You will be able to explain why default AI writing is generic and recognise its common patterns.
Farah asked an AI assistant to write the homepage for her workwear shop. She typed one line: "Write a homepage headline and intro for a modest workwear brand in Singapore." In a few seconds she had a headline, "Redefine your everyday elegance", and an intro about "timeless pieces crafted for the modern woman who effortlessly balances style and substance". It was fluent and grammatical, and it sounded exactly like the homepages of three other shops she had looked at that week. She could not find a single line in it that her customers had ever said.
This is the most common experience people have when they first use AI for business writing. The output looks finished and says nothing in particular. This lesson explains why, so you can do something about it in the rest of the module.
A language model, the kind of AI behind assistants such as ChatGPT, Claude and Gemini, is trained on a very large amount of text. When you give it a prompt, it produces text by predicting, piece by piece, what is likely to come next, given everything it has seen.
That is what makes it fluent. It is also what makes it generic. Given a vague prompt, the most likely continuation is the most typical version of that kind of text. Ask for a homepage for a workwear brand with no other information, and you get something close to the average of all the workwear homepages and fashion copy it learned from. The average of many homepages sounds like no business in particular.
The model is not being lazy. It is doing what you asked, with the information you gave it. A one-line prompt gives it almost nothing to work with except the category, so it fills the gap with what is typical for the category.
Once you have read a few AI drafts, you start to see the same habits. Learning to spot them is the first step to removing them.
Stacked adjectives: "timeless, versatile and effortlessly elegant pieces". Several adjectives in a row, none of them specific.
Lists of three: "style, comfort and confidence", "quality, care and community". Three items, usually abstract, often with a rhythm that sounds good and means little.
Dramatic contrasts: "It's not just a blouse, it's a statement", "More than tuition, a partnership". A setup and a twist that make an ordinary claim sound deep.
Claims with no specific detail: "trusted by thousands", "designed with you in mind", "exceptional quality". Statements that could be about anything and cannot be checked.
Certain favourite words: the kind that appear far more often in AI drafts than in ordinary writing, such as "redefine", "effortless" and "timeless". Farah's draft had all three in its first two sentences.
None of these patterns is wrong in a single instance. Human writers use lists of three and the occasional contrast. The problem is when they pile up, because then the reader feels the text was not written by anyone, and stops trusting it.
An AI assistant knows a great deal about the world in general. It knows nothing about your business unless you tell it. It does not know your customers, the words they use, your prices, your delivery times, your reviews, what makes you different from the shop down the road, or what went wrong last month.
So when you ask it to write about your business without that information, it has two choices: stay vague, or make something up. It usually stays vague, and when it does invent, that is worse.
This is why the voice-of-customer file and the benefit and proof sheet you built earlier in this course matter so much here. They are exactly the information an AI assistant lacks. Lesson 8.2 shows how to give it to the AI.
Language models can produce statements that are false, stated with the same confidence as statements that are true. They might give a wrong date, invent a statistic, misstate a rule, attribute a quote to someone who never said it, or describe a feature your product does not have. The text reads smoothly either way, so the error is easy to miss.
For business copy, this matters more than most uses. A wrong number on your landing page is a claim you have made to customers, and as lesson 3.3 explained, a misleading claim is your responsibility whoever drafted it. Every number, name, date, rule and claim in an AI draft needs checking against a source you trust, such as your own records or an official website, before it is published.
Farah's draft said her blouses were "made from sustainably sourced fabrics". She had no idea where her fabric supplier sourced from. The AI had added the phrase because it is common in fashion copy. Had she published it, she would have made a claim she could not support.
None of this means AI is useless for writing. It means a one-line prompt gets you a generic result, and the fix is in how you brief it, where you use it, and how you edit what it gives you. Lessons 8.2 to 8.4 cover each.
For the activity, start by doing what Farah did: ask an AI assistant for a homepage headline and intro with no context at all. Then go through the result line by line, using the patterns above, and mark each generic phrase. Keep the result, because you will compare it with a properly briefed version in lesson 8.2.
Ask an AI assistant to write a homepage headline and intro for your business with no context and list every generic phrase in the result.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).