A company we worked with was paying roughly €1,500 a year to an agency for its blog. The deal was one article a month. Twelve pieces a year, written by someone who had never seen the products, published into a section of the site nobody visited.
This is not an unusual contract. It might be the single most common line item in a small company's marketing budget.
Why anyone signs it
Because you've been told, correctly, that a company that never publishes anything slowly disappears from search. And because writing is work that nobody inside the company has time for — the people who know the subject are the same people who are busy doing the job.
So a supplier offers to take the problem away for a fixed monthly fee. It's a reasonable decision, made with no way of judging the result. That's the part that matters: almost nobody who signs one of these contracts has a way to tell whether it worked.
What twelve generic articles a year actually do
Very little. Not because the writer is bad, but because one piece a month, written by someone outside your trade, competes against everything else on the internet about that topic. It reads like it was written to fill a slot, because it was.
The uncomfortable question is not "is it good writing?" It's: in two years, has one customer ever mentioned an article to you?
What replaced it
A publishing system rather than a writing service. A human decides the topics — that part stays human, because knowing what your customers actually ask is the only real advantage you have. The drafting is automated. And every draft goes through a check before it can be published.
The check is the entire point. It refuses anything that invents a statistic, anything that reads as an advert for the company, anything that opens with the same tired formula every generated article opens with. If a piece fails, it doesn't go out.
The obstacle: the first version threw away good work
Our early version had a flaw that took a real failure to expose. When the quality check didn't respond at all — not a rejection, just silence — the code discarded the article. A finished, fully written piece, deleted because the reviewer never answered.
It now parks that article as a hidden draft with the reason attached, and returns the topic to the queue instead of burning it. A gap is better than a bad article. But throwing away a good one is worse than both.
What it costs now
- The recurring fee: gone
- Output: several articles a week instead of one a month
- Topics: chosen by someone who knows the business, not by a supplier filling a slot
- Nothing published without passing the check — a missed week costs less than one embarrassing article
The honest limit
Automated articles are good at the things anyone in your industry could write: how to choose between two types of machine, how to maintain something, what a regulation actually requires. They are useless at the thing only you know — why a specific customer chose you, what went wrong on a difficult job and how you fixed it.
Automate the articles anyone could write. Keep the ones only you can.
Those are also the ones that turn a reader into a phone call. Which is why the system exists to buy you the time to write them, not to replace them.
More field notes
Ownership · 10 September 2026
Your website works fine. You just don't own it.
From the outside everything looks normal. You find out the day you want to change a price — and discover that the hosting, the domain and ten years of search ranking sit in somebody else's account.
Operations · 8 September 2026
Somebody retypes the price list for three weeks. Twice a year.
Every supplier raises its prices at least twice a year, thousands of items at a time. In most distributors a person retypes them by hand. Nobody calls it a problem, because it has always been done that way.