Thirty posts a month is the red flag, not the writing

If you show me a content programme and let me ask one question before I look at anything, I ask how many pieces a month it publishes. The answer predicts what I am going to find better than any sample of the writing does.

This is not snobbery about volume. It is arithmetic. Volume does not create a structural error - it multiplies one. Publish thirty pages a month with no keyword map, and by Christmas you do not have one cannibalisation problem, you have a hundred and eighty of them, all interlocking.

Why cadence is the tell

Think about what has to happen for a page to be good. Somebody decides the page should exist. Somebody decides which search it owns and checks nothing else already owns it. Somebody supplies the facts. Somebody decides what links to it and what it links to. Somebody reads it for invented claims. Somebody rewrites the opening.

Only one of those is writing, and it is the only one AI made faster. The other five take about the same time they always did. So a programme that went from four pages a month to thirty did not get seven times more capable at deciding, checking and linking. It got seven times faster at the one step that was never the bottleneck.

The generation step is the only one the AI does, and it is the smallest part of the job. Cadence tells you which of the other five got dropped.

That is why the number is diagnostic. Thirty a month with two people doing the deciding is not a productivity story, it is a story about which checks were quietly removed to hit the number.

What the bill looks like when it arrives

The costs do not show up in the month you publish. They show up six to eighteen months later, in a shape most owners do not recognise as being caused by publishing:

• Your positions on your main terms start oscillating between two of your own URLs, because you now have several pages half-covering the same thing.
• Your best-performing old page starts slipping, and nobody connects it to the twelve newer pages that overlap it.
• Search Console shows impressions climbing and clicks flat, which is what a site full of pages that rank at position 14 for things nobody buys on looks like.
• Somebody eventually has to work out what to merge and what to delete, and that job is far more expensive than writing was, because it requires understanding every page rather than producing one.

That last one is the real bill, and it is worked through in the hidden cost of AI articles at scale.

What a sane cadence looks like

There is no universal number, but there is a test: you can publish as fast as you can decide, and no faster.

In practice, for a local service business with one person doing the deciding, that lands somewhere between two and six substantial pages a month. Not because more is impossible, but because more than that and the decisions start getting made by default - which page owns which term gets decided by whichever page Google happens to prefer, rather than by you.

Two other things worth saying about rate:

• A burst is fine if it is planned. Building fifteen service and location pages in a fortnight from a map you drew first is a completely different activity from publishing fifteen blog posts because the calendar said so. The first has a structure; the second is filling slots.
• Consistency matters less than people claim. There is a widespread belief that Google rewards regular publishing. I have never seen it in client data, and I have seen plenty of sites do very well publishing rarely. Ten good pages beat a hundred islands, and cost far less to maintain.

Where the effort should go instead

For a local service business the pages that earn money are your service pages, your location pages and your Google Business Profile. Not your blog. Cheap generation does not change that arithmetic; it just makes it cheaper to build the wrong thing quickly. I have made that case at length in why blogs are mostly a waste of time for local businesses.

If a programme is producing thirty blog posts a month and your main service page has not been touched in two years, the cadence is not the only problem, but it is the one you can see from outside.

Checking your own

Three checks, none of which need a tool beyond Search Console:

1. Count what you published in the last six months, and count how many of those pages have brought in an enquiry. If the second number is zero, the cadence is producing volume rather than results, and slowing down costs you nothing.
2. In Search Console, look for any search term where more than one of your URLs appears. Each one is a page competing with another page you paid for. If there are more than a handful, cannibalisation is already the main issue.
3. Compare impressions against clicks over twelve months. Impressions rising while clicks stay flat is the signature of publishing a lot of pages that rank where nobody looks.

If those come back badly, the answer is not to stop using AI. It is to publish slower and decide more, which is the whole structure set out in using AI for your website content without wrecking it.

When volume genuinely is the right answer

I do not want to leave this sounding like small is always right, because there are cases where high volume is correct and I have built them.

• A site that legitimately needs a page per suburb or per product, where the pages are genuinely different because the underlying facts are different, and where the map was drawn before anything was generated. Forty location pages built from a plan is not the same activity as forty blog posts built from a calendar.
• A migration or a rebuild, where the volume is a one-off and the structure already exists.
• A large catalogue where the alternative is having no page at all for things people are actively searching for.

What those have in common is that the deciding was done up front, in bulk, deliberately. The rate of publishing is high; the rate of decision-making was also high, it just happened in one sitting. That is a completely different thing from a standing commitment to thirty pieces a month, which guarantees that at some point the deciding will be skipped to hit the number.

The test is not the number. It is whether the number was derived from a plan or the plan was derived from the number.

If you are judging an agency

Cadence is the cheapest question you can ask, and you can ask it before you look at a single piece of work. Thirty pieces a month for a retainer that cannot cover thirty pieces of thinking tells you what you are buying before you read a word of it.

Two follow-ups that are more revealing than asking for samples:

• Ask which search term each of last month's pieces owns. If there is no answer, or the answer is a topic rather than a term, nobody is holding the map and the pieces are competing with each other.
• Ask what they decided NOT to publish last month, and why. A programme doing the deciding properly rejects things. One filling slots never does.