Anything a language model can write, every competitor can write. Differentiation comes from what models cannot reach: your customers’ actual words, your own data and scars, and the live conversation in your market.
There is a simple reason so much B2B content now sounds the same. The large language models that write a growing share of it were trained on broadly the same public internet. When the inputs are similar, the outputs converge. Ask five competitors’ tools to write about your category and you will get five competent, fluent, nearly interchangeable articles.
The implication is not that AI is useless for content. It is that the advantage has moved. It no longer sits in the ability to produce words. It sits in what the words are built from — the context a model is given. Better context, better output.
The Three Voices are the three sources of context that no model has on its own. Every piece of genuinely differentiated content draws on at least one of them.
1. The voice of the customer
What buyers actually say. Not what you think they care about, not your positioning statement, but their words: how they describe the problem, what they objected to, what nearly stopped them, and what finally tipped them.
This is the most underused source of material in most companies, and it is sitting in plain sight:
- Sales calls — especially the first call, when the buyer describes the problem in their own terms before they have learned your vocabulary.
- Support tickets and onboarding questions — where the real friction lives.
- Reviews — yours and your competitors’. Competitor reviews are a free list of what your market dislikes about the alternatives.
- Win and loss interviews — a short conversation with a recent win and a recent loss will often teach you more than a quarter of analytics.
The value is that customers describe their problem in words other customers recognise instantly. People believe you understand them when you say it the way they say it.
2. The voice of the company
Your own data, opinions and scars. The things only you could publish because only you lived them:
- Results from your own work, with the numbers attached
- Teardowns of what went wrong and what you changed
- Opinions you are prepared to put a name to, including unpopular ones
- What you have learned to stop doing
A useful test: if a competitor could publish this piece verbatim, with their logo on it, it is not in your voice. It might be accurate and well written. It is not differentiating you from anything.
3. The voice of the market
The live conversation around your category: what is being argued about in communities and on social platforms, what analysts are saying, what competitors are launching, what regulation is changing, what your buyers are worried about this month.
This voice supplies timing and relevance. Content that responds to what the market is already discussing gets read, because it arrives when attention is already there. Content produced from a calendar alone is always answering last month’s question.
It is targeting data, not just content
The same work pays off twice. Customer-voice mining that produces a good article also produces:
- Sharper ad copy, written in the words buyers actually search and respond to
- Better keywords, because buyers describe symptoms in terms you would never have guessed
- A tighter ideal customer profile, because patterns in who describes the problem most acutely tell you who feels it most
- Better sales enablement, because objections collected systematically can be answered systematically
That is why this sits in the demand and attention part of The System rather than being filed under “content marketing”. It is market research that happens to produce content as a by-product.
A monthly voice harvest
You do not need a research budget to start. A simple monthly routine is enough:
- Customer: pull ten recent call notes or recordings and copy out every phrase a buyer used to describe their problem. Keep them verbatim.
- Company: write down one result, one mistake and one opinion from the month. Each is a potential piece.
- Market: note the three conversations your buyers were having in public this month, and whether you had anything to say.
- Use it: brief every piece of content, and every new ad, from that document — including anything a model helps you draft.
That last step is the point. AI tools become far more useful once you feed them this material. The model does the miles; the voices supply the part nobody else has. The same idea is covered in machines do the miles, humans do the last one.
Where to go next
Educate to differentiate explains how to turn company voice into teaching that buyers value. The ICP targeting guide shows how customer voice sharpens targeting. And Sold on Social is the practice built around this framework.