Everybody Is Publishing More and Saying Less

Read ten company blogs in your sector this week and you will notice something. They are longer than they used to be, there are more of them, and they are interchangeable.

Same structure, same reasonable tone, same three subheadings, same closing paragraph about how the landscape continues to evolve. Nothing wrong with any of it and nothing memorable in any of it either.

Why this happened all at once

Stanford’s AI Index for 2025 reports that 78% of organisations said they used AI in 2024, against 55% the year before. That is one of the fastest adoption curves ever recorded for a business technology.

Worth being precise about what the figure means: it counts organisations reporting use in at least one function. It is not a claim that anything is working. But for marketing specifically, the direction is not in doubt, and it explains the convergence. When most of a sector starts drafting from similar tools with similar prompts, the output moves toward the middle, because the middle is what those tools are built to produce.

The middle is exactly where you do not want to be

The reason this matters commercially rather than aesthetically: the entire point of publishing is to be chosen over somebody else.

A prospect reading your article three days before a meeting is not evaluating your prose. They are looking for evidence that you have done this before and know something they do not. Competent, unobjectionable, structurally identical content provides no such evidence. It reads as a company that has a content calendar, which is not the same as a company that knows what it is doing.

And the volume increase makes this worse rather than better. When there were three articles on a subject, being one of them mattered. When there are three hundred, being indistinguishable from the other two hundred and ninety nine is the same as not existing.

What is missing is not effort, it is input

The failure is not that AI wrote it. It is that nothing was put in.

A prompt saying “write a post about workflow automation for small businesses” contains no information. The output can only be an average of everything already written on the subject, because there is nothing else for it to work from. It is not the tool being generic. It is the request.

Everything that would make the piece worth reading lives inside your business and has to be supplied: the number a client actually reported, the objection you hear in every third sales call, the thing you used to believe and stopped believing, the specific way this goes wrong in your sector that nobody outside it knows about.

None of that is on the internet. It is in your head, in your team’s heads, and in your project history.

How to use AI properly for this

The right division of labour is the opposite of the common one. Most people use AI to generate the substance and then edit the wording. Do it the other way round.

Bring the substance yourself. Talk for fifteen minutes about a real job that went well or badly. Record it. That recording contains more genuine insight than any prompt will produce, and it is the raw material.

Let AI do the work it is genuinely good at. Transcribe the recording. Find the three arguments buried in it. Point out where the reasoning skips a step. Structure it. Draft the version that says the same things in a readable order. Check it against what already exists on the subject and tell you which parts are ordinary and which are not.

That last one is the most underused. Asking a system to tell you which of your paragraphs could have appeared in anybody’s article is a fast and unflattering edit, and it is the difference between publishing something and publishing something worth reading.

Then research at a scale you could not before. Reading forty competitor pages, ten regulatory documents and your own last two years of client notes to find what nobody has said is genuinely useful work, it is now cheap, and almost nobody does it because they are busy generating drafts instead.

What this means for search

The obvious worry is that publishing less means ranking for less. In practice the arithmetic has changed.

Search engines are now sorting through an enormous increase in competent, near identical material on every subject, and the signals that separate one page from another are increasingly the ones volume cannot fake: whether anybody links to it, whether people who arrive stay, whether the site shows a consistent body of work on the topic rather than one page.

A cluster of ten genuinely distinct pieces that reference each other outperforms fifty generic ones, and it is considerably less work. The generic fifty compete against everything ever written on the subject. The distinct ten compete against almost nothing, because nobody else has your project history.

The test before publishing

One question, and it is brutal: could a competent competitor have published this exact piece.

If yes, it does not go out. Something in it has to be true because of your specific experience, and if you cannot point to that thing in a sentence, the piece is not finished.

A useful second check: would you send this to a prospect three days before a meeting. Most published content fails that test, which is a strange thing to be true of material produced by a business to win work.

Fewer, better, and specific

The businesses winning at this are not publishing more than the average. Several are publishing considerably less. What they have is a small number of pieces that could only have come from them, which get sent, forwarded and remembered.

That is a far better use of AI than producing four times the volume of things nobody finishes reading.

Sources

• Stanford Institute for Human-Centered Artificial Intelligence, Artificial Intelligence Index Report 2025. “78% of organizations reported using AI in 2024, up from 55% the year before.”

AI Optimize builds content systems that start from your material rather than a prompt, which is why what comes out sounds like you and not like everybody else. That work sits under Organic Content Engine.