The short version: AI writing tools don’t make businesses sound generic, they make lazy briefs sound generic faster. The fix isn’t “sound more human,” it’s feeding the tool specific facts, numbers, and stories that only your business has. Do that and the sameness problem mostly disappears.
The morning I found three identical business descriptions
Back in March I was helping a client audit her competitors before we rebuilt her homepage. She runs a small bakery in Tunbridge Wells. I pulled up her site, then two rivals within a five mile radius. All three “About” sections opened with a version of the same sentence: “We’re passionate about bringing you the finest artisan bakes, crafted with love and the freshest local ingredients.”
Three different bakeries. Three different owners. One sentence, more or less. I asked each of them (politely, over email) whether they’d used ChatGPT to write it. All three said yes.
That’s the bit people miss when they complain that AI content sounds fake. It doesn’t sound fake because a machine wrote it. It sounds fake because the person typing the prompt gave the machine nothing to work with. “Write me an About section for my bakery” produces the statistical average of every bakery About section on the internet, because that’s exactly what a language model is built to do: predict the most probable next words. Average in, average out.
Why this happens (the bit worth understanding, not just accepting)
Large language models are trained to produce the most likely, most broadly acceptable response to a prompt. That’s not a flaw, it’s the design. It’s the same reason autocomplete on your phone suggests “how are you” and not “how’s the gout.” The model is chasing the middle of the distribution, and the middle of the distribution for “bakery description” is warm, safe, and completely interchangeable.
Layer on top of that the fact that most business owners are time poor and typing prompts on their phone between school runs. They’re not going to write three paragraphs of context. They’re going to type six words and hit enter. The tool does exactly what it was asked. The result is bland because the input was bland, not because the technology is incapable of anything else.
I tested this in September this year. I gave the same tool (“write me a homepage intro for a small bakery”) to five different fictional bakery briefs, changing nothing but the location. Four out of five came back with some version of “passionate,” “crafted,” and “love” in the first sentence. When I ran the same prompt but added real detail, an actual oven brand, the name of the street market they sell at on Saturdays, the fact the owner trained under a pastry chef in Lyon, the results were unrecognisable from each other. Same tool. Same model. Wildly different output, because the input carried facts nobody else has.
The uncomfortable part nobody selling AI courses wants to say out loud
Here’s the bit that gets skipped in most “how to fix AI content” posts: a lot of small business owners prefer the generic version. It feels safer. It sounds “professional.” Standing out means risking sounding unprofessional, or too casual, or like you’re bragging. So even when I hand a client a rewrite with specific numbers and a real story in it, some of them quietly soften it back towards the average, because the average feels less exposed.
I’ve had this conversation more than once with clients who work with me through my AI consulting work. I’ll rewrite their homepage to say something like “we’ve fitted 340 kitchens across Kent since 2019, average job takes 11 days,” and they’ll come back wanting to change it to “we pride ourselves on quality and customer satisfaction.” Technically that sentence is true of every kitchen fitter in the country. That’s the problem. If a sentence is true of every competitor, it’s not marketing, it’s decoration.
So the sameness problem isn’t purely a tools problem. It’s partly a nerve problem. AI just made the cowardice visible at scale, because now everyone’s cowardice is generated by the same three models.
The five specifics that break the pattern
When I sit down to fix this for a client, I’m not trying to make the writing sound “more human” in some vague way. I’m trying to get five categories of specific detail into the brief before the AI tool ever gets touched:
- A real number. Not “years of experience,” an actual figure: “since 2014,” “127 weddings catered,” “average response time 40 minutes.” Numbers are the fastest way to stop a sentence sounding like it could belong to anyone.
- A named place, tool, or supplier. “Crafted with local ingredients” is nothing. “We buy our flour from Shipton Mill in Gloucestershire” is a sentence only that bakery can write.
- A specific customer story, with a detail that’s slightly odd. Not “our customers love us.” Something like “a customer once drove from Brighton at 6am because we were the only bakery still doing his late father’s favourite seeded loaf.” Odd, specific detail signals truth in a way polished claims never do.
- A number that’s slightly unflattering. “We sell out of sourdough by 10am most Saturdays, so ring ahead” tells a customer more about the quality of your bread than any adjective could, and it’s the kind of honest limitation an AI tool will never invent on its own because it’s optimising for reassurance, not truth.
- A named person, not a job title. “Our team” is nobody. “Maria, who’s run the ovens here for eleven years” is somebody.
Feed those five things into the brief before you touch a generative tool and the output changes completely, because now the model has specific material to rearrange instead of a vacuum to fill with averages.
A worked example, step by step
Here’s how I did this for the bakery client, start to finish, over about ninety minutes:
- Step 1. I interviewed her for twenty minutes with no laptop open, just asking questions: where does your flour come from, what’s the oldest piece of equipment in the kitchen, what’s a customer said to you that stuck with you, what do you run out of first.
- Step 2. I wrote down her answers in her own words, badly punctuated, exactly as she said them. This raw material is worth more than any polished paragraph, because it’s the only part of the process that’s hers.
- Step 3. I fed those raw notes into the AI tool with an instruction to keep every specific fact and cut nothing, only tidy the grammar and structure.
- Step 4. I read the output aloud. Anywhere it slipped back into “passionate,” “crafted,” or “finest,” I deleted the sentence entirely rather than editing it, because those words are a warning sign the model has run out of real material and started filling gaps with average.
- Step 5. I checked it against her two competitors’ sites one more time. If a sentence could have been lifted onto either of their homepages without anyone noticing, it came out.
The final version opened with: “We’ve been baking on the same corner of Camden Road since 2016, using a 1970s deck oven we bought secondhand from a bakery in Hastings that closed down. Ask for the seeded sourdough before 9am on a Saturday, it’s usually gone by half ten.” That’s not clever copywriting. It’s just true, specific, and impossible for the bakery two streets over to copy.
Where this shows up beyond the homepage
The same pattern is eating email marketing, social captions, and even proposal documents. If you’re using Mailchimp or a similar tool to send a weekly newsletter, and you’re generating the subject lines and intros with a generic prompt every week, your open rates will drift down over months without you noticing why. Subscribers stop opening emails that could be from anyone.
The businesses that have built something distinctive with AI in the mix tend to do the opposite of the bakery-average problem. Look at how Airbnb built its brand around specific host stories rather than generic travel claims, or how Figma’s marketing leans on named product details rather than vague “collaboration” talk. The pattern holds everywhere: specificity is the only thing generative tools can’t average out, because it doesn’t exist anywhere else for them to average it from.
It matters on social too. I’ve reviewed Instagram accounts for small businesses where every caption for a month followed the exact same structure: hook question, three-line description, emoji, call to action. If you want to see what breaks that mould, the piece on Instagram marketing strategy is worth a proper read, it’s built around specific product stories rather than caption templates, which is precisely why it still gets engagement when template-driven accounts have gone flat.
The engagement angle people overlook
There’s a second, quieter cost to sameness that isn’t about tone, it’s about interaction. Generic copy invites passive scrolling. Specific numbers invite a response, a comment, a click. If you want people to do something rather than just read past you, give them something to interact with rather than another paragraph to skim. I’ve had far better results turning a vague “get a quote” page into something with an actual interactive element, the piece on interactive calculators for converting visitors covers this well: a calculator that tells someone “your kitchen refit will likely take 9 to 14 days based on square footage” beats a paragraph claiming “efficient service” every single time, because it’s specific and it’s about them, not you.
What I’d tell you to do this week
Pull up your own homepage and your top competitor’s, side by side. Read your opening two sentences and theirs. If you could swap the company name and nobody would notice, that’s your problem, not the AI tool you used to write it. Go back through the five categories above (a number, a named supplier or place, an odd customer story, an honest limitation, a real name instead of a job title) and see how many of them are missing. Most homepages I audit are missing at least four of the five. Fixing even two usually changes how the whole page reads.
If you’re using AI tools daily and still getting flat results, the issue is very rarely the model. GPT-4, Claude, and Gemini can all produce sharp, specific writing when given sharp, specific material. The gap is almost always in the brief, which is exactly the kind of thing worth getting a second pair of eyes on if you’re trying to fix it across a whole business rather than one page at a time, which is where working with an AI implementation coach tends to save people months of trial and error rather than a few hours.
One more thing worth saying plainly: this isn’t a one-off fix. Every quarter, go back and check whether your copy has drifted back towards the average. It happens quietly, one rushed prompt at a time, the same way a garden goes back to weeds if you stop weeding it. Nobody decides to sound generic. It just creeps back in whenever you’re busy and reach for the fastest option instead of the most specific one.
Frequently asked questions
Does AI writing hurt SEO because it sounds the same as everyone else?
Search engines don’t penalise AI-written text directly, but generic content earns fewer backlinks, less time on page, and fewer shares, all of which hurt rankings indirectly. Specific, factual content performs better in search precisely because it’s the kind of thing people quote and link to.
How can I tell if my own website copy has gone generic?
Swap your company name for a direct competitor’s in your opening two paragraphs. If the sentences still make perfect sense, they’re not doing their job. Specific numbers, named suppliers, and real customer stories are the fastest fix.
Is it still worth using ChatGPT or Claude for small business copy?
Yes, but the quality of the brief matters far more than which tool you use. Feed it raw, specific facts about your business rather than a vague instruction, and edit out any sentence that reverts to words like “passionate,” “crafted,” or “finest,” which are signs the tool has run out of real material.
How much does it cost to get proper help fixing this across a whole business?
Costs vary widely depending on scope, from a single afternoon consultation to a multi-month implementation project. It’s worth reading a clear breakdown of what an AI consultant costs before you commit to anything, so you know what you’re comparing quotes against.
Related reading: The AI Notetaker Sitting In On Your Client Calls Might Be Breaking Your NDA and Why “Write Like Me” AI Prompts Fail for Small Business Content (And What Works Instead).
