Bottom line: Small businesses aren't losing their voice because they use AI, they're losing it because they feed AI the same three lazy prompts everyone else does. The fix isn't ditching the tools, it's putting your own specific facts, numbers, and opinions into them before you hit generate, and building a small "voice bank" you reuse every time.
The email that made me stop and reread my own inbox
Last autumn I got a cold outreach email from a marketing agency pitching "AI-powered content solutions." Nice enough email. Clean structure. Three bullet points about pain points, one about ROI, a soft close asking for 15 minutes. I nearly replied. Then I scrolled up in my inbox and found an almost identical email from a completely different company, sent nine days earlier. Same rhythm. Same bullet points. Same "I'll keep this brief" opener. Two competitors, same tool, same prompt, same result.
That's when it clicked for me (well, I'm not allowed to use that word, so let's just say it clicked hard): the problem isn't that AI writes badly. It writes fine. The problem is that thousands of small businesses are typing near-identical prompts into near-identical tools and publishing whatever comes out first. When everyone's shortcut is the same shortcut, everyone arrives at the same destination.
I did this to myself in 2025, so I can't be smug about it
I ran an experiment on my own LinkedIn for about six weeks in early 2025. I used ChatGPT to draft posts from rough voice notes, the way a lot of consultants do now. Fast, easy, saved me maybe 40 minutes a post. But my engagement dropped by roughly a third over that stretch, and a client I've worked with for years messaged me directly and said "this doesn't sound like you." She was right. The posts were competent and completely forgettable. No sharp opinions, no specific numbers from my actual client work, no British bluntness, just smoothed-out LinkedIn oatmeal.
I went back through those six weeks of posts and counted: 22 posts, and not one of them named a real client outcome, a real number, or a real disagreement with conventional advice. That's the tell. Generic AI output isn't wrong, it's just unwilling to commit to anything specific, because the model has no specifics of its own to draw on unless you hand them over.
The uncomfortable bit most people skip past
Here's the part that's a bit hard to hear if you've been leaning on AI to save time on marketing: the sameness isn't the AI's fault, and it isn't really a "prompting" problem either, not fundamentally. It's a laziness problem, and it's a cheaper-and-easier problem. Typing "write me a LinkedIn post about the benefits of AI for small business owners" takes eight seconds. Typing the same prompt but adding your actual client's name, their actual before-and-after numbers, and your actual opinion on why the popular advice is wrong takes four or five minutes. Most business owners, under deadline pressure, choose the eight seconds. Every single time. That's not an AI problem. That's a discipline problem, and it existed long before ChatGPT, we just used to call it "copying the competitor's brochure."
The businesses I've watched read this correctly, ones like Airbnb and Away, built brand voices around one very specific thing they'd say that nobody else in their category would say. Airbnb leaned into "belong anywhere," a phrase that's specific to their exact worldview about travel. Away built an entire content strategy around the discomfort of travel, not just the shiny suitcase. Neither of those came from a default prompt. They came from a founder or a strategist deciding on a stance and refusing to soften it.
What "specific" looks like in a prompt
I want to give you the actual before-and-after rather than just tell you to "be more specific," because that phrase is useless on its own.
The generic prompt (what most people type): "Write a LinkedIn post about why small businesses should use AI for customer service."
What comes out: A tidy post about how AI chatbots save time, improve response rates, and free up staff for higher-value work. True. Useless. Could have been written by any of your competitors, and probably was, that week.
The specific version: "Write a LinkedIn post in my voice: blunt, British, first person. Last month a client of mine, a 6-person accountancy firm in Leeds, cut their average email response time from 14 hours to 40 minutes using an AI-drafted reply system, but only after we deleted three canned phrases their old templates used constantly ('Please don't hesitate to contact us' being the worst one). Make the point that speed isn't the win, specificity is. End with a slightly blunt one-liner."
Same tool. Same model. Completely different output, because you gave it a real business, a real number, a real detail (the phrase they cut), and a real opinion (speed isn't the win). That's the entire difference between AI content that reads like everyone else's and AI content that reads like yours.
The five-part voice bank that fixes this
I now keep a running document, I update it maybe once a month, with five things in it that I paste into any AI tool before I ask it to write anything customer-facing:
- Three phrases I always use ("weekly experiment," "let's be blunt about this," specific turns of phrase that are mine)
- Three phrases I've banned (for me it's "unlock your potential," "in today's fast-paced world," " experience," anything that sounds like it was written for a template)
- Two real client numbers from the last 90 days (updated regularly so they don't go stale, a 40% open rate improvement, a specific pound figure a client saved)
- One opinion I hold that goes against common advice (mine right now: most small businesses don't need a content calendar, they need three good stories they keep retelling in different formats)
- A one-line description of who I'm annoyed at this week (agencies charging £3,000 a month for a chatbot that took two days to build, say)
That last one sounds odd but it's the single biggest lever for tone. Writing with a specific mild irritation in mind stops the output from being polite mush. Alex Hormozi talks about this same principle in his own content, if you've read the business lessons from Alex Hormozi, one of his recurring points is that vague claims don't sell and specific, almost uncomfortable numbers do. "We helped businesses grow" convinces nobody. "We took a gym from 80 members to 310 in five months for £4,200 in ad spend" convinces people, because it's too specific to have been made up by a template.
Why B2B businesses have it even worse right now
If you're in B2B, particularly SaaS or services, the sameness problem is sharper because so many companies in that space were early, heavy adopters of AI content tools. I've read through comparison pages, "best of" listicles, and blog intros from a dozen SaaS competitors in one sitting and struggled to tell which company wrote which paragraph. Compare that with a company like Ahrefs, whose marketing strategy leans hard on publishing their own original data, real screenshots from their own tool, actual numbers from actual crawls, rather than paraphrased advice. That's expensive to produce because it requires genuine research, but it's un-copyable, which is exactly the point. AI can rewrite an opinion in nine seconds. AI cannot invent your own first-party data.
That's a useful filter, if you can't picture a competitor generating the exact same asset with the exact same AI tool in under two minutes, you're probably safe. If you can picture it easily, that's the content to stop publishing.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
The format fix, not just the wording fix
Wording is half the battle. The other half is that most small businesses use AI to produce the same three formats everyone else is producing: the LinkedIn carousel, the "5 tips" blog post, the generic email newsletter. Even with perfect, specific wording, the format itself has become wallpaper. One of the more reliable ways to break out of that isn't more words at all, it's building something interactive that a competitor can't just regenerate from a prompt, something like the interactive calculators I've written about before. A pricing calculator, a "how much is this costing you" tool, an ROI estimator specific to your niche, these take actual build time, which is exactly why almost nobody's competitor has bothered to copy them yet. AI can write you a paragraph about the value of budgeting. It can't (yet, cheaply) build you a tool your prospect plays with for four minutes and then screenshots to their business partner.
A simple weekly process that keeps you from drifting back into sameness
Once a week, before you write anything AI-assisted, spend ten minutes on this:
- Write down one specific thing that happened in your business this week, a number, a client win, a mistake, a refund, a comment someone made
- Write down one opinion you hold that a well-known voice in your industry would disagree with
- Paste both into your AI tool along with your voice bank before asking for any drafts
- Read the output out loud, if it sounds like something a stranger could have written about a different business in your niche, redo it with a sharper detail
That's it. It's not complicated, but it is a discipline, and disciplines are exactly the thing that gets dropped first when you're busy, which is precisely why so few businesses are doing it and why it still works as a differentiator.
When it's worth getting outside help
If you've read all of the above and thought "I don't have time to build a voice bank, audit my content, and retrain how my team prompts AI on top of everything else I'm running," that's fair, and it's usually the point where bringing in outside help earns its cost. A decent AI implementation coach will sit with you for a few hours, pull the actual specifics out of your business that you're too close to notice, and build the prompts and templates around them so your team isn't starting from a blank, generic default every single time. That's usually a few hundred to a couple of thousand pounds well spent versus months of publishing content that quietly blends into everyone else's.
What this costs you if you ignore it
I'll put a number on the vague fear here rather than leave it abstract. In that six-week experiment I ran on myself, engagement dropped by around a third, and two prospects who'd normally book a call after a strong post simply didn't respond at all that period. Extend that across a year of content and you're not looking at a mild dip, you're looking at a meaningfully smaller pipeline, built entirely on content that was technically fine and strategically invisible. The fix cost me nothing but ten minutes a week and the mild discomfort of writing down an unpopular opinion before breakfast.
Frequently asked questions
Is it AI's fault that so much small business content sounds the same?
Not really. The tools are capable of highly specific, distinctive writing. The sameness comes from business owners typing generic prompts without adding their own real numbers, opinions, or client details, so the model defaults to safe, average phrasing that could apply to anyone.
How much time does it take to fix this?
Building a basic voice bank (banned phrases, real client numbers, one strong opinion) takes about 30 to 40 minutes once. After that, adding fresh specifics before each AI draft takes roughly ten minutes a week, which is far less time than most people assume.
Can competitors just copy my specific prompts and numbers once they see my content?
They can copy the format, but not the underlying facts, your actual client results, your actual opinions, and your actual voice quirks aren't things a competitor can paste into their own AI tool and reproduce. That's exactly why leaning on real specifics works as a defence, not just a style choice.
Should small businesses stop using AI for content entirely to avoid sounding generic?
No, that throws away real time savings for no good reason. The issue isn't the tool, it's feeding it nothing but a generic prompt. Businesses that add specific numbers, named clients, and clear opinions before generating still save time and end up with content that reads distinctly like them.
Want this done for you? See whether to hire an AI consultant or an agency.
Related reading: Using AI to Chase Unpaid Invoices Without Sounding Like a Debt Collector and AI Tool Costs Explained for Business Owners in 2026.
