SEO Content Strategy: A UK Business Guide to Building Topic Authority Instead of Chasing Keywords in 2026

An SEO content strategy built one keyword at a time eventually starts working against itself. The mechanism is unglamorous and almost universal: a business identifies twenty promising search terms, commissions twenty articles, and publishes them over eighteen months. Six of those terms mean substantially the same thing to a search engine. Now six pages compete for one query, none of them accumulates the signals that would have made one of them strong, and the page that ranks is whichever one the algorithm picked — frequently not the one with the commercial value.

This guide starts with that problem because it is the one most UK businesses actually have, and because it is invisible from the reports most of them receive. It covers how keyword cannibalisation develops, how to diagnose it in Search Console in about twenty minutes, and the hierarchy of fixes from consolidation through to de-indexing. It then covers the structural remedy — organising content into pillars and clusters rather than a flat pile of posts, with the pillar model covered in more depth in our guide to creating topical authority — how to map search intent by reading what already ranks rather than guessing, what to measure instead of per-keyword positions, and an honest account of which parts of this are documented mechanism and which are practitioner inference.

How cannibalisation actually happens

Nobody sets out to publish six pages about the same thing. Cannibalisation is an emergent property of a particular and very common workflow, and recognising the workflow is most of the cure.

It starts with a keyword research export. A tool returns several hundred terms with volumes attached, somebody filters for reasonable volume and achievable difficulty, and the survivors become a content calendar. The problem is that the export lists terms, not intents, and several terms frequently express one intent. “Managed IT support cost”, “how much does managed IT cost”, “managed IT pricing UK” and “IT support cost per user” are four rows in a spreadsheet and one question in a user’s head. Treated as four briefs, they produce four articles that each answer the same question slightly differently.

The second contributor is time. Cannibalisation rarely arrives all at once; it accumulates. A page published in 2023 ranks acceptably. In 2025 somebody writes a fresher piece on a near-identical topic without checking what exists, because the person who wrote the first one has left and nobody audits the archive before commissioning. The new page splits the signals with the old one and both settle lower than the original sat alone.

The third is well-intentioned refreshing without consolidating. An organisation decides its content needs updating, writes new and better versions, and publishes them alongside the originals rather than replacing them. It now has two good pages competing instead of one strong page, and the older URL frequently retains the links and history that would have made the new content rank.

What it looks like from the outside

The symptoms are specific enough to recognise. Several of your URLs appear for the same query in Search Console, each with a modest impression share. Positions oscillate week to week for a term that should be stable, because the algorithm keeps reselecting which of your pages to show. A page that ranked reliably drops shortly after you publish something similar. And the page ranking for your most commercially valuable term is a blog post rather than the service page you wanted, which is cannibalisation in its most expensive form.

What it does not look like is a penalty or a technical fault. Nothing is broken, no warning appears, and a technical audit will not flag it, because every page is individually fine. That is why it persists for years in organisations that are otherwise doing sensible things, and why it is worth looking for deliberately rather than waiting for it to announce itself. Technical health is a separate axis, covered in our guide to the technical SEO audit.

Content strategy in UK businesses — the numbers

The figures below reflect what we find across UK organisations of 20 to 500 staff with an established blog or resource section — typically thirty or more published articles and at least two years of publishing history.

The first two figures together describe the scale of self-inflicted competition. Roughly two-thirds of sites have it somewhere that matters, and where it exists there are typically three pages involved rather than two — which means the signals that could have made one page competitive are divided three ways. Consolidation is frequently the highest-return SEO work available to an established site, precisely because the content already exists and has already been paid for.

The third figure is the one that should reframe content budgets. Almost a third of articles on a mature UK business site receive essentially no search traffic. Some of those are legitimately non-search assets — announcements, event write-ups, material written for customers rather than searchers. Many are not: they are keyword-led pieces that were published, never linked to from anywhere meaningful, and never revisited. Before commissioning new content, it is worth establishing what proportion of the existing library is in this state, because the answer changes whether the right investment is production or repair.

The fourth figure is the expectation to set with whoever approves the budget. Content restructuring does not produce movement in six weeks. Consolidation can act faster than new publishing — occasionally within a month, because the pages already have history — but the full effect of a restructured programme takes a couple of quarters to be legible. Anyone promising otherwise is describing something other than organic search, and we cover realistic timelines in our guide to how long SEO takes to show results.

Where content strategies fall down

The grid below groups the recurring failures we find when reviewing an established UK content programme. The badges reflect how much each issue suppresses the performance of content that has already been paid for, rather than how difficult it is to correct.

The first row of the second card is where the recoverable value sits. Competing pages left in place are an ongoing tax on content you have already bought, and consolidation is usually a day of work per topic rather than a project. It is also the intervention most likely to show movement quickly, because you are concentrating existing signals rather than asking a search engine to evaluate something new.

The orphan pages row is worth separating out because it is so easily fixed. A page with no internal links pointing to it is a page you have told search engines nothing about, and in a cluster model internal links are the mechanism that communicates which pages belong together. Orphans are common on sites where content is published through a blog template that generates a chronological listing and nothing else — the page exists, it is technically reachable, and nothing in the site structure says what it relates to.

On the third card, reporting on per-keyword positions is the practice that hides everything else in this grid. A ranking report shows position by term and cannot show that three of your pages are splitting one query, that a third of your library gets no impressions, or that your coverage of a topic has gaps. Organisations reporting only positions are structurally unable to see the problems that most affect them, which is why measurement appears in the checklist later rather than as an afterthought. The alternatives are covered in our guide to SEO reporting metrics that matter.

What an audit of an established content library finds

The chart below shows how often each finding appears when we audit a UK business content library of thirty or more articles. The figures are not mutually exclusive — most sites have several — and together they describe why adding more content to an unaudited library often produces less than expected.

The fourth row is the one that wastes the most production budget, because a mismatched page cannot be fixed by writing it better. If the results for a query are all comparison tables and the business published a thousand-word narrative essay, the essay is not competing on quality — it is competing in the wrong category. Format follows intent, intent is visible in the results, and checking takes thirty seconds per query before a brief is written.

The fifth row is worth understanding as a symptom rather than a problem. A blog post outranking a service page for a commercial term usually means the blog post answers the query better than the service page does, which is information about the service page. The fix is rarely to suppress the article; it is to improve the service page so it earns the position, and to have the article link to it clearly with descriptive anchor text so the commercial intent is served either way.

The duplicate row at twenty-two per cent is mostly migration residue: a site move that left both old and new URLs live, a category and tag archive generating thin variants, or a print version indexed alongside the main page. These are the cheapest findings to resolve and the ones most likely to have been present for years, because nobody looks at a site’s indexed page count until something goes wrong.

The fix hierarchy for competing pages

Once the cannibalisation register exists, each entry needs one of four treatments. Working through them in order of preference avoids the common error of differentiating pages that should simply have been merged.

Consolidation is the default and it is underused, largely because deleting content feels like destroying an asset. It is not: the content moves into the surviving page and the redirect carries the accumulated history with it. The instinct to differentiate instead is understandable and usually wrong, because genuine differentiation requires the two pages to answer materially different questions, and if they did they would not have been competing in the first place.

Two practical cautions. Choose the surviving URL on evidence rather than preference — the page with existing links, history and impressions is usually the right host even if its content is weaker, because content is easy to improve and history is not. And implement the redirects: a consolidation where the old URLs are deleted without redirects discards exactly the signals the exercise was meant to concentrate, which converts a good decision into a net loss.

For the retire category, be honest about what is actually off-strategy. A page with no impressions is not automatically worthless — it may serve customers, support a sales conversation, or exist for a legitimate non-search reason. The test is whether anyone would miss it, not whether search engines currently send it traffic.

Search intent mapping by reading the results

Intent mapping sounds like a theoretical exercise and is in practice the most concrete step in the whole process, because you do not have to infer intent. The search results already encode a decision about it, and reading them is faster and more reliable than any framework.

Classify, then check

Start with the conventional categories: informational (someone wants to understand something), commercial investigation (someone is comparing options before buying), transactional (someone is ready to act), and navigational (someone wants a specific organisation). Assign each target query a provisional category. Then search it, on google.co.uk, and look at what occupies the first page.

If the results are guides and explainers, the query has been judged informational and a service page will not rank for it. If they are comparison tables and listicles, the expected format is a comparison. If they are product or service pages, a blog post is the wrong asset. If the page is dominated by local results, the query carries local intent regardless of how it reads. This takes half a minute per query and overrides your provisional classification whenever the two disagree, because the results reflect an assessment made from far more behavioural data than you have.

Map intent to asset type

Once intent is established, the asset follows. Informational queries want cluster content: guides, explainers, how-tos, comparisons of approaches rather than vendors. Commercial investigation wants comparison content and case material, and is where content most directly supports revenue. Transactional queries want service or product pages, optimised as such, with the blog content linking into them. Navigational queries want your own pages to be findable and require little content work.

The commercial investigation band is the one UK businesses most often under-serve. It is where a buyer is actively choosing and where the content that helps them choose is disproportionately valuable, yet content plans tend to cluster at the informational end because that is where the search volume is. Volume and value diverge sharply here, and planning by volume systematically over-invests in the top of the funnel.

A note on UK specifics

Two things matter for UK planning. Volumes are considerably smaller than US equivalents, so a term with a few hundred monthly searches may be a genuinely strong commercial target rather than something to dismiss — and tools that report global or US-weighted figures will mislead you. And terminology diverges: the words UK buyers use are frequently not the words US-authored competitor content uses, which creates openings for anyone researching against UK data rather than importing an American keyword set.

What a structured content programme costs in the UK

The figures below are indicative UK costs for 2026, excluding VAT, for a business-to-business organisation targeting a handful of commercial topics. They assume genuinely researched long-form content rather than volume production, because the economics of the two are not comparable and only one of them tends to survive contact with a competitive result page.

The ordering of that table is deliberate, because it is also the recommended spending order. An audit costing a couple of thousand pounds routinely identifies enough consolidation opportunity to lift existing pages that collectively cost tens of thousands to produce. Spending on new production before auditing is the equivalent of adding stock to a warehouse nobody has inventoried, and it is what most content budgets do.

The per-article range deserves comment because it is where buyers are most often misled in both directions. Content at the bottom of the range can be perfectly good for straightforward informational topics with a competent writer who knows the subject. Content substantially below the range — and it is widely available — is generally assembled rather than researched, and its failure mode is not being obviously bad but being indistinguishable from everything else targeting the same query, which is a slower and more expensive way to fail. Fuller pricing context sits in our guide to SEO pricing in the UK.

Note what is absent from the table: link acquisition. That is a separate discipline with separate economics, and bundling it into a content budget obscures both. Content structure and topical coverage are what this guide addresses; they are necessary and, in competitive commercial categories, not always sufficient.

Restructuring a content programme — six months

The sequence below assumes an established library of thirty or more articles with no topic structure. A greenfield programme skips the audit and consolidation phases and starts at month two, which is both faster and rarer.

The sequencing decision that matters most is consolidating before producing. It is counter-intuitive because consolidation looks like housekeeping and production looks like progress, and organisations under pressure to show activity reliably invert it. Inverting it means new content joins an environment where signals are already divided, and frequently adds another competitor to an existing conflict.

The second is pillars before clusters. A cluster page exists partly to demonstrate that a topic is covered in depth, and that demonstration requires something for it to point at. Publishing twelve cluster articles and adding the pillar later works eventually, but it wastes several months of the structure not being legible.

Keyword-led against topic-led planning

The comparison below highlights the topic-led column, which in this case is a straightforward recommendation rather than a qualified one: for organisations publishing more than a handful of pieces a year, keyword-led planning creates the problems the rest of this guide is about. The exception is genuinely small programmes — if you will publish six articles in total, structural overhead may exceed its benefit.

The reporting row is the one that changes behaviour rather than merely describing it. Position-per-keyword reporting cannot see cannibalisation, coverage gaps or a dead archive, so an organisation using it will keep commissioning articles and keep being puzzled by flat results. Switching to cluster-level impressions and a count of pages competing per query makes the actual problems visible, at which point the planning tends to correct itself.

The honest entry is the topic-led failure mode. It is slower to show results, it requires somebody to own the map, and it is less satisfying to report on month by month because the unit of progress is coverage rather than output. Programmes that adopt it and then lose the owner drift back to keyword-led commissioning within about a year, usually without anyone deciding to.

Content structure maturity — where most UK sites sit

Combining the assessment areas gives an indication of how well an organisation’s existing content is organised to compete, independent of how much of it there is or how well written it is. The gauge reflects a first review of a UK business with an established blog and no topic structure.

A score around thirty-two usually has nothing to do with writing quality. Production frequently scores well — the articles are competent, on-brand and professionally made. What scores badly is everything structural: no pillars, internal linking that happens when a writer remembers, no intent mapping, no register of competing pages, and a third of the library receiving no impressions because nothing points at it and nothing signals what it belongs to.

The distinctive feature of this particular gap is that the expensive part has already been paid for. Unlike most maturity gaps, closing this one is mostly reorganisation of existing assets rather than new investment: an audit, a map, a consolidation pass and a linking pass. Organisations that do those four things typically find that content they had written off starts performing, which is a better return than the same money spent on additional production.

The caveat worth stating: a low score matters in proportion to how much content exists. A business with eight articles does not have a cannibalisation problem or a structure problem — it has a coverage problem, and the answer is to publish more with a sensible plan rather than to audit what little exists. The structural discipline earns its overhead somewhere around twenty to thirty published pieces, which is roughly where conflicts start appearing without anyone noticing.

What to measure instead of keyword positions

Per-keyword position reporting is the default because it is easy to produce and looks like accountability. It is also structurally incapable of showing the problems that most affect an established content programme, which is why changing the measurement usually changes the strategy without anyone arguing about strategy.

Pages competing per query

The cannibalisation metric, and the one nobody reports. For each priority query, how many of your URLs receive meaningful impressions. The target is one. Tracking this quarterly turns cannibalisation from an invisible condition into a number that goes down when you work on it, which is what makes the work fundable.

Cluster-level impressions and coverage

Group Search Console data by topic cluster rather than by page, and track total impressions and the number of distinct queries the cluster appears for. This is the closest practical proxy for whether your coverage of a topic is being recognised, and it moves earlier and more legibly than positions on head terms.

Queries per page and pages with zero impressions

A page appearing for many related queries is a page that has been understood as covering a topic rather than matching a phrase. Conversely, the count of pages with effectively no impressions is your dead-inventory figure — typically around a third on an unmanaged library — and reducing it is usually cheaper than producing replacements.

Commercial outcomes, attributed honestly

Organic-assisted enquiries and the pages involved, acknowledging that attribution across a long consideration period is imprecise. Informational cluster content rarely converts directly and frequently appears earlier in journeys that convert elsewhere, so judging it on last-click conversion will systematically undervalue it and lead to cutting exactly the content that supports the commercial pages.

One honest caution about all of this: organic search data is noisy, seasonal and subject to algorithm changes nobody controls. Month-to-month movement is largely meaningless and quarter-on-quarter comparison is the shortest sensible interval. An agency or internal team reporting confident causal explanations for a four per cent monthly movement is over-reading the data, and the appropriate response to most short-term fluctuation is to keep going.

Benchmarks — content practice against what we find

The figures below show how often each practice is present across UK organisations of 20 to 500 staff running an established content programme. They describe process rather than output volume.

Eighty-four per cent publish to a plan and twenty-two per cent have a topic map, which tells you what most plans consist of: a schedule of titles. A schedule is not a structure, and the gap between those two numbers is where cannibalisation is manufactured — content produced on cadence, against terms rather than topics, with no map to check a new brief against.

The sixteen per cent auditing the archive before commissioning is the most consequential row for budget efficiency. It means roughly five organisations in six commission new content without establishing what they already have, which is how the same topic gets covered twice and how a third of libraries end up receiving no impressions.

The bottom two rows are the ones that determine whether any of this persists. Reporting segmented by topic makes the structure visible to whoever funds it; a named owner means somebody checks new briefs against the map. Without both, a restructured programme reverts to keyword-led commissioning within about a year, and the audit gets commissioned again two years later.

The number that decides where to spend first

If one measurement had to direct a content budget, it would be the proportion of an existing library that receives effectively no organic impressions. It establishes whether the organisation has a production problem or a structure problem, and those have different and non-interchangeable answers.

Roughly a third of the library doing nothing is a striking figure precisely because every one of those pages was commissioned, briefed, written, reviewed and published at full cost. It is the clearest available evidence that content performance is not primarily a function of how much you produce, and it is the argument for auditing before commissioning that finance teams find most persuasive.

Some of that third is legitimately non-search content and should be excluded before drawing conclusions — announcements, event write-ups, material written for existing customers. Do the exclusion honestly rather than using it to explain the whole figure away; in most libraries the genuinely search-intended proportion of the dead inventory is still substantial.

Where the figure is high, the first money goes on audit, consolidation and internal linking, because you are recovering assets already paid for. Where it is genuinely low — under about ten per cent — the library is working and the constraint is coverage, so production is the right investment. Establishing which situation you are in takes a Search Console export and an afternoon, and it changes what the next twelve months of budget should buy.

What is documented and what is inference

This field contains a great deal of confidently stated mechanism that is not actually documented, and a guide that does not separate the two is asking you to take inference on faith. It is worth being explicit about which parts of the above rest on what.

Well established. Search engines crawl and index individual URLs, and internal links are used both to discover pages and to understand relationships between them. Duplicate and near-duplicate content causes a selection problem where the engine picks a canonical version, and you can influence that selection with canonical declarations and redirects. Redirects pass accumulated signals to the destination. None of this is contentious.

Consistent with published guidance. That content should be created primarily for people rather than to manipulate rankings, and that demonstrated experience and expertise matter in how quality is assessed. Google publishes quality rater guidelines describing what human evaluators look for, which shape how systems are trained and evaluated. That is a genuine and useful signal about direction; it is not a list of ranking factors, and anyone presenting rater concepts as directly optimisable levers has overstepped.

Practitioner inference. “Topical authority” as a quantity a site accumulates within a subject area is a model, not a documented metric. It describes something practitioners observe repeatedly — sites that cover a subject comprehensively and coherently tend to rank more easily for new queries within it — and the underlying mechanism is not published. The pillar-and-cluster structure is similarly a planning discipline that works, rather than a format search engines are known to recognise as such.

That distinction matters practically rather than academically. The reason to adopt the pillar-and-cluster model is not that a search engine rewards the shape; it is that the discipline of building it forces you to identify duplicate intents before publishing, link related content deliberately, cover topics with fewer gaps, and avoid the cannibalisation that demonstrably does cause harm. The benefits are real and largely traceable to mechanisms that are documented. You do not need the unproven part to be true for the method to be worth using — which is a considerably more robust basis than most content strategy advice rests on.

How AI search changes the picture, and how much

By 2026 a meaningful share of informational queries are answered in place — in AI overviews within search results, or in assistants that synthesise an answer and cite sources. This has two genuine implications for content strategy and rather more speculation attached to it than evidence supports.

The first implication is that clearly structured, specific, genuinely informative content is better positioned to be cited than content written to hit a word count around a keyword. Systems that synthesise answers need extractable claims: concrete figures, named specifics, direct answers to direct questions, clear headings. Content that buries a thin answer in five hundred words of preamble is poorly suited to being quoted, and the thin keyword-led page is the format most exposed.

The second is that some informational traffic that previously arrived on your page now does not, because the answer was delivered without a click. That is a real effect and it lands hardest on exactly the top-of-funnel informational content that volume-led keyword planning over-produces. It strengthens the argument made earlier for weighting content plans towards commercial investigation queries, where a buyer comparing options generally still wants to read sources.

What is not yet reliable is any account of how to make citation happen. Citation behaviour varies between systems, changes without notice, and is poorly measurable — most organisations cannot yet tell you how often they are cited or what it is worth. Anyone selling a service that guarantees or systematically engineers AI citation is describing something nobody can currently verify. The defensible position is that content which is structured, accurate, specific and attributable is better placed under any of these systems than content which is not, that this is the same content that serves human readers well, and that the measurement will improve. We look at the specifics separately in our guides to optimising for Google AI overviews and appearing in ChatGPT and Perplexity answers.

What this looks like in practice

A UK commercial insurance broker with 55 staff had been publishing consistently for four years — 118 articles, produced by an agency on a two-per-month retainer at roughly £9,600 a year. Organic traffic had been flat for eighteen months despite the library nearly doubling in that period. The agency’s monthly report showed position tracking for 40 keywords, most of them stable, several improving slightly.

The audit found what flat-despite-publishing usually means. Of the 118 articles, 39 received effectively no impressions. On the firm’s eleven most commercially valuable queries, nine had multiple competing URLs, with one query — a variation on professional indemnity cover for consultants — being contested by five separate articles published between 2022 and 2025, each answering substantially the same question. None of the five had ever reached the first page. There were no pillar pages; the site had an eight-page paginated blog listing and nothing else connecting related pieces. And 61 articles had no internal links pointing to them from anywhere except that chronological listing.

The intent problem was the one the firm found most surprising. Three articles targeted terms whose live results were entirely made up of product and service pages — transactional queries being addressed with explainer articles. Those pieces could not have ranked regardless of quality, and one of them had been refreshed twice on the assumption that it needed to be better.

The remediation ran about fourteen weeks. Consolidation merged 23 articles into 9 surviving URLs with redirects, and retired 14 more that served no purpose. Four pillar pages were built around the firm’s actual commercial lines, two of them by upgrading existing strong articles rather than writing new ones. Internal linking was added deliberately across the clusters. Three transactional targets were reassigned from articles to the relevant service pages, with the articles rewritten to serve the informational question and link through. Total remediation cost was approximately £7,400 — less than a year of the existing retainer.

At four months, the consolidated pages were the visible change: the professional indemnity query moved from nothing on page one to a single URL in the lower half of the first page, and aggregate impressions across the four clusters were up materially while total published pages had fallen by 28. Enquiry volume attributed to organic rose, though over a period short enough and a base small enough that the firm was careful not to over-claim causation.

We had been told for two years that content takes time and we needed to keep publishing. Both of those things are true, which is what made it hard to argue with. What nobody had told us was that five of our own articles were fighting each other for the query we cared about most, and that we were paying to add a sixth competitor every couple of months. Deleting content turned out to be the thing that worked.

Two points generalise. The first is that a flat trend alongside a growing library is a structural signal rather than a reason to publish harder, and the diagnosis takes an afternoon in Search Console. The second is that the position report was accurate throughout — the tracked keywords genuinely were stable — and it was incapable of showing any of what was wrong. Changing what was measured was what changed the decision.

Common content strategy mistakes

The errors below recur across UK content programmes. Most are planning failures rather than writing failures, which is why better writing does not fix them.

• Commissioning one article per keyword variant. Several terms frequently express one intent. Four rows in a spreadsheet can be one question in a user’s head, and treating them as four briefs manufactures competition between your own pages.
• Publishing without auditing what exists. Roughly five organisations in six commission new content without checking the archive, which is how topics get covered twice and how a third of libraries end up receiving no impressions.
• Differentiating pages that should be merged. Genuine differentiation requires the pages to answer materially different questions — and if they did, they would not have been competing. Consolidation is the default and it is underused.
• Consolidating without redirects. Deleting the losing URLs discards exactly the accumulated signals the exercise was meant to concentrate, turning a good decision into a net loss.
• Guessing intent instead of reading the results. The live results encode a decision about intent made from far more behavioural data than you have. Checking takes thirty seconds and overrides your classification whenever they disagree.
• Planning by search volume. Volume and commercial value diverge sharply. Volume-led plans cluster at the informational end and under-serve the commercial investigation queries where buyers are actually choosing.
• Publishing clusters before pillars. A cluster page partly exists to show a topic is covered in depth, and that requires something for it to point at. Pillar first, or lose several months of the structure not being legible.
• Treating internal linking as tidying. The links are the structural signal, not decoration. Adding them on publication rather than in a later pass is the difference between a cluster and a pile.
• Reporting positions per keyword and nothing else. It cannot show cannibalisation, coverage gaps or dead inventory, so it leaves an organisation structurally unable to see its actual problems while looking accountable.
• Judging the programme at three months. Consolidation can move faster because the pages have history, but restructured publishing needs four to nine months. Changing course on one quarter of noisy data is how programmes never compound.

The 12-point content strategy checklist

Items one to four diagnose what you have. Items five to eight build the structure. Items nine to twelve keep it working once the initial effort is over.

1. Export every published URL with its impressions, clicks and queries. Your inventory. Establish how many pages receive effectively no impressions — typically around a third — because that figure decides whether to invest in repair or production.
2. Build the cannibalisation register. For your top twenty commercial queries, filter Search Console by exact query and check the Pages tab. More than one URL with meaningful impressions means those pages compete. Twenty minutes, no tooling.
3. Check intent against live results for every priority query. Search it on google.co.uk and look at the format that occupies the first page. Where this contradicts your assumption, the results win.
4. Identify orphan pages. Anything with no internal inbound links beyond a chronological listing. Over half of libraries have them, and they are cheap to fix.
5. Build the topic map from commercial priorities, not a keyword export. Three to six pillars matching what the business sells, each with the clusters that genuinely belong to it. Map every existing page onto it.
6. Work the fix hierarchy before producing anything new. Consolidate by default, differentiate only where intents genuinely differ, canonicalise near-duplicates that must stay live, retire what serves nobody. Implement the redirects.
7. Build or promote the pillar pages. Substantial pages that genuinely cover the topic and link out to every cluster piece, with clusters linking back. Existing strong articles can often be promoted rather than replaced.
8. Assign asset type by intent. Informational and comparison queries to cluster content; transactional queries to service pages, with articles linking through rather than competing with them.
9. Add internal links on publication. Into the pillar and across to siblings, with descriptive anchor text, as part of publishing rather than a later pass.
10. Change the reporting. Pages competing per query, cluster-level impressions and query counts, pages with zero impressions, and organic-assisted enquiries. Quarter-on-quarter, not month-on-month.
11. Establish a refresh and retirement cycle. Annual review of what needs updating rather than replacing, and what should be removed. Refresh in place; do not publish a better version alongside the original.
12. Name an owner for the topic map. Somebody who checks every new brief against it. Without this the structure reverts to keyword-led commissioning within about a year, usually without anyone deciding to.

At a glance — SEO content strategy summary

How Cloudswitched approaches content strategy

Cloudswitched provides SEO services for UK organisations, and for any site with an established library the work starts with an audit rather than a publishing schedule. In practice that means the inventory and the cannibalisation register first, intent checked against live results before any brief is written, a topic map built from what the business actually sells, and a consolidation pass before new production — which regularly means recommending that content be merged or retired rather than added to. Reporting is by cluster and by pages-competing-per-query rather than by keyword position, because the latter cannot show the problems that most affect an established programme. Where the honest finding is that the existing library needs repair rather than expansion, that is what we will propose.

Frequently Asked Questions

What is keyword cannibalisation and why does it matter?

Cannibalisation is when several of your own pages compete for the same search query, dividing the signals that would otherwise have made one page competitive. It matters because nothing appears broken: every page is individually fine, no warning is issued, and a technical audit will not flag it. The symptoms are multiple URLs of yours receiving impressions for one query, positions oscillating week to week as the algorithm reselects which page to show, a page dropping shortly after you publish something similar, and a blog post ranking for a commercial term instead of the service page you wanted. It affects roughly 68 per cent of established UK business sites.

How do I check whether my site has cannibalisation?

Use Search Console and allow twenty minutes. Open Performance, set the range to the last three months, go to the Queries tab and pick a commercially important query. Add a filter for that exact query, then switch to the Pages tab. If more than one URL appears with meaningful impressions, those pages are competing. Repeat for your ten or twenty most valuable queries and you have a cannibalisation register, which is the most useful SEO artefact most organisations do not have. No paid tooling is required, and only about 9 per cent of organisations maintain one.

Should I merge competing pages or make them different?

Merge, in most cases. Genuine differentiation requires the two pages to answer materially different questions, and if they did they would not have been competing in the first place. Consolidation means moving the best content into the strongest URL, redirecting the others permanently, and updating internal links to point at the survivor. Choose the surviving URL on evidence rather than preference: the page with existing links, history and impressions is usually the right host even where its content is weaker, because content is easy to improve and accumulated history is not. Always implement the redirects — deleting the losing URLs without them discards the signals you were trying to concentrate.

What is the pillar and cluster model?

A pillar page covers a broad topic comprehensively and targets the head term; cluster pages each address a specific subtopic with its own distinct intent. Every cluster links to the pillar and the pillar links out to every cluster. The internal linking is the mechanism rather than decoration — it is how you indicate which pages belong together and which is the principal page for the topic. Build pillars before clusters, because a cluster pointing at nothing cannot signal what it belongs to. We cover the model in more depth in our guide to creating topical authority.

Is topical authority a real Google ranking factor?

It is a practitioner model rather than a documented metric, and it is worth being clear about that. What is well established is that search engines index individual URLs, use internal links to understand relationships between pages, and have to select a canonical version when pages are near-duplicates. What is consistent with published guidance is that content should be made primarily for people and that demonstrated expertise matters in how quality is assessed. That sites covering a subject comprehensively accumulate some transferable advantage within it is a repeated observation whose mechanism is not published. The method is still worth using, because its benefits trace to mechanisms that are documented — chiefly avoiding cannibalisation and linking deliberately.

How do I work out search intent for a query?

Search it and look at what ranks. Assign a provisional category first — informational, commercial investigation, transactional or navigational — then check the live results on google.co.uk. If the first page is guides and explainers, the query has been judged informational and a service page will not rank. If it is comparison tables, that is the expected format. If it is product or service pages, an article is the wrong asset. If local results dominate, the query carries local intent regardless of wording. Where the results contradict your classification, the results win, because they reflect an assessment made from far more behavioural data than you have.

Why is so much of our published content getting no traffic?

Around 31 per cent of articles on an established UK business site receive effectively no organic impressions, and the usual causes are structural rather than editorial. The page may be competing with two or three of your own pages for its target query. It may be orphaned, with no internal links pointing at it beyond a chronological blog listing — true of over half the libraries we audit. Or its format may be mismatched to the intent that actually ranks, in which case it could not have competed however well written it was. Exclude genuinely non-search content before drawing conclusions, then treat the remainder as recoverable assets rather than sunk cost.

Should we publish more content or fix what we have?

Establish your dead-inventory figure first, because it decides the answer. If around a third of your library receives no impressions, the constraint is structure and the first money should go on audit, consolidation and internal linking — you are recovering assets already paid for, and consolidated pages move faster than new ones because they carry existing history. If the figure is genuinely low, under about ten per cent, the library is working and the constraint is coverage, so production is the right investment. Working out which situation applies takes a Search Console export and an afternoon.

How long before a restructured content programme shows results?

Four to nine months for the effect of restructured publishing to be reliable, with consolidation acting faster — sometimes within weeks, because the pages already have history and you are concentrating rather than creating signals. Judge it quarter-on-quarter rather than month-on-month: organic data is noisy, seasonal and subject to algorithm changes nobody controls, and confident causal explanations for a few per cent of monthly movement are over-reading it. The common failure is changing course on one quarter of data, which prevents the programme ever compounding.

What should we report on instead of keyword rankings?

Four things. Pages competing per query, which is the cannibalisation metric and should trend towards one. Cluster-level impressions and the number of distinct queries each cluster appears for, which is the closest practical proxy for coverage being recognised. Queries per page, and the count of pages with zero impressions, which is your dead inventory. And organic-assisted enquiries, acknowledging that attribution across a long consideration period is imprecise — judging informational content on last-click conversion will systematically undervalue it and lead to cutting the content that supports your commercial pages.

How much does this cost in the UK?

Indicatively for 2026, excluding VAT: a content and cannibalisation audit £1,500 to £4,500; a topic and intent map £2,000 to £6,000; consolidation work £150 to £400 per topic; researched long-form articles £300 to £900 each, more for specialist subjects; or a managed retainer covering planning, production, linking and reporting at £1,200 to £5,000 a month. Spend in roughly that order, because an audit costing a couple of thousand pounds frequently identifies enough consolidation opportunity to lift existing pages that collectively cost tens of thousands to produce.

Does AI search change content strategy?

In two real ways and rather fewer than is often claimed. Structured, specific, attributable content with concrete figures and direct answers is better placed to be cited by systems that synthesise answers than content written to hit a word count around a keyword — the thin keyword-led page is the format most exposed. And some informational traffic now never arrives, because the answer was delivered without a click, which lands hardest on the top-of-funnel content that volume-led planning over-produces. What is not yet reliable is any method for engineering citation: behaviour varies between systems, changes without notice, and most organisations cannot measure it at all. Treat guarantees in this area with scepticism.

Related reading

More guidance on planning, building and measuring UK business technology and marketing:

• Headless CMS versus traditional CMS: choosing the right architecture
• Virtual CIO versus IT consultant: understanding the difference
• Network documentation: building a map that actually gets used
• Measuring Microsoft 365 Copilot ROI
• In-house versus managed IT support: a UK cost comparison