AI Has Made Content Easier to Produce. Being Worth Reading Is a Different Problem.

We use AI a lot at Marketing 101, probably more than some people would expect from an agency that talks so much about keeping marketing human. There is no contradiction in that. If a tool can help us analyse an hour-long meeting transcript in minutes, organise research, interrogate an idea or turn one strong piece of content into useful material for another channel, we are going to use it.

There is very little value in spending three hours doing something well that technology can help us do well in thirty minutes. The more interesting question is what we do with the time we get back, because AI has made one part of content marketing dramatically easier: producing words. Producing something that is worth reading, however, is still a much harder job.

Average content has become incredibly cheap

A few years ago, creating a reasonably competent 800-word article needed time, writing ability or budget. Today, almost anybody can open ChatGPT, Claude, Gemini or one of the many other tools available, give it a subject and have a perfectly presentable article before they have finished their coffee.

The output will usually be grammatically sound. It will probably have a tidy introduction, sensible headings, a few useful points and a conclusion that wraps everything up neatly. There is nothing inherently wrong with that, but the bar has moved. If everybody can produce competent copy cheaply, competence on its own becomes much less valuable.

We see the effect of this everywhere. LinkedIn is full of posts that are perfectly readable but strangely interchangeable. Company blogs answer questions accurately enough, yet tell you almost nothing about the company behind them. The problem is often not bad writing. It is the absence of anything distinctive, specific or hard-earned.

That matters because the job of content marketing has never really been to fill a space on a website or maintain a posting schedule. It is there to help people understand what you know, how you think and why they should pay attention to you rather than someone else.

The best material rarely starts life as content

A lot of the strongest content we work on does not begin with somebody saying, "I've got a great idea for a blog." More often, it appears halfway through a client meeting when someone talks about a problem customers keep bringing to them, a sales director mentions an objection they have heard several times that month or a founder explains why they disagree with the usual way their industry handles something.

Sometimes a technical specialist spends five minutes explaining a complicated subject in a way that suddenly makes complete sense, and then the meeting moves on. Those moments are often more valuable than a list of content ideas generated from scratch because they already contain experience, context and a point of view.

We regularly work with meeting transcripts containing an hour or more of conversation. AI is very good at helping us search that material, group themes and identify moments worth returning to, but we would never simply ask it to "turn this transcript into ten social posts" and publish whatever comes back.

The useful part still needs judgement. Which comment actually matters? What fits the client's wider marketing strategy? Is there enough evidence behind the point? Does it tell us something useful about what the business knows or believes? Would the audience care?

AI helps us get through the material faster. It does not remove the need to know what good looks like.

Your AI should know more about your business than one prompt

One of the biggest mistakes businesses make with AI is judging its ability from an almost empty chat window. They type "write me a LinkedIn post about leadership" and then wonder why the result sounds like every other LinkedIn post about leadership.

If we brought a new content writer into Marketing 101, we would not give them one sentence about a client and expect them to understand the brand immediately. They would need context. They would need to know who the business sells to, what it wants to be known for, which services matter commercially, how its audience talks and which subjects need more care than others.

They would also need to understand what the company sounds like when it communicates well. That includes the language it prefers, how strongly it tends to express an opinion and the difference between sounding confident and sounding overblown.

AI needs a version of that briefing too.

For some of our clients, the guidance around voice becomes quite detailed over time. We might know that somebody dislikes certain marketing phrases, prefers direct language, explains one subject carefully but is far more opinionated about another, or has very clear views about terminology. Sometimes we have specific sentence structures we want to avoid because they repeatedly make the writing feel artificial.

British English matters. So does knowing that a particular founder would never describe something as "game-changing", regardless of how enthusiastic they were about it.

That context builds gradually. Every piece of work teaches us a little more about what sounds right, and that is exactly how we want an AI-assisted content process to develop as well.

"Professional, friendly and approachable" is not much of a brief

This is one area where traditional brand guidelines can become less useful when you start applying them to AI. Lots of businesses describe their tone of voice as professional, friendly and approachable. Those words sound sensible, but on their own they tell a writing tool almost nothing.

Professional for an employment lawyer is different from professional for an independent restaurant. Friendly in financial services is unlikely to sound the same as friendly in retail. The adjectives are not wrong, they are simply too broad to guide the writing in any meaningful way.

We get much better results by describing behaviour. Use ordinary language when an ordinary word will do. Explain technical terms rather than assuming everybody understands them. Get to the useful point quickly. Avoid exaggeration. Do not turn every paragraph into a slogan. Use examples where we have them and be confident without pretending the evidence tells us more than it does.

Those instructions give AI something practical to work with. Strong examples are even more useful. Previous articles, emails, presentation transcripts, good LinkedIn posts and website copy can all help establish what the business actually sounds like rather than what a brand guideline says it should sound like.

We also find rejected examples useful. Telling the AI "this sounds too corporate", "we would never use that phrase" or "you have softened the point so much that it no longer says what I meant" helps define the boundaries of the voice. In practice, teaching an AI what to avoid is often just as useful as telling it what you want.

We do not ask AI to "make it sound human"

"Make this sound more human" is one of the least useful instructions you can give an AI because human is not a tone of voice. Left to interpret the instruction itself, the tool often reaches for things it associates with informal writing, such as contractions, rhetorical questions, punchier sentences or conversational phrases.

Sometimes that helps. Other times it simply swaps one kind of artificial writing for another.

We prefer to diagnose the actual problem. We might ask the AI to identify claims that could apply to almost any business, find repeated sentence structures, flag corporate language or show us where the conclusion is simply repeating something we have already covered. We might ask it to compare a draft against established brand examples and highlight where the voice shifts, or show us where the original opinion has become weaker during the drafting process.

That keeps the editorial decision with the person. AI is very good at fixing a defined problem, but "make this better" is not a defined problem.

The human needs to be there before the first draft

There is plenty of discussion around keeping a human in the loop when using AI, and we agree with the principle. Where we differ is in how early that human needs to appear.

If AI chooses the subject, develops the argument, invents the examples and writes the piece, having somebody proofread it for two minutes before publication does not make the process especially human-led. The useful raw material should come from somewhere real in the first place.

That could be an experience, a customer conversation, something learned through doing the work, original research or a point of disagreement. It might be a result, a mistake or a pattern the business has noticed before somebody else has put it into words.

AI can then help explore the thinking. This is one of our favourite uses for it.

Instead of asking an expert to sit in front of a blank document and write 1,000 words, we can give AI the beginning of an idea and ask it to interview them. Why do you think that? What have you seen? Can you give an example? When does that advice not apply? What changed your mind?

After a few answers, there is often far more useful material than there would have been from a cold request for a finished article. The expert has supplied the thinking and the AI has helped draw it out. Neither has been asked to do the other's job.

The real opportunity is making expertise work harder

This is where we think AI becomes genuinely useful for businesses.

A good customer conversation does not necessarily need to disappear when the call ends. A useful presentation can become more than something delivered once. An internal expert who has never written a blog may still have years of knowledge worth sharing, and a detailed piece of research may contain several different ideas for different parts of the audience.

AI makes it much easier to find, organise and reuse that material. It can help a business get more value from the knowledge it already has, but only if the source remains visible.

The risk is not that everybody starts using AI, because that is already happening. The risk is that businesses use the same tools, ask the same generic questions and gradually remove the things that made their marketing recognisable in the first place.

I do not think businesses need to be frightened of AI-generated content. I think they need to be much more demanding of it.

At Marketing 101, we want AI doing the work that saves time and helps us see more in the material we already have. We want people doing the work that depends on judgement, curiosity, experience and a point of view.

That division makes sense to us, and it is the one we think will continue to matter as the technology gets better.

Want to see how we stress test AI-written content?

We have created a free Humaniser Prompt based on the checks we use when reviewing AI-assisted content. It looks for repetitive structures, unnecessary filler, generic claims, common AI language and some of the smaller writing habits that creep in when the first draft goes unchallenged.

It is not intended to replace your own brand voice rules, and we would not recommend blindly following every instruction in it. Its value is in giving you a more useful editing framework than simply asking AI to "make this sound more human".

You can download the Marketing 101 Humaniser Prompt here: https://mailchi.mp/marketing-101/rvxe8vsxij

Use it as a starting point, then make it specific to your business. Add the phrases you would never use, the examples that sound right and the patterns your AI keeps getting wrong. Over time, that level of detail is what helps stop your content sounding like everybody else's.