Content marketing that answers questions people actually ask

What content marketing is supposed to do

Content marketing has one honest job: to be the best available answer to a question a potential customer is already asking, in a place they can find it, published by someone they end up trusting.

Everything else people call content marketing is downstream of that. Rankings are a consequence of being the best answer. Shares are a consequence. Being cited by an AI assistant is a consequence. Treating any of those as the goal produces the familiar failure — a blog full of competent, sourceless, interchangeable articles that rank for nothing and are read by nobody except the person who commissioned them.

The second job, less discussed and often more valuable, is internal. A good content programme gives your sales team something to send. If five prospects a week ask the same question and the answer currently lives in an account manager’s sent folder, writing that answer properly is more commercially useful than three months of top-of-funnel articles about industry trends.

Strategy comes before the calendar

An editorial calendar is an output, not a strategy. Ours starts by answering three questions in writing.

Who is this for, and what are they trying to find out? Not a persona document with a stock photograph and a fictional name. A list of the actual questions, in the actual words, that someone asks at each stage — from “what is this category and do I need it” through “how do I choose between two suppliers” to “how do I justify this internally”.

Which of those questions is this business credible answering? Credibility is topic-specific and it is not transferable. A firm with deep experience in one domain has no standing in an adjacent one, and search engines are increasingly good at noticing. Google frames content quality around experience, expertise, authoritativeness and trust, and is explicit that trust is the most important of the four. That is a strategic constraint, not a checklist item.

What are we deliberately not writing about? The most useful line in a content strategy is the one that rules things out. Without it, the programme drifts toward whatever has search volume, and you end up with a page on a subject you cannot defend in a sales call.

Audit what you already have before commissioning anything

Almost every established site we look at has more pages than it needs and fewer good ones than it thinks. Before writing anything new, we inventory what exists and mark each page: keep, improve, merge, or retire.

The audit usually finds four things. Pages that rank on page two and would reach page one with an afternoon of work — the cheapest wins available to anyone. Pairs of pages targeting the same query, neither of which wins, because they are competing with each other. Pages that were accurate in 2021 and are now quietly wrong, which is worse than having no page. And a long tail of thin posts published to hit a monthly quota, which dilute the site without contributing anything.

Improving an existing page that already has links and history is nearly always faster than starting a new one that has neither. This is inconvenient advice for an agency to give, because rewrites bill less than new commissions.

Topic clusters, pillar pages, and one page per job

Authority accrues to subjects, not to pages. A cluster is a pillar page covering a subject broadly, surrounded by supporting articles that go deep on individual facets, all linked to each other deliberately rather than by accident.

The discipline that makes it work is one page per job. Every query you target is assigned to exactly one URL. When two of your own pages chase the same search, neither accumulates the signals it needs and search engines have to guess which one you meant — a problem that shows up as unexplained flatness in your reporting long before anyone diagnoses it as self-competition.

Search intent is the assignment, not the keyword

The same keyword can imply completely different jobs. Someone searching a term to understand what it means needs a definition and an explanation. Someone searching it with “vs” or “best” is comparing and needs a structure that makes comparison easy. Someone searching it with “how to” wants steps they can follow now, and will resent three paragraphs of context first.

The practical test is to look at what currently ranks. If the results are all comparison tables and you write an essay, you have not written a worse article than the winners — you have written the wrong format for the question, and no amount of quality fixes that.

Briefs are where content quality is decided

The difference between a good article and a generic one is usually settled before a writer starts. A brief that says “1,500 words on content strategy, target keyword content strategy services” will produce exactly what you would expect.

A brief we would hand to a writer contains: the reader and what they already know; the specific question the piece answers; the argument the piece makes, in one sentence; the sections and what each must establish; the sources to consult and any that must be cited; the internal links to include and the anchor text to vary; the title and meta description; and a short list of claims the piece must not make, which is often the most important part.

That last item exists because unbriefed writers fill gaps with plausible-sounding assertions. A brief that says “do not claim a percentage improvement, and do not name a client” prevents a whole category of problem before it reaches an editor.

Production: writers are not the bottleneck, experts are

You can hire competent writers. You cannot easily hire someone who has spent years actually doing the thing your buyers are asking about — and that person almost certainly works for you already and will not write an article.

So do not ask them to. The pattern that works is a structured interview: the writer arrives having read the brief, asks specific questions for half an hour, records it, and produces a draft the expert corrects rather than composes. Correcting is fast and comfortable; drafting from a blank page is neither, which is why the “we will get our engineers to blog” plan fails on every occasion it is attempted.

AI, honestly

Language models are good at some parts of this work and bad at the part that matters. They are useful for research triage, for generating structural alternatives, for finding inconsistencies across a long document, and for producing a first pass at something formulaic. They are not useful for the thing that makes content worth reading, which is a specific view formed by doing the work.

Google’s stated position is that its focus is on content quality rather than how content was produced — while also warning that using automation to produce content primarily to manipulate rankings violates its spam policies, and defining scaled content abuse as generating many pages mainly to rank rather than to help users. Both halves of that are true at once and the resolution is not complicated: build every piece on something a model does not have, and have a human edit and verify it.

Formats beyond the blog post

The default output of most content programmes is an article, because articles are easy to commission. Several other formats reliably outperform them for specific jobs.

The question page. One question, answered completely, at a URL whose sole purpose is that question. Unglamorous and disproportionately effective for search, and easy for a salesperson to send.

The comparison or decision guide. How to choose between options, including honestly saying when the answer is not you. Buyers are running this comparison whether or not you help them, and being the source that helped is worth more than being the source that pretended there was no choice.

The teardown or worked example. An analysis of a real thing, with the reasoning shown. Expensive to produce and very hard for a competitor to copy.

Original research. Survey your customers, aggregate anonymised data you already hold, or measure something nobody has measured. This is the only content type that other people are obliged to cite, because you are the source rather than a summary of one. It is also the format most likely to be quoted inside an AI answer, for the same reason.

If a diagram would do a paragraph’s job better, we will say so and have it drawn — that is graphic design work, and a good chart is cited more often than a good sentence.

Editing standards and the accuracy check

Editing and proofreading are different jobs and we do them at different times. Structural editing asks whether the argument holds, whether the sections are in the right order, and whether anything can be cut. Line editing deals with clarity and rhythm. Neither of them is a factual check.

The factual check is separate and deliberately boring: every number, claim, standard, product name and citation is opened and verified against its source. Anything that cannot be verified comes out. It does not get hedged into a vaguer form of the same assertion — that is how “studies show” enters a document.

On structure, the reader research is unambiguous and forty years old. Nielsen Norman Group’s testing found that 79% of users always scanned any new page while only 16% read word by word, and that concise writing improved measured usability by 58%, scannable layout by 47%, and all their improvements combined by 124%. The same study found objective language outperformed promotional language by 27% — which is a measured argument against the tone most marketing copy is written in.

That is why our pages carry meaningful subheadings, short paragraphs, front-loaded conclusions, and no promotional adjectives that a reader would have to discount.

Internal linking and distribution

New content that nothing links to is content nobody finds. On the day a piece publishes, it should be linked from the existing pages where the connection is genuine, with descriptive anchor text — Google’s own guidance is that link text should tell users and search engines what the linked page contains, and “click here” tells them nothing.

Structured data belongs here too, with a caveat worth having. Article markup, using the Article, NewsArticle or BlogPosting types, helps machines understand authorship and dates, though it has no required properties and is not needed for indexing. FAQ markup is a different story: the FAQ rich result was restricted in 2023 to well-known government and health sites and then removed from Google Search entirely in May 2026. Anyone still selling you FAQ schema for the rich result is selling something that no longer exists. We still write FAQs, because readers use them — we just do not pretend the markup earns a feature.

Distribution is the half of the plan that most programmes skip. Email to a list you own, adaptation for the social channels your buyers actually use, and internal circulation so your own sales team knows the page exists. That last one costs nothing and is skipped almost universally.

Refresh, consolidate, retire

Content decays. Facts age, competitors publish better versions, and platforms rename features. A programme without a maintenance cycle spends progressively more of its budget replacing ground it already held.

The decision tree is simple. If a page still ranks and only the facts have aged, refresh it in place and keep the URL. If the structure is wrong but the URL has history, rewrite in place rather than starting a new one. If two pages compete for the same query, merge the better content into one and redirect the other — redirects and canonical annotations are both treated as strong consolidation signals, while listing a URL in a sitemap is a weak one. If the topic is dead and the page earns nothing, retire it.

Reviewing published pages on a schedule is unglamorous and it is where a mature programme gets most of its remaining upside. It also pairs closely with technical SEO, since consolidation is a redirect and indexing question as much as an editorial one.

Measuring content against pipeline, not pageviews

Pageviews are a diagnostic. They are not a result, and reporting them as one is how content programmes get cancelled in year two by a finance director who was never given a real number.

What we instrument: impressions and clicks per page from Search Console, so visibility and click-through can be separated; engaged sessions and assisted conversions in Google Analytics 4, so a page that participated in a journey is not invisible because it was not the last click; and enquiry and stage progression in your CRM, which is where the actual answer lives. If those systems cannot be joined, joining them is the first piece of work — before any article, because otherwise you will be arguing about the programme with anecdotes.

Two signals we add that most reporting omits. First, which pages your sales team actually sends to prospects: a page a salesperson reaches for every week is doing more commercial work than one with a large session count and no conversations attached to it. Second, content decay — pages losing impressions month over month, caught while a refresh is still cheap.

What drives the cost

We do not publish prices, because the honest range before scope is known is wide enough to be useless. The variables:

How much research each piece needs. A question page drawing on your existing expertise is a fraction of a teardown requiring three expert interviews and original analysis.

Whether expert access exists. If we can get half an hour with someone who knows the subject, quality goes up and cost goes down. If nobody is available, the writer has to reconstruct expertise from public sources, which is slower and produces a weaker page.

The state of what you already have. A large archive needing an audit and consolidation is real work before anything new is written — and usually a better investment than the new writing.

Whether original research is in scope. Surveys, data analysis and the visualisation that goes with them are the most expensive content you can commission and the most durable, because nobody else can publish it.

Volume, last. It is the variable clients want to lead with and the one we weight lowest, because it is the one most likely to produce spend without return.

Where this connects to everything else

Content is one part of an organic search programme, and it fails without the other parts: a crawlable site, a sensible information architecture, and pages that load quickly enough to be read. It also increasingly answers a second question — whether your material is retrievable and citable inside AI assistants, which is the subject of generative engine optimization and a different discipline from ranking a page.

Distribution reaches beyond your own site through social channels, and the whole programme should sit inside a measured digital marketing plan rather than running as an unattributed cost centre.

If you already have an archive and no idea what it is doing for you, start with the audit. It is the cheapest piece of work available and it occasionally concludes that you do not need us to write anything at all.