Content marketing — what actually happens to content after you publish it

Content marketing is sold in one sentence: publish consistently and the customers will come. The sentence is true to the extent that "train and you will win a marathon" is true — it omits everything that decides.

This text is about what happens after you press publish. We start with our own numbers, because they are the most honest material we have on the subject, and they are not flattering.

What our own content production does

We have run this site for years, we publish regularly, and we have full access to Search Console. Here is what it says for the last ninety days.

Google shows 14% of our articles. Not "shows them high" — shows them at all, for any query, even once. The other 86% do not exist for anyone who does not already know the address.

On the impressions that do exist, the click-through rate rounds to zero. Clicks happen, but against the number of impressions they do not register on the scale. The most-shown article on the whole site sits at an average position of 70 — the seventh page of results, where an impression is a statistical event rather than a visit.

We put this at the top of a text about content marketing for one reason: every article on the subject you will read was written by somebody who did not show these numbers of their own. We show ours before advising anything.

A text nobody reads looks exactly like a good one

This is the heart of the problem and the reason content marketing is so easy to run badly for two years.

An article that brings customers and an article nobody has seen look identical in the admin panel. Both are published, both have a date, a title and a decent heading. The difference only surfaces in data most companies do not open, because opening it is unpleasant.

The most common cause is mundane and has nothing to do with writing quality: the text answers a question nobody asks. It exists because somebody in the company decided the subject was important, or because Wednesday came round in the calendar.

We have our own uncomfortable example — the Polish-language edition of the very section you are reading.

Six documents, over twenty-one thousand words. The total demand for every phrase they declared: about 1,200 searches a month on the Polish market. Four of the six stood on a phrase the Keyword Planner does not return at all, or returns below the threshold of relevance — including the two longest texts.

This is not an example borrowed from somebody else's company for illustration. It is work we paid for ourselves and are now rewriting. This text is one of the rewritten ones.

Two of your own texts on one query

The second way content wastes itself quietly is less obvious than the first: not a shortage of text, but a surplus of texts about the same thing.

In our Search Console data, twenty-eight queries are served by more than one of our addresses. The record holder — a question about the cost of a web application — has four of our pages, standing at positions 21, 37, 66 and 89. None of them is bad. Each simply takes part of the same signal from the others, and Google has to pick one anyway.

Companies fall into this from good motives: somebody writes about costs, six months later somebody else writes about budgeting, a year later a calculator appears with a description. Each text sensible on its own; together they compete for the same question.

Checking takes a minute and needs no tool: type `site:yourdomain.com` and the phrase into Google. If three of your pages come up, you have your answer.

What we do about it, because that question follows the diagnosis. The order is always the same and the first step is the hardest:

1. Decide which text survives — not the longer or the older one, but the one standing on a phrase with measured demand.
2. Move across what is missing from it — a merge is not a deletion. Two or three sections usually survive from the absorbed text; the rest turns out to be repetition.
3. Redirect the old address permanently to the survivor. Without that the old address keeps answering, only now without content — and a search engine reads that as a page in good order.
4. Repoint the links that led to the absorbed text. This is the step that gets forgotten and takes longest to bite. In the last merge in this section we found thirteen such references across four language editions — none visible in the body text, all sitting in relation fields.
5. Declare a canonical where a merge is impossible — for a product in two categories or a print version, a redirect is not an option and rel="canonical" consolidates the signals without removing the page.

You lose less than it seems: a redirected address passes on what it managed to gather. Keeping two texts to "have two chances" does not work — Google will pick one regardless, only it will pick for you.

Where we actually rank well

We went through every query where we sit in the top fifteen — the range where a result is genuinely reachable. Almost all belong to two groups: our own company name together with its misspellings, and local service queries pointing at the home page. From our entire article output, only one subject is in there.

You rank well on a brand because nobody else is called the same — that is not a credit to the content. You rank well on a local query because there are a dozen competitors within a few miles, not a few thousand. Only the third group — subject queries — is what content marketing is about, and it is the slowest to arrive.

This pattern is the rule in small companies rather than the exception, and it is almost always mistaken for success, because the report shows growth. The distinction takes five minutes: filter out of the report every query containing your company name. What remains is the result of work on content. The rest is the result of existing and having a website.

What Google wrote about content written for search

In the Google Search Central documentation, on the page "Creating helpful, reliable, people-first content", there is a sentence that settles most arguments about how to write:

> If the "why" is that you're primarily making content to attract search engine visits, that's not aligned with what our systems seek to reward.

The same page proposes a test anyone can run on their own text: is this a page you would want to bookmark, share with a friend or recommend?

It is worth setting that against the previous section. Google's objection is not "you wrote for a phrase" — phrase research is fine, and without it you get our 86%. The objection is: the text exists purely in order to exist in search, and contributes nothing beyond that. Those are two different things, and the difference shows in whether a reader finishes knowing something they did not know.

"Scaled content abuse" — the name Google gave to what many companies do

The Google Search spam policies carry a separate category, worth reading by anyone who orders content by the piece.

The definition:

> Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users.

Note the third of the five examples listed: "Stitching or combining content from different web pages without adding value".

That is precisely the disease we found in ourselves. Six texts in one section repeated each other: three explained the same customer journey from scratch, three described the same keyword tools, two quoted the same permissible length for a meta description. Nobody was stitching this from other people's pages — we were stitching ourselves to ourselves, which is a variant of the same error and leads to the same result: a reader who lands on two of those texts gets two answers to the same question. In our case they contradicted each other: one advised publishing monthly, the other weekly.

AI for writing text — what Google actually says

This is now the most frequently asked question in the field, so the answer plainly: Google does not forbid content written with AI. It forbids something narrower, written in the same documentation:

> If you use automation, including AI-generation, to produce content for the primary purpose of manipulating search rankings, that's a violation of our spam policies.

Read that literally, because it makes the rest of the debate about AI in content beside the point. Google judges purpose and added value, not the tool. There is not a word in the spam policies about text having to be produced by hand; there is a great deal about why it was produced and whether it helps anyone. The same provision covers human writing, incidentally — an article written by a person purely to fill a calendar falls under the same definition.

Google also states an expectation worth reading as a question to your own editorial process:

> Are you providing background about how automation or AI-generation was used to create content?

In practice the difference is easy to spot from one question: does the text contain anything the model could not have known? Your prices, your client's case, your measurement, a photograph of your work, a sentence you heard in a meeting. If the whole text could be generated from the title alone, it does not matter who wrote it — it contributes nothing and joins the 86%.

Where to start instead of the calendar

The order we see most often is the reverse of the effective one. First a publishing calendar appears, then topics are hunted to fill it.

We will show what that leads to on ourselves, because the example is fresh. Until yesterday the Polish edition of this text declared the phrase "tworzenie treści dla firm" — content creation for business. It sounds sensible, it describes the contents, it was chosen by somebody reasonable. The Keyword Planner returns no volume for it at all. The same session showed that the market asks about the same thing differently: content marketing, several thousand searches a month, and nobody in our site held the phrase. Four thousand words of work stood under a heading nobody types, and one check was enough to see it.

The one-topic test — twenty minutes, four questions

This is not a methodology, just four questions that must be answered before the text is commissioned. Each has a binary outcome, so the step cannot be passed on instinct.

1. Does anyone type it? Keyword Planner, free, a Google Ads account without campaigns is enough. Enter the phrase and two natural variants of it.

• zero volume → not a subject for an article. It may belong on an offer page or in a sales conversation, but do not expect traffic.
• there is volume → carry on, but check whether the whole family reduces to one inflection of the same thing — three variants at 50 each is not 150, it is 50.

2. Who types it, and why? Enter the phrase into Google and look at the first page of results.

• shops and offer pages only → a buying query; an article will not get in,
• definitions only → a definitional query, and a caution: this is today's most exposed category, because an AI summary answers it instead of you,
• guides and comparisons → this is where an article belongs.

3. Have you already written about it? site:yourdomain.com plus the phrase, in Google, thirty seconds. If two of your pages come up, you do not write a third — you merge the two that exist. We skipped this step six times in a row in one section, which is why it has a number of its own here.

4. What in this text will be yours? One specific sentence: a price, a client's case, a measurement, a photograph of your work, a decision you regret. If you cannot point to it now, the text will be made of what is already online — and will be replaceable by two sentences of summary.

Only after four answers does a calendar make sense — one that follows from how much you have to say, rather than the other way round.

On the question "how often should we publish" we have no honest numerical answer and will not pretend otherwise. We do have an observation from our own data: our problem was never the cadence, it was that eighty-six per cent of the work was invisible. Publishing faster would have deepened that result, not improved it — because the fourth question would have gone unanswered sooner.

Which formats, and for what

Not everything needs to be an article. Matching format to purpose saves more work than any editorial technique.

A guide article works when the question is typed into a search engine and needs more than a paragraph. The only format that works on its own for years, and the only one where phrase research is mandatory.

A client case is the strongest material a service business can have and the rarest to appear, because it needs the client's consent and figures from before and after. Without them it is an advertisement with a name attached; if you cannot give the name or the data, write that plainly instead of rounding upward.

Something to download (a report, a template, a calculator) turns anonymous traffic into an address — provided it has value in itself, including for somebody who will never buy from you. What happens to that address next: email marketing.

Video is the most overrated and the most underrated format on this list, depending on what it is made for.

When video makes sense, and when it is spending out of momentum

There is one control question: does the client have to see how it works in order to decide? If so, video replaces a conversation. If not, it replaces a paragraph of text and costs several dozen times more.

Three uses that hold up:

• A product in action — software, a machine, a device, a process. A two-minute screen recording can replace an hour-long sales presentation and, unlike it, works at two in the morning.
• A client speaking — provided they say something specific ("we stopped entering invoices by hand") rather than a generality ("we recommend this company"). A generality filmed is still a generality.
• Explaining something convoluted — an insurance process, the stages of an audit, how a charge is calculated. Anything that takes a page in writing and thirty seconds in a drawing.

Three things about budget. A phone recording is sometimes better than a studio one — a client speaking in their own office is credible precisely because it looks that way. A screen recording needs no camera at all and is the cheapest way to show software. Professional production makes sense where the material will sit on the home page for two or three years — then the cost spreads across thousands of views.

Technically: H.264, 1080p at most for a website, hosted on YouTube or Vimeo rather than your own server (video served from your own server can wreck the loading time of the whole page), and autoplay only for short clips without sound. How that plays into page speed we cover separately under search positioning.

Who writes it: in-house, agency, freelancer

This question usually comes too early — before it has been settled how much content you actually need. The answer follows two things, and neither is a preference: how much of it there is and whether it repeats on a steady rhythm.

Irregular work in bursts, of varying scope, behaves differently from five texts a month in the same format. The first survives being paid for by the piece; the second needs a process, because with repetition the cost moves from writing to coordination — and coordination is what starts to cost most.

A freelancer suits irregular work and specific commissions. There is one serious drawback: on specialist subjects you pay for the time somebody spends learning your industry, and you pay for it again with every change of supplier.

An agency makes sense for recurring work meant to run without your involvement, and where you need skills you would not hire separately — text, graphics, video and publishing at once.

A staff post only pays once content is part of the product rather than its promotion, and once there is so much of it that outsourcing costs more than a salary. In practice that covers few companies and is sometimes a decision taken for reasons of ambition.

Whatever the model, one thing decides whether the result can be published without rewriting: the brief — specifics, not "a professional tone". Instead of "not too promotional": "cut the paragraphs about features, keep the ones about what changes for the client". Instead of "in our style": two texts you consider good and one you consider bad. The most useful style guide we have seen was a list of forbidden words.

The tip that saves most: put the phrase and its volume in the brief. A writer who knows they are writing for a query searched a hundred times a month writes a different text from one given only a subject.

What it costs and when anything shows

We are not going to quote a price range here, and the reason matters more than the range would. Our figures come from the Polish market; converting them at an exchange rate would read as credible and would be invented, because rates for writing do not track currencies — they track how much of the work stays on your side.

What does transfer to any market is what decides where in the range you land:

• Prior knowledge. A text about accounting for an accountancy firm costs differently from the same text for a company outside the field, because in the second case you are also paying for the time spent getting into the subject.
• Who supplies the material. If you hand over a finished interview, data and examples, you are paying for editing. If the writer has to find them, you are paying for research, which is often longer than the writing.
• Responsibility for the outcome. "A text for a given phrase, with research and a measurement after three months" is a different service from "a text on a given subject", though both are called copywriting.
• Rhythm. A standing arrangement costs less per text than one-off commissions, because the onboarding cost disappears from each of them.

On time there is one thing worth knowing that sellers omit: the effect in search is measured in months, and the first threshold is not position but indexing. A text Google has not taken into its index has no position to improve. With us that threshold turned out to be the main bottleneck — we set it out under search positioning.

What happens to a text after publication

Publication is treated as the end of the work and is at most half of it. Three things are worth doing the same day, because nobody comes back to them later: send it where you already have people (the mailing list, a company profile, an industry group), link to it from a page that already works — a link from your own visible page is worth more than three social posts, and is also how Google learns the new address exists — and check after a week whether it has been indexed. Not the position, the presence; if the address is not in the index after two weeks, no promotion will fix that.

One caveat so as not to sell hope: with us distribution fixed nothing, because the problem was earlier, in the choice of topics. Sending out a text nobody was looking for increases the number of people who do not need it.

Then comes the decision about what to do with an old text, and here refreshing is sold as a universal remedy. It is effective within a narrow band. A text with impressions sitting just past the first page — the cheapest work in the field, because Google has already decided the subject suits you; a better title and tidier sections can change something in weeks. A text with impressions but sitting far back — polishing does not suffice, because the text does not answer the question as well as ten others; the decision is rewrite or merge. A text with no impressions at all — there is nothing to refresh; first establish whether it is in the index and whether anyone is searching for what it is about.

How to measure without deceiving yourself

Three measurements, in this order, because each only makes sense after the previous one:

1. Does Google show this text at all? Search Console, performance report, filtered to the address. Zero impressions means every further metric is moot. That is the question we opened with, and our answer was "no" in 86% of cases.
2. Does anyone click? Impressions without clicks are usually a problem with the title and description in results, or a position so distant that an impression is an accident. The distinction is simple: check the average position.
3. Do the people who arrived do anything? Only here do enquiries, sign-ups and contacts come in. What to do with that traffic we set out in the text on conversion rate.

What not to measure as a mark of success: the number of published texts, the number of characters, the number of impressions divorced from position. All three rise as the work piles up, regardless of whether it brings anything — with us they rose throughout the period when the click-through rate was rounding to zero.

What this text deliberately leaves out

• How to write for search sentence by sentence — structure, headings, meta descriptions, weaving in phrases. That is a separate document: SEO copywriting.
• Images, graphics and where to get them legally — stock libraries, licences, optimisation: image copyright.
• Whether a business needs a blog and what it costs in time — business blog.
• How search works and what decides position — website positioning.
• Running social profiles — social media and the company website.