SEO content written by people
Content AI cites and people actually read.
Anyone can generate a thousand articles in an afternoon now. That is exactly why generic content is worth nothing and content that adds something is worth more than ever. We write with people who know the industry, with an SEO method, and with the rigour that gets a model to cite you as a source.
What changed when writing stopped costing anything
For twenty years SEO content was a volume problem: publish more than the next company. That is over. Producing text is no longer the barrier, and what used to set you apart — having articles — everyone now has by tomorrow morning.
What is scarce is what cannot be generated: original data, real experience, judgement, and someone willing to take a position. That is exactly what Google looks for through E-E-A-T signals — experience, expertise, authoritativeness and trust — and what models look for when deciding who to cite.
And there is a deeper shift: around 60% of the sources AI cites are not in Google's top 10. Good content competes inside the AI answer before it competes in the search results, and there the biggest publisher does not win.
How we write: the method, not the template
First, search intent. What the person searching expects to find: a definition, a comparison, a procedure or a price. It is the decision that has sunk the most articles, because an excellent piece that answers a different question does not rank.
Then the brief: what it has to answer, which entities need to appear, what has already been said and what is missing. We work in clusters — a pillar piece and its satellites — to build topical authority over a subject, instead of a pile of unconnected articles.
And the writing, with someone from the industry. Not a generic writer with a brief: someone who has done the job and knows where the half-truths are. It shows by the second paragraph, and your reader notices before Google does.
How it works
From the search to content that gets cited
Step 3 is the one that cannot be automated, and the only one your competitors cannot copy in an afternoon.
How to write so that AI cites you
A model does not read like a person: it retrieves fragments. So structure carries as much weight as substance. Citable content answers the question in the first paragraph of each section, uses headings that are real questions, defines terms the first time they appear, and does not hide the conclusion in the final line.
Underneath there is technical work almost nobody does: structured data, well-defined entities, an llms.txt that says what to read, and self-contained paragraphs so a retrieval system — RAG — can quote them without dragging in context. It is the same GEO and AEO work as in the other solutions, applied to content.
And the part that matters most and happens least: publishing something worth citing. Original data, a small study, a table nobody else has. One number of your own, with a method behind it, gets cited for years.
What about AI for writing? Where it works and where it does not
We use it, and we will tell you what for: research, organising material, generating headline variants, checking consistency across a hundred articles, spotting gaps against competitors. At all of that it is better and faster than a person.
Where we do not use it is judgement. A piece generated end to end is correct and says nothing: no experience behind it, no position taken, no data that was not already on the internet. Google does not penalise it for being AI-written — it says so explicitly — it ignores it for adding nothing. And a model will not cite it either: why would it, when it already knows that.
That is why we call it human-first and not “no AI”. The tool is there; the byline belongs to a person.
This is not romanticism: your competitors already have everything in the right-hand column, and nothing in the left.
How to choose a content agency: six questions before you sign
Plenty of people put the question to an AI first. It is worth looking at what comes back: when you ask a model to recommend an agency, the sources it cites are almost never the agencies' own websites. They are third-party studies, press releases, roundups in trade media, forum threads and LinkedIn profiles. We track this ourselves, month after month.
It is the same mechanism you want working for your content, so the question answers itself: a content agency that cannot get itself cited is unlikely to manage it for you.
These are the questions we would ask. They work on any agency, and we answer them the same way if you put them to us.
- 01
Who writes, and what have they done in our industry? Ask for names and bylined pieces, not a “content team” profile.
- 02
Do you use AI, and for what exactly? The honest answer is not “no”: it is naming the parts where you do and where you do not.
- 03
How do you decide what to write about? If the answer is a topic calendar rather than a map of search intent, someone is publishing blind.
- 04
What original data are we going to publish? It is the most cited thing there is, and almost nobody proposes it, because it is work.
- 05
How do you measure whether AI cites us? With per-model tracking and history, or it is not a measurement.
- 06
What happens to the published work if we stop? The content is yours and should keep earning. If it only works while you keep publishing, it was a treadmill, not an asset.
What gets measured, beyond words published
Articles per month is not a result, it is an invoice. We look at which pieces reach positions that bring someone with intent, which ones are being cited in AI answers, and which ones need updating before anything new gets written.
That last part is usually the most profitable and the one no agency proposes, because it bills less: updating ten articles that already rank beats publishing ten new ones. And if you sell in several countries it crosses over with international SEO, because translated content and localized content do not compete the same way.
What we actually do for you
- A search-intent map and clusters built around pillar content.
- Briefs with the entities and the questions that have to be answered.
- Writing by people from your industry, with their name on it.
- Structure designed so AI can cite the fragment.
- Original data and small studies: the thing that actually gets cited.
- Updating what already ranks, before publishing more.
FAQ
- What is SEO content?
- Content written to answer a specific search with the depth the searcher expects, and structured so search engines and AI models can understand it. It is not about inserting keywords: it is about getting the intent right, covering the semantics of the subject and keeping it readable for a person, who is the one you have to convince.
- What is SEO content writing, in practice?
- Starting from search intent, not from the keyword. Then: a brief with what must be answered and which entities must appear, someone who knows the subject doing the writing, clear structure with the answer near the top of each section, and something of your own — a number, an example, an opinion — that is not already in the ten results above you.
- Do you use AI to write?
- For research, organising material, headline variants and consistency checks across many pieces, yes. For judgement and for taking a position, no. Google does not penalise content for being AI-written, but it ignores content that adds nothing, and a fully generated piece rarely adds anything: no experience behind it and no data that was not already online.
- How many articles a month do we need?
- That is the wrong question, and the one that burns the most budget. It depends on how much room there is in your subject and how much of what you already have is untapped. On many projects the first quarter returns more from updating what is published than from adding anything new.
- How long does content take to work?
- What already exists and gets updated moves within weeks. New pieces take months, and it depends on your domain's authority and how contested the subject is. AI citations show up sooner: there you do not have to wait to climb positions, because the answer is not built from the order of the SERP.
- How do you get AI to cite your content?
- Three things at once. Structure, so the model can retrieve a fragment that stands on its own. Entities and structured data, so it knows what and who you are talking about. And something worth citing: original data, a small study, a table nobody else has. The third matters most and is the one almost nobody does.
Keep exploring
