Why AI Content Sounds Generic (and the Rule That Fixes It)

Try this. Type "Write a LinkedIn post about why small businesses need a bookkeeper" into any AI tool. Read what comes back. Now have a friend in another city type the same line into a different tool.

Put the two drafts side by side. Same opening. Same three points. Same soft question at the end. "What's your biggest bookkeeping challenge? Let me know in the comments!"

That's not a coincidence, and it's not because the tools are dumb. It's the most common reason AI content sounds generic, and almost nobody names it correctly. People blame the model. The problem is what they handed it.

Here's the rule I build everything else on. AI can arrange what you give it. It cannot invent your experience. No Raw Material, no human content. I call it the Raw Material Rule, and once you see it, you can't unsee it.

Why AI content sounds generic: it fills every gap with the average

A language model writes by predicting the most likely next word, given everything in front of it. When you give it one line, almost nothing is in front of it. So it reaches for the most likely post about bookkeeping ever written. Not your post. The average one, sitting in the middle of thousands of articles and ads on the same topic.

Every question your prompt leaves open is a gap. Who's reading? Why now? What do you believe? What happened to you last Tuesday? AI doesn't stop and ask. It fills each gap with the most common answer it has seen. One gap filled with the average is fine. Twenty gaps filled with the average is a post that sounds like everyone else's.

Most people think this means the tool isn't smart enough. That's backwards. The tool did exactly what you asked. You asked for a post about bookkeeping with no audience, no reason, no story and no opinion. You got a post about bookkeeping with no audience, no reason, no story and no opinion.

Think about how you'd brief a freelancer. You'd never send a new copywriter one sentence and expect something usable. You'd tell them who the client is. You'd forward an old email so they could hear how you talk. You'd tell them the story about the client who almost lost their house over a tax bill. A good freelancer would ask questions until they had enough. AI usually won't. So you hand it everything up front.

You are the source. AI is the arranger.

The Raw Material Rule flips how most people use AI. Most people treat the AI as the source of ideas and themselves as the editor. Ask for a post, then fix it.

Flip it. You're the source. AI is the arranger. It's very good at taking a messy pile of your notes, stories and opinions and turning them into clean paragraphs in a sensible order. It's useless at producing the pile.

Why? Your experience isn't in the training data. The model has read millions of posts about bookkeeping. It has never sat across from a client who opened a shoebox of crumpled receipts and started crying. It doesn't know your price, your best client result, the mistake you made in year two, or what you say to every new customer on the first call. Those are the details that make a reader think, "This person actually does this work." They only come from you.

Raw Material comes in five forms:

• Facts: what's true about your business, your offer, your process.
• Stories: specific things that happened, with a person, a moment and an outcome.
• Opinions: what you believe that others in your field might disagree with.
• Examples: concrete cases that show your point in action.
• Numbers: prices, percentages, time saved, dollars found, client counts.

A decent prompt has at least one of each. A great one has more than you need, so the AI has something to choose from.

How to get the Raw Material out of your head in three minutes

The reason people skip this isn't laziness. Staring at a blank box labeled "stories" is hard. Your best material is in your head, and it doesn't come out on command. So don't write it. Talk it.

The voice memo. Open the recorder on your phone. Talk for three minutes about the topic as if a client just asked you over coffee. Don't plan. Don't fix sentences. Then transcribe it with any free tool. People who say "I don't have any stories" usually have five. They tell them out loud all the time.

The 5-question interview. Answer these five questions about your topic, in your own words:

1. What happened?
2. What did you learn?
3. What do most people get wrong?
4. What's your opinion?
5. What's the proof, or the number?

Each question pulls out a different kind of Raw Material. Here's the difference between a weak answer and a strong one, using a bookkeeper writing about tax season as the example:

Question	Weak answer	Strong answer
What happened?	Clients often struggle at tax time.	In March, a roofer brought me eight months of receipts in two grocery bags. We found $4,200 in duplicate vendor charges.
What do most people get wrong?	They don't plan ahead.	They think tax prep is a spring job. It's a monthly habit. Spring is just when the bill for skipping it arrives.
What's your opinion?	Bookkeeping is valuable.	Shoebox receipts aren't a system. If you can't find last month's numbers in five minutes, you don't have books. You have a pile.

Look at the weak column. That's exactly what AI writes on its own. The strong column is what it can't write without you.

The reply mine. Search your sent email for the topic. Find the last three times a client asked about it and you answered in writing. Those replies are Raw Material, already in your voice. This works especially well for FAQs and sales pages, because the questions real buyers ask are the questions your page has to answer.

What it looks like in practice

Here's an example. Picture Marco, who runs a landscaping company and needs a blog post on fall lawn care. His first pass at Raw Material is one line: "Fall is a good time for lawn care." Hand AI that, and you'll get the same post as every lawn site on the internet.

So he records a voice memo in his truck between jobs. Three minutes in, he's said that most homeowners stop mowing in September, which is a mistake, because the grass is still growing. That he aerates every client's lawn in the first two weeks of October and has for eleven years. That one client skipped fall fertilizer two years running and spent $1,100 reseeding the next spring. And that he thinks the big-box spring fertilizer bags are mostly a waste of money.

None of that was in the one-liner. All of it was in his head. Now the AI has something to arrange: a mistake, a habit, a cost, and an opinion people will argue with. That's a post someone would actually read to the end.

Don't clean the Raw Material up first. Fragments are fine. Bad grammar is fine. A note like "roofer, grocery bags, $4,200, he laughed about it later" is fine. AI will arrange it. What matters is that every detail is true and every detail is yours.

How much is enough? For a social caption, one story or opinion plus one number. For a LinkedIn post, one story, one opinion, one proof point. For a newsletter or blog post, answer all five interview questions. Quick check: if the Raw Material part of your prompt is shorter than the rest of the prompt, it's too thin.

The honest caveat: no story in, fake story out

Here's what happens when you skip this. AI doesn't leave the gap empty. It fills it with something that sounds right.

You'll see "I once worked with a client who..." followed by a person who never existed. You'll see "Studies show that 73% of small businesses..." with no study behind it. These invented stories and fake numbers are the worst thing AI can put in your content, because they don't just sound off. They're false, and they go out under your name.

So add one line to every writing prompt:

"Use only the stories, examples and numbers I provide. Do not invent any. If you need one I haven't given you, write [NEED STORY] or [NEED NUMBER] and leave it for me."

That placeholder is worth a lot. Instead of a smooth fake, you get a visible gap. A gap is honest. It shows you exactly where the piece needs you. Some of your best material will come from staring at a [NEED STORY] marker and asking, "Okay, what actually happened?"

One more limit. Your Raw Material is yours to use. Other people's lives aren't. If a client story could identify the client, ask first, change the details, or keep it general.

Before you ask AI to write anything, ask yourself one question: what am I giving it that nobody else could? If the answer is nothing, go get your Raw Material first. That's the part only you can do.

Key takeaways

• AI fills every gap in your prompt with the most common answer, which is why one-line prompts sound like everyone else.
• The Raw Material Rule: AI can arrange what you give it; it cannot invent your experience.
• Raw Material comes in five forms: facts, stories, opinions, examples and numbers.
• Talk it out instead of writing it: a three-minute voice memo or the 5-question interview pulls it out fast.
• Tell AI to mark missing details as [NEED STORY] or [NEED NUMBER] instead of inventing them.

Frequently asked questions

Why does AI-generated content sound so generic?

AI-generated content sounds generic because a language model predicts the most likely next word, and every detail a prompt leaves out gets filled with the most common answer. A one-line prompt leaves dozens of gaps: audience, purpose, opinion, stories, numbers. Filled with averages, the result reads like everyone else's post. Giving the AI your own facts, stories, opinions and numbers closes those gaps.

What is the Raw Material Rule for AI writing?

The Raw Material Rule says AI can arrange what you give it, but it cannot invent your experience. Raw Material means the facts, stories, opinions, examples and numbers only you can supply. Without it, AI either writes generic content or makes details up. With it, AI does what it's good at: turning a messy pile of your notes into clear, well-ordered paragraphs.

How do I stop ChatGPT from making up statistics and stories?

Add an explicit instruction to every writing prompt: use only the stories, examples and numbers I provide, do not invent any, and if one is needed that I haven't given, write [NEED STORY] or [NEED NUMBER] instead. This turns invented filler into visible gaps you can fill with real details. Still check every number and claim before you publish.

How much personal detail should I give AI before it writes?

Match it to the length of the piece. A social caption needs one story or opinion and one number. A LinkedIn post needs one story, one opinion and one proof point. A newsletter or blog post needs answers to five questions: what happened, what you learned, what most people get wrong, your opinion, and your proof. If that section is shorter than the rest of your prompt, it's too thin.

What's the fastest way to come up with stories for AI content?

Talk instead of writing. Record a two- to three-minute voice memo about the topic as if a client asked you about it over coffee, then transcribe it with any free tool. You can also search your sent email for past answers to client questions on the topic. Both methods produce real stories and opinions already in your own voice.

Where this goes next

Raw Material is the heart of a Voice Brief, the instruction packet you hand AI before it writes. My class Create AI Content That Sounds Human covers the full brief in Session 2, "Brief It Like a Human."

See the course