Can AI write your marketing content?

The capability question is largely settled and it is the wrong question. A model can produce competent, fluent marketing copy quickly. What follows from that is not that everyone should, but that the value of competent fluent marketing copy has collapsed — because if it can be generated on demand, an answer engine can generate it directly and has no reason to send anyone to your version of it.

That is not a prediction. Scaled sets of near-identical pages lost a very large share of their rankings in the March 2026 core update, and the property penalised was interchangeability rather than volume or authorship. Generated content is not penalised for being generated; it is penalised when it is indistinguishable from everything else, which is the default outcome when the input was a topic and nothing else.

Which points at the actual division of labour. What a model cannot produce is what only you know: what you charge, what you observed across two hundred jobs, why a common assumption in your trade is wrong, what happened when a customer tried the obvious approach, the specific local constraint that changes the answer in your area. Those inputs do not exist in training data because they were never written down by anyone. Supplied to a model, they can be turned into a page quickly and well. Not supplied, no amount of prompting recovers them.

So the useful arrangement is generation as a drafting instrument over your material rather than as a source of material. Give it the transcript of how you actually explain something to a customer, the notes from a job, the figures from your own records, and ask it to structure and tighten. That produces something no competitor can reproduce, arrives faster than writing from scratch, and preserves the only thing that makes the page worth publishing.

There is a second failure worth naming because it is subtler than duplication. Generated text about a business's own operations tends to be confidently generic in a way practitioners notice immediately — it describes what a business like yours would typically do rather than what you do. To a prospective customer comparing suppliers, that reads as a business that has not thought about it, which is worse than a short plain page. The tell is usually specificity: real operations have odd details and generated descriptions do not.

The honest summary: use it, and change what you feed it. A page that could have been written without knowing anything about your business was not worth publishing when a person wrote it either. The difference now is that the cost of producing that page has fallen to nearly zero, so the supply is unlimited and the value is correspondingly gone.