There is a version of AI marketing that turns out posts by the dozen. It is fast, it is cheap, and it sounds exactly like everyone else's, because everyone is pointing the same tools at the same prompt. We took a different approach with a client who wanted a stronger presence online. We put the effort into working out what was actually worth saying, and we put a person in front of everything before it goes anywhere near an audience.
The problem was not output, it was direction
The client was not short of things to say. They were short of a plan for saying it. There was no clear picture of who they were talking to, no consistent themes running through the year, and a website that quietly failed to answer the questions their buyers were actually asking.
Producing more content would not have fixed any of that. Producing the right content would.
Working out what is worth saying
Who the audience actually is. We built out the segments the business sells to, what each one cares about, the language they use to describe their own problems, and where they go looking for answers. Not demographics for their own sake, but the practical detail that tells you what is worth writing about.
A content strategy with a reason behind it. Themes, pillars and a calendar where every slot exists for a purpose. The test we apply is simple: if you cannot say which audience a piece is for and what it is meant to do, it should not be on the plan.
Content packs, not one-off posts. For each campaign we assemble the angle, the key messages, the supporting points worth making, the formats that suit them, and a brief clear enough to work from without guessing.
An audit of the SEO gaps. We compared the questions buyers ask against what the site actually covers, and against what competitors rank for. The interesting findings were rarely about keywords. They were about whole topics the business had strong opinions on and had simply never published.
Setting up for GEO. Generative engine optimisation is about being visible when the answer arrives from an AI assistant rather than a page of blue links. In practice that means content structured so it can be quoted accurately, factual statements that stand on their own, and machine-readable signals that describe what the business does and who it serves.
Nothing ships without a person
Everything passes a human gate before it is published, and that gate is not a formality. It is where tone gets corrected against how the business actually talks, where a claim that sounds plausible but is not quite right gets caught, and where anything that could have been written for any company in the sector gets sent back.
That matters more in a technical business than a general one. Copy that is fluent and confident but subtly wrong about what a product does is worse than no copy at all, and a model has no reliable sense of which of its own sentences are the risky ones. A person who knows the business does.
The analysis is a different matter. Segmentation, gap-finding, structuring and comparison are exactly the tasks where these tools are genuinely strong, and where the output can be checked against what we already know about the business.
Why the separation matters
The engine we use holds no client information at all. Each client's brand rules, audiences and product knowledge live separately, so improving the engine improves every client's output, and changing one client's rules changes only theirs.
That separation is what keeps this repeatable without it turning into the same content with different logos on it.
What changed
The client moved from posting when someone remembered to a plan with a reason behind it: a defined set of audiences, a backlog of topics drawn from real gaps rather than guesswork, and a site being built out to answer the questions people are genuinely asking. What goes out sounds like them, because someone who knows how they sound decides whether it does.
If your marketing feels busy but directionless, the useful question is not how quickly you can produce posts. It is whether you know what those posts should be about. We are happy to talk that through.