An editorial system with AI inside it, not a text generator. First-party research, verification before publishing, and upkeep of what you already have.
Producing more isn’t the problem
Generating a hundred articles with a model is trivial and worth nothing. Google has been acting against mass-produced content with no added value since 2024, and assistants don’t cite what adds nothing either. The usual result of an operation like that is a bigger site, crawled worse, with less traffic than before.
Illustrative. The usual outcome: a bigger site, crawled worse, with less traffic than before.
The real bottleneck was never the writing. It was having something to say, checking it, and keeping it current. That is where AI genuinely helps, and where an editorial process is needed for it to help.
What we do
An editorial system, not a generator. We define what your brand talks about, in what voice, from what sources and within what limits. The model works inside that frame; without it, it produces correct, empty text, which is the worst outcome because it looks good.
Research and first-party data. What sets a piece apart is what only you can tell: your figures, your cases, what you have seen fail. We build the process to get that out of your company, which is the part no tool does on its own.
Translation and adaptation across markets. Not literal translation: adapting examples, references and the search engines of each market. It is where AI performs best and where it shows most when it is done badly.
Verification before publishing. Every figure, date and quote is checked against its source. An invented fact on a commercial page costs far more than the time saved generating it.
Maintaining what you already have. Updating twenty pages that already get visits almost always pays better than publishing twenty new ones. We find the ones that have aged, the ones that contradict each other and the ones competing for the same query.
Where we draw the line
We don’t mass-produce near-identical pages to cover query variants. We have seen it fail, and we have seen it fail on this very site: a set of pages like that, published years ago, ended up diluting the crawling of the whole domain. We retired them.
We also don’t sign AI-written content as a person when the format implies real authorship, and we don’t publish customer testimonials nobody gave.
What we measure
Organic traffic and conversions on the pages worked on, rankings for the target queries, pages cited by AI assistants, and cost per published piece against the previous process. We always compare against the same period last year, not against last month.
Why us
We have spent years producing content in four languages for sectors where a factual error is expensive. That is why the process has so many checks: they aren’t bureaucracy, they are scars.
And we use AI across the whole flow, by name: Claude and GPT for research, drafting and adaptation, and people to decide what gets said, check it, and answer for it.
