a sitemap or an SEO platform.

The trade off. Its ground is arranging the cluster and seeing the shape, not sourcing the terms. Pair it with Semrush or Ahrefs or you are planning from no evidence.

4. Screaming Frog SEO Spider

The verdict. The cheapest honest look at the link graph you already have, and on that single criterion it beats every platform here.

Best for. Crawling your site and getting exact inlink counts, orphan pages and crawl depth per URL.

Pricing. Free for up to 500 URLs. Licence roughly £199 per year, around $250, as of August 2026. One licence, no per-seat escalation.

Why it ranks here. Criterion one is whether a tool can list every URL that nothing links to. Screaming Frog does this better than Semrush and better than Ahrefs, because it is a real crawler running on your machine with no credit limits and its Inlinks column is exact rather than sampled.

The reports that matter are unglamorous. Inlinks per URL tells you which cluster children are structurally invisible. Crawl Depth tells you how many clicks from the homepage each page sits, and a child at depth six is a child nobody reaches. Connect the Search Console and Analytics APIs and the Orphan URLs report lists pages receiving impressions but linked from nowhere in the crawl. The force-directed crawl diagram helps too, because a healthy cluster looks like a hub with spokes and an unhealthy one looks like a chain.

It ranks fourth because it answers one of the three questions completely and the other two not at all. No demand data, no SERP data, no notion of what a page is for. It sees structure, exactly and cheaply, and structure is half the job.

Strengths.

• Exact inlink counts and crawl depth per URL, with no sampling or credit limits.
• Orphan URL detection via the Search Console and Analytics APIs.
• One flat annual licence, cheaper than two months of any platform here.

Limitations.

• Desktop software, so large crawls eat memory and it is not a shared dashboard.
• No keyword demand data, no SERP analysis, no content guidance.
• Free tier stops at 500 URLs, under the size where clusters get messy.

The trade off. Buy the licence, run it quarterly, and you will find cluster pages nothing has linked to since the day they shipped.

5. Keyword Insights

The verdict. The most defensible grouping method on this page, because it clusters by live SERP overlap rather than by how similar the strings look.

Best for. Turning a raw export of two thousand keywords into groups that reflect what Google treats as one query.

Pricing. Credit-based, from roughly $58 per month at entry level to roughly $299 for agency volume, as of August 2026. You pay per keyword processed rather than per seat.

Why it ranks here. Every clustering tool decides when two keywords belong on one page, and there are two ways to do it. Semantic similarity asks whether the phrases mean the same thing. SERP overlap asks whether the same URLs rank for both, and if Google returns eight of the same ten results for two queries, Google has already told you they are one page.

That distinction is the direct answer to the cannibalisation problem. Most duplicate cluster pages get commissioned because two keywords looked different in a spreadsheet. SERP overlap clustering catches a large share of those pairs at the grouping stage, which is the cheapest moment to catch them.

It ranks fifth because it only sees the SERP. With no crawler and no knowledge of your site, it will group beautifully and still hand you a cluster where six of the fifteen groups are pages you published last year. Run it, then check the output against a crawl, in that order.

Strengths.

• SERP overlap clustering is the most defensible grouping method available.
• Intent classification per cluster is usable rather than decorative.
• Credit pricing scales to occasional heavy use with no annual platform commitment.

Limitations.

• No crawl and no knowledge of your existing pages, so its groups can duplicate what you published.
• Credits run out fast on large sites and the cost becomes unpredictable.
• Output needs a second home. It is not where the plan lives.

The trade off. The best answer to which keywords are one page, and no answer at all to whether that page already exists.

6. MarketMuse

The verdict. The one optimiser that starts from your existing content inventory rather than a blank keyword, which is the right instinct for clusters.

Best for. Seeing what you already published on a topic before commissioning more of it.

Pricing. Free tier with limited queries. Paid plans from roughly $99 per month into several hundred for team tiers, as of August 2026.

Why it ranks here. MarketMuse's Topic Inventory reads your existing pages and maps them against topics, so its first answer to "should we build a cluster on X" is a list of what you already have on X. Every other optimiser here starts from the SERP and assumes your site is empty. Personalized Difficulty follows from that, estimating difficulty relative to what your site already covers. Treat that number as a planning heuristic rather than a measurement of anything Google publishes.

The feature closest to cannibalisation detection is its flagging of multiple pages competing on one topic inside your inventory. It is topic-level rather than query-level, so it is coarser than filtering Ahrefs by query, and it runs without you knowing which query to check, which helps when auditing a cluster you inherited. It ranks sixth because pricing has moved around repeatedly, inventory analysis is slow on large sites, and the writing guidance is weaker than Clearscope's.

Strengths.

• Content inventory analysis starts from your pages rather than a blank SERP.
• Flags multiple pages competing on one topic without you specifying the query.
• Free tier is enough to evaluate whether the inventory view helps you.

Limitations.

• Pricing and packaging have changed repeatedly, so budget with caution.
• Inventory analysis is slow on sites over a few thousand pages.
• Topic-level flagging is coarse next to query-level rank data.

The trade off. Buy it for the inventory, not for the editor.

7. Surfer SEO

The verdict. A strong on-page optimiser with a topical map generator attached, and the map is built from the SERP rather than from your site.

Best for. Getting a single page's on-page coverage right once you have decided the page should exist.

Pricing. Essential from roughly $99 per month, Scale from roughly $219, as of August 2026.

Why it ranks here. Surfer's Content Editor is genuinely good at its job. It reads the pages currently ranking, extracts the terms and structural patterns they share, and scores your draft against them in real time. Writers like it because the feedback is immediate and specific.

The Topical Map feature is why it appears in a cluster list at all. Give it a seed term and it produces a pillar plus supporting page ideas, generated with no knowledge of your website, which means it will confidently propose eight pages, three of which you have already published. Surfer will never recommend a merge, because it does not know there is anything to merge.

Used correctly it sits after the cluster decision: decide the page should exist, then check it covers what the SERP expects. Used incorrectly, the content score becomes the goal, and you end up with fifteen pages that all score 82 and all say the same thing, which describes the cannibalisation problem rather than solving it.

Strengths.

• Real-time content scoring that writers will actually use.
• Terms and structure pulled from the pages currently ranking, not a static database.
• Google Docs and WordPress integrations keep writers in place.

Limitations.

• No crawl of your own site, so its maps duplicate pages you already have.
• No rank tracking and therefore no cannibalisation signal at all.
• Priced close to a full SEO platform for a narrower job.

The trade off. Excellent for the page, blind to the cluster. Never let it decide what to publish.

8. Clearscope

The verdict. The cleanest content optimiser on the market and the one with the least to say about clusters.

Best for. Editorial teams who want one clear brief per page and no dashboard sprawl.

Pricing. Essentials from roughly $189 per month, higher tiers into the several hundreds, as of August 2026.

Why it ranks here. Clearscope does one thing with unusual discipline. You give it a target term, it reads the ranking pages, and it produces a graded report of terms to cover. The reports are readable enough to hand to a freelance writer with no training, which is worth more than it sounds at volume.

For cluster work its contribution is page level only. No crawler, no rank tracking, no site inventory, no cluster feature. It cannot tell you that the brief you just generated describes a page you published in 2024. Where it helps a cluster is indirectly: two reports side by side for two proposed pages show the overlap in the term lists faster than reading two outlines. That is a manual technique rather than a feature, and it works.

It ranks eighth because the price is high for a single-page tool, and at that price the marginal thing you buy over Frase is polish rather than capability.

Strengths.

• The clearest reports of any optimiser, usable by writers with no SEO training.
• Google Docs integration keeps the grading where the writing happens.
• Comparing two reports exposes overlap between two proposed pages.

Limitations.

• No crawl, no rank tracking, no site inventory, no cluster features.
• Expensive for a page-level tool, with limited report allowances on entry tiers.
• Nothing stops you optimising two near-identical pages perfectly.

The trade off. Buy it for writers, not for planners.

9. Frase

The verdict. Most of the SERP research job at a tenth of the optimiser prices, which for cluster work is the right trade.

Best for. Cheap, fast briefs across a lot of cluster pages.

Pricing. Solo from roughly $15 per month, Basic from roughly $45, Team from roughly $115, as of August 2026.

Why it ranks here. Frase pulls the top ranking pages for a query, extracts their headings and questions, and assembles a brief you can edit. Doing that by hand takes forty minutes per page. Frase takes about four.

Across a cluster of eighteen pages, that is the difference between briefing all of them properly and briefing four and improvising the rest. Brief quality matters more than any single page's optimisation score, because the brief is where the Distinct Job Test gets enforced in practice: if you cannot write the brief without repeating another brief, you have found your duplicate. Question extraction is the underrated piece, and seeing the actual questions people ask around a subtopic often tells you a proposed page is really two, or that two are really one.

It ranks ninth because the scoring is weaker than Surfer's and Clearscope's, the AI drafting needs heavy editing, and it knows nothing about your site.

Strengths.

• Cheapest credible SERP research here by a wide margin.
• Question extraction reveals whether a subtopic is one page or two.
• Fast enough to brief an entire cluster in an afternoon.

Limitations.

• Optimisation scoring is less reliable than Surfer's or Clearscope's.
• AI drafting output needs substantial editing.
• No crawl, no rank tracking, no awareness of your existing pages.
• Interface is busier than the job requires.

The trade off. Buy it to brief the whole cluster rather than buying a premium optimiser to perfect three pages of it.

10. Airtable or Google Sheets

The verdict. The unglamorous register where the job sentence lives, and the only place here that gives question three a permanent home.

Best for. One row per cluster page, with the URL, the job sentence, the primary query and the current status.

Pricing. Sheets is free with a Google account. Airtable has a free tier with paid plans from roughly $20 per seat per month as of August 2026.

Why it ranks here. Every tool above either sources data or arranges it, and none keeps the record. Six months after a cluster ships, the question is which page owns which query and what each one was for, and that answer lives in a table somebody maintained or it does not exist.

The structure that works is five columns: URL, primary query, the one-sentence job, the parent pillar, and last reviewed date. The job column is load-bearing. Sort by it, read down, and duplicate sentences sit next to each other. That catches pairs no keyword tool flags, because two pages can have entirely different primary queries and the same job.

It ranks last because it is entirely manual. It knows nothing, detects nothing, and decays the moment nobody owns it. It earns its place because the one thing no other tool does is the thing this does for free.

Strengths.

• The job sentence gets a permanent, sortable home, which is where duplicates surface.
• Linked records in Airtable model the pillar-to-child relationship directly.
• Free or near free, with no per-page limits.

Limitations.

• Entirely manual. It detects nothing and warns about nothing.
• Goes stale within a quarter without a named owner.
• Rows cannot show the shape of a cluster.

The trade off. Ten minutes a week buys you the only record of what each page was for.

What to Actually Pay For

Pay for one platform, not two. Semrush or Ahrefs, not both. The overlap is roughly ninety percent of what you will use, and the saving covers Screaming Frog, Frase and a canvas. Choose Semrush if rank tracking discipline is weak on your team, Ahrefs if internal linking is your known weak point.

Pay the £199 for Screaming Frog above 500 pages. It is the cheapest item here that answers a question the expensive items answer worse. Run it quarterly and diff the orphan list.

Do not pay for a premium optimiser until briefs are consistent. Clearscope improves pages that were correctly commissioned. It does nothing for a cluster where three pages should have been one.

Pay for SERP overlap clustering above a thousand keywords. Below that, group by hand against the actual search results and you will learn more about the topic than any export teaches you.

Tools to Avoid for This Job

Any AI tool that generates a cluster from a seed keyword with no access to your site. It produces a tidy list of fifteen page titles, and on an established site a quarter of them describe pages you already have. The output looks like a plan and is a proposal to duplicate yourself.

Keyword volume as the sole criterion for whether a page exists. Volume tells you how many people ask. It does not tell you whether the answer belongs on a page you already own.

A content score as the definition of done. Optimising fifteen cluster pages toward the same term list makes them more similar to each other, which is the opposite of what a cluster needs.

An internal linking plugin that auto-links on keyword match. It builds hundreds of edges nobody chose, points them at whichever page matches a string, and makes the link graph unreadable at the moment you need to read it.

The pillar page nobody will maintain. A 6,000 word pillar commits you to keeping 6,000 words current through every change to the cluster, and most teams write it once. It becomes the stalest page in the cluster and the one everything links to.

What No Tool on This List Does

None of them writes the job sentence. Question three is a human judgment about what a reader gets here that they cannot get elsewhere on your site. Tools can surface the overlap. Deciding which of two overlapping pages deserves to live is editorial.

None of them tells you when to merge rather than differentiate. Two partially overlapping pages have two valid futures: make them properly distinct, or combine them and redirect. That depends on backlinks, on how the two actually rank, and on whether the distinct angle is real or invented after the fact.

None of them, including Storyflow, verifies that a proposed cluster does not already exist on your site. Storyflow has no crawler, no Search Console connection and no rank data, so it will let you plan a twenty page cluster while four of those pages are already live. It arranges what you put on the board and never checks the board against your website, which is a real gap at exactly the moment duplication is cheapest to prevent.

None of them proves topical authority exists as a ranking mechanism. It is a working model that produces good planning behaviour, because it forces you to think about coverage and internal linking. Google documents canonical selection among similar pages and recommends consolidating them. It has never documented a topical authority score.

The Bottom Line

The pillar and cluster model is taught as keyword grouping and practised as information architecture, and the gap between those two things is where the money goes. What decides the outcome is whether each page has a job nothing else on your site has, and whether the links between them describe a hub rather than a chain.

Buy one platform. Semrush if you want demand data, a crawl and rank tracking bundled where your team will trip over them. Ahrefs if internal linking is your known weakness, because internal link opportunities is the best single cluster feature anyone ships. Add Screaming Frog for £199 a year and run it quarterly.

Then, before you commission anything, put the proposed cluster somewhere you can see all of it at once and write one sentence per page. That is the narrow ground Storyflow ranks third on, and it is a planning surface rather than a data source. The expensive mistake is not a missing cluster page. It is two pages that should have been one, and the only reliable moment to catch that pair is while both are still just a sentence.