Is AI making every small business brand look the same?

By Kate Dole · Published

I use AI tools in my own work, and I'd rather say so up front. Here's what the research says about brands blurring together, and which parts of your brand I'd keep human-made.

Yes, and it’s speeding up a drift that began before ChatGPT. Researchers who compared 10,000 big-company websites found layout differences fell by over 30% after peaking around 2008–2010. AI tools push further in the same direction, because they hand everyone the likeliest answer, and in any category the likeliest answer is the average. I use these tools myself. The way to stay recognisable is to feed them what only you know and keep what people remember human-made.

Did brands look alike before AI?

They did. The sameness was well underway by the time anyone could type “make me a logo” into a chat box.

Sam Goree and his colleagues analysed nearly 200,000 screenshots of 10,000 websites, from the biggest US companies to Webby nominees. Visual differences between sites peaked around 2008–2010 and then shrank, with layout differences “declining over 30%”. Sites built on the same shared toolkits looked the most alike (Goree et al., via The Next Web, 2020). Those were large companies with large budgets, and they converged anyway.

Advertising went the same way. Orlando Wood’s book Lemon (IPA, 2019) tracks how ads became flatter and more abstract, and one review sums up the era he describes as “standardisation, specialisation, repetition and avoidance of risk” (Marketing.ie, 2020).

“Avoidance of risk” describes most of the small-business websites I open: a template with a handshake photo up top, a lot of blue, and a sentence about quality service you can trust. Every one of those choices felt safe to the person who made it, and safe is invisible. AI arrived in a market that was already copying itself, and it made the copying instant.

Why does AI give everyone the same answer?

Because predicting the most likely output is what it’s built to do. Ask for a logo for a landscaping company and you’ll probably get a leaf, a green and a clean sans-serif, since that’s what most landscaping logos it learned from look like.

That’s useful when you want the expected answer, like the right format for an invoice. It works against you when the whole point is to be picked out of a lineup.

The industry people who make brands for a living are saying the same thing. In the IPA’s September 2026 AI magazine, McCann’s Brandon Kaplan puts the risk in accepting the first answer AI gives you, since that answer drifts toward the average. Tyler Berry of the studio YeahNice says the result is a generic “industry house style” (IPA AI Magazine, September 2026).

Researchers are looking at it too. A March 2026 paper from Shin and colleagues examines “vibe coding”, where you describe a website to an AI and it builds one (arXiv, 2026). It’s a position paper, an argument with no measurements, and the argument goes like this: when building a site gets frictionless, design homogenization can get worse. The authors propose “productive friction”, moments built into the process that make you stop and push back on the default.

My version is shorter. AI gives you the average, and nobody remembers the average. Memory runs on what’s different, which is why Jenni Romaniuk’s research on brand assets finds that of fame and uniqueness, “uniqueness is most important” (Ehrenberg-Bass Institute). I go deeper on that in seen to be remembered.

I’d use AI for the dull parts and spend the hours it saves on the conversation that makes a brand specific.

Does AI make being distinctive worth more?

Yes, and that’s the good news for small businesses. When everyone can produce a competent, category-shaped brand in an afternoon, competent and category-shaped stops setting anyone apart.

Miles Marshall of Turner Duckworth makes this point in the same IPA piece: brands looked alike before AI, AI speeds up the work of fitting a category, and so distinctiveness becomes more valuable. He’s talking about global brands. It holds even harder at the scale of a Moncton bakery, because you’re competing for the same few seconds of a stranger’s attention on a phone screen.

Here’s what a small business has that a model can’t scrape: your town, your customers’ actual words, your face, the story of why you started, the odd detail people mention when they recommend you. Those are the raw materials of a brand people remember. A tool trained on everyone else’s brands has none of them until you hand them over.

Use it to explore, and have a person make the mark you’ll live with.

Romaniuk’s Grid sorts brand assets by how famous and how unique they are. Anything famous but common lands in a box labelled “Avoid”, because such elements “will trigger competitors to copy”. A logo generated by averaging thousands of logos in your category lands there by design. It’s familiar, and it belongs to nobody in particular, including you.

AI is handy at the start. Use it to look at a hundred directions in ten minutes, to spot the clichés in your category, or to rough out an idea you can’t sketch. Then take what you learned to someone who’ll ask about your business before drawing anything.

When I worked with Foreword Films on their tenth-anniversary rebrand, the logo came out of seven concepts across three rounds, and the concepts came out of a long first conversation. Andrew Buss, one of the founders, put it this way afterwards: “The two hours that we spent talking the first time, I didn’t know why she needs to know all this stuff. And now I realize why.” Those two hours are the input no tool had. You can see what came out of them in the Foreword Films case study.

How do I use AI and keep what makes my brand yours?

Let AI handle the forgettable work, and keep anything people will remember in human hands.

Here’s how I’d split the work.

Hand it to AIKeep it human-made
A first outline of a page or postYour first sentence and headline
Draft alt text, tidied meeting notes, a typo checkYour logo, colours and type
Sorting customer reviews to find phrases that repeatYour photos of real people in your real place
A list of the clichés in your category, so you can avoid themYour stories and opinions

A few habits make the left column work better:

  • Feed it the specific stuff. Paste in customer emails, review text, the way you explain your work out loud. The more of your own material it has, the less it fills the gaps with everyone else’s.
  • Treat the first answer as the average. Ask for ten more. The useful one is often the one that makes you slightly uncomfortable. That’s productive friction, done by hand.
  • Keep the rough edges. When an AI edit sands down the line that sounded like you, put the line back. I wrote about this in Unhidden: “it’s the rough edges and mess of being a human that’s ultimately going to help us stand out.”
  • Run the cover-the-logo test. Hide your logo and read your homepage next to two competitors’. If someone who knows you couldn’t pick yours, the tools have flattened it. The test and four others are in my guide to whether your website is outdated.

Structure is a different matter. Familiar layouts help visitors find their way, and AI is fine at producing them. Keep the structure familiar and the identity unmistakable. I explain why that split matters in how people form a first impression of a website.

What I’d do

I’d use AI for the dull parts and spend the hours it saves on the conversation that makes a brand specific.

That’s how I work. I use AI tools in my studio, and I’d sooner tell you than have you wonder. Mostly I use them to check my own work once it’s built: debugging, testing a site for security holes, and red-teaming it, which means asking a tool to attack the build and find the blind spots I can’t see from inside. I mostly run private AI models on my own machine, so client work stays with me.

The plain-language version: AI takes the slow, important checking off my plate, so I can spend those hours on the work itself, and that keeps budgets friendly. The parts people remember still come from listening to a client for two hours and noticing what lights them up: the colour, the face, the first sentence, the photo style that keeps coming back.

If you’re a small business owner short on time, start with your first sentence. Write it yourself, with a place, a person or a number in it, and resist the urge to have a tool polish it. Then look at your photos. Real ones of your real space will do more for you than anything generated, because they’re the only images on earth that belong to you.

If you want help finding the specific things worth repeating, that’s what my brand strategy work for New Brunswick businesses is for. And if you’d rather start on your own, take this line from my first newsletter: “The weirder the better in the age of AI. Our pimples make us human, friends.”

Quick answers

Can AI replace a graphic designer for a small business?

For a one-off social graphic or a quick mock-up, often yes. For the logo, colours and voice you'll repeat for years, you want a person asking you questions a tool won't think to ask, because your answers are what make the brand yours.

Will customers notice AI-written content?

Plenty of them will suspect it. Havas found 57% of people think brands use AI to create shortcuts, as reported in the IPA's September 2026 AI magazine (Havas hasn't released the full data yet). Generic phrasing gives it away faster than any detector.

Can I build my website with AI?

You can get a working site from a prompt in an afternoon. Expect it to look like the sites the tool learned from, so swap in your own words, photos and colours before launch, and test it on your phone.

Is AI content bad for Google?

Google's guidance, updated October 5, 2026, accepts automation that helps readers and calls content made with AI "for the primary purpose of manipulating search rankings" spam. It also asks whether your use of AI is clear to visitors.

One letter a month

Unhidden in Plain Sight is my monthly letter about being seen: the patterns that keep good businesses invisible, and what to do about them.