Today a client pointed me to the new design on a CRM website, with that nice “on trend” serif font that’s starting to turn up everywhere. I immediately recognised the Claude Code and Claude Design style that seems to be spawning all over the internet.
It looks nice. But if it were a website for a local business, I think it would struggle to convert.
Why? Because a website that converts isn’t a design problem. It’s a “who’s reading this and what are they worried about” problem.
AI doesn’t know your customers yet. It knows the average of every website on the internet, and the average website is American, generic and trying too hard (if you ask an Australian).
Australians are different. We read web pages differently. We think and process information differently. Roy Morgan asked 1,024 Australians about advertising late last year. 88% said advertising should reflect community standards, and only 37% said what they see actually does [1]. We’re a sceptical bunch, and we can smell a sales pitch from the next suburb. A website that’s trying too hard triggers that instinct before the visitor has read a word.
Here’s what we’ve found actually works, and where AI keeps tripping.
You Have Seconds, Not Minutes
Before we get to the individual pieces, it helps to know how little time you actually have.
People judge the look of a web page in about 50 milliseconds. That’s the finding of a well-known study at Carleton University, and it means the visitor has an opinion of your business before the page has finished loading [2]. Stanford’s web credibility research found that when people explain why they do or don’t trust a site, the visual design comes up in 46% of the comments, more than any other factor [3].
Then they start scanning. Nielsen Norman Group’s eye-tracking research found people read about 20% of the words on an average page, and most decide whether to stay or leave within the first 10 seconds [4][5]. The same research found pages with a clear value proposition hold attention much longer.
So every point below is really about the same thing: make the right impression fast, then don’t get in the way.
The Headline Has to Cut Through
We’ve found local visitors tend to hang around on a website when the Unique Value Statement (that big headline at the top of the page) cuts through the noise of the internet. When it says something different and authentic to the values of the business. Something with a bit of personality and fun. Something simple and easy to process.
The eye-tracking backs this up. People read in an F-shaped pattern: across the top first, then a shorter sweep lower down, then a skim down the left edge. In NN/g’s words, “first lines of text on a page receive more gazes than subsequent lines” [6]. Your headline is the one line almost everyone reads. If it’s generic, the visitor assumes the business is too.
AI writes the safe version every time, because the safe version is what most websites say. “Quality service you can trust” is the average of a million home pages. Nobody in Perth has ever picked a plumber because of it. For more on this, see what every page now needs since AI changed content.
Copy That Converts Is About the Visitor
They want to know what’s in it for them. They don’t want to hear about the business. They want to know their frustrations are understood and addressed by the service on offer.
Plain language matters more than most business owners think. Unbounce analysed tens of thousands of landing pages for its 2024 Conversion Benchmark Report. Pages written at a 5th to 7th grade reading level converted at 11.1%, roughly double the 5.3% for pages written at a professional level. Adding long, difficult words had a stronger negative effect on conversion than adding more words [7].
That’s exactly how Australians talk. Short words, straight to the point, no fluff. AI tends to write the opposite, because so much of what it learned from was written to sound impressive rather than to be understood.
One Idea Per Section
They’ll leave if the design is too complex. The site needs to present one simple idea in each clearly defined section as they scroll. The message needs to flow down the page in order, so the visitor never has to look in two places to understand one thing.
That’s where the AI style falls down for a local business. Every section is fighting for attention and none of them win.
Remember the visitor is only reading a fifth of your words. If a section carries three ideas, two of them are invisible. A page that flows one idea at a time gives the skimmer a story they can follow with their thumb. A page of competing panels gives them a reason to go back to Google. Good information architecture makes this easy to plan before a single section is built.
They Want to See a Real Person
A photo of you, your team or your van does more than any stock image. If stock is used, it has to look authentic, like it’s been taken in the visitor’s location.
This is where AI slips most. It reaches for American kitchens, American faces and American weather, and locals notice, even if they can’t say why.
The evidence on this one is consistent. NN/g’s eye-tracking found people simply ignore stock photos of generic people, but spend real time looking at photos of the actual staff who work at a business [8]. A moving company that swapped its stock photos for a picture of its own crew lifted conversions by 45% [9]. A debt counselling service that replaced a stock “smiling woman in a headset” with a photo of its founder saw 35% more consultation sign-ups [10].
AI-generated imagery makes this worse, not better. Getty Images surveyed more than 30,000 people across 25 countries. 98% said authentic images are pivotal to trust, 87% said it’s important an image is real, and respondents felt less favourably towards brands using AI-generated visuals of people or products [11]. A quick phone photo of you at a job in Joondalup beats a perfect render of a man who doesn’t exist standing in a kitchen that isn’t here.
| Business | Change made | Result |
|---|---|---|
| Moving company | Swapped stock photos for its own crew | +45% conversions |
| Debt counselling service | Replaced a stock headset photo with the founder | +35% sign-ups |
They Want Proof Before They Call
A couple of short, punchy reviews. Or perhaps locally recognised logos of who you’ve worked with. Photos of finished jobs. Enough to back up the headline at the top.
Reviews are close to universal now. BrightLocal’s 2026 Local Consumer Review Survey found 97% of consumers read reviews for local businesses, and 74% want to see reviews from the last three months [12]. Recent matters. A glowing review from 2019 tells the visitor nothing about who’s turning up this week.
You don’t need dozens. Northwestern University’s Spiegel Research Center found purchase likelihood for something with five reviews is 270% higher than with none, and the benefit of each extra review drops away quickly after that [13]. Five recent, specific, local reviews on the home page do the job. A wall of fifty doesn’t add much, and it starts to look like a wall of text.
One note for health businesses: if you’re regulated by AHPRA, you can’t use testimonials at all. Your proof is credentials, years in practice and real photos of the team.
The Call to Action Has to Fit the Job
They want to take action conveniently. A button top right, because that’s where people have learned to look. A form on the page so they can enquire without hunting for it. On a phone, a number they can tap.
That last one is bigger than it sounds. Roughly four in ten web visits in Australia are on a phone [14], and for a tradie or an emergency service it’s usually more. Google’s research with Ipsos found more than 40% of mobile searchers use click-to-call, and that the option to call matters most at the point of purchase [15]. If your number isn’t tappable, you’re asking someone with a burst pipe to memorise ten digits.
AI may know to put a button near the top of the page, but it doesn’t always understand which call to action will work better. Someone looking for a new home needs to enquire or have a chat. Someone with an urgent plumbing job needs a fast quote. A physio patient wants to book a time. Understanding what the audience needs is critical, and it changes the wording, the placement and what happens after the click.
The Quieter Things
A few quieter things sit underneath all that. A familiar layout, so nobody has to learn your site. Mostly neutral with a few pops of colour, because light and calm converts better than dark and busy for most local businesses.
The familiar layout has a name in the usability world. Jakob’s Law says users spend most of their time on other sites, so they expect yours to work the way those do [16]. Logo top left, menu across the top, button top right, phone number visible. A clever layout that breaks those habits isn’t clever. It’s a small tax on every visitor, and the busy ones don’t pay it.
Then the local details. Australian spelling. A phone number that looks like an Australian phone number. Suburbs and landmarks the reader knows. A .com.au domain. None of these convert on their own, but each one is a small signal that says “this business is here”, and AI gets them wrong by default because most of what it learned from wasn’t. This is the same thinking behind what local SEO actually covers.
What AI Gets Right
To be fair, plenty. The structure is clean, the code is fast, the spacing is consistent and the pages come out in hours rather than weeks. Those things used to cost real money and now they don’t. That’s good for local businesses, and we use it every day. We’ve written before about letting AI do the robot work so people can be human.
What it can’t do yet is sit across the table from a Mandurah mortgage broker and understand why her clients are anxious, or know that a Perth homeowner reads “premium” as “expensive”.
AI Builds. A Human Still Decides.
None of this is complicated. But it’s local knowledge, built up from years of watching what people in this market actually do. AI may get more granular one day. It isn’t there yet. For now it still needs a human who knows the area to pick the target market and understand what they’re worried about before a single section gets built.
That’s how we use it. AI does the building. We decide what goes on the page and why.
A 5-Minute Check
Open your home page on your phone. Without scrolling, can you find:
- Who you are and why you’re different
- A real photo
- One piece of proof
- A way to contact you
Read the headline out loud. Would you say it to a customer at the counter? If anything’s missing or sounds wrong, that’s the first thing to fix.
Want a second opinion on your website?
We’ll tell you what’s working, what isn’t, and why in plain English.
Let’s ChatFrequently Asked Questions
Can AI build a website for a local business?
Yes, and the build quality is often good. The gap is in the decisions before the build: who the customer is, what they’re worried about, what proof they need and what action they’re ready to take. Those still need someone who knows the market.
Are stock photos bad for conversion?
Generic stock photos of people tend to get ignored, and real photos of your team and work consistently test better. If you must use stock, choose images that look like they were taken where your customers live.
How many reviews do I need on my website?
Around five recent, specific reviews is enough to move the needle. Research shows the benefit of each additional review drops off quickly after that, and people increasingly want to see reviews from the last three months.
Does website design affect trust?
Strongly. People form an impression in a fraction of a second, and design is the most common reason they give for trusting or distrusting a site. Design matters for trust, not for decoration.
More Articles
Sources
- Roy Morgan, Advertising expectations and experience: Australians have their say (March 2026): roymorgan.com
- Lindgaard, Fernandes, Dudek and Brown, “Attention web designers: You have 50 milliseconds to make a good first impression!”, Behaviour & Information Technology (2006): semanticscholar.org
- Stanford Web Credibility Project, How Do People Evaluate a Web Site’s Credibility (2002): credibility.stanford.edu
- Nielsen Norman Group, How Little Do Users Read?: nngroup.com
- Nielsen Norman Group, How Long Do Users Stay on Web Pages?: nngroup.com
- Nielsen Norman Group, F-Shaped Pattern of Reading on the Web: nngroup.com
- Unbounce, Conversion Benchmark Report (2024): unbounce.com
- Nielsen Norman Group, Photos as Web Content: nngroup.com
- Conversion Sciences, Which Pictures Convert Best (Harrington Movers case): conversionsciences.com
- MarketingExperiments, This Just Tested: Stock images or real people?: marketingexperiments.com
- Getty Images, Building Trust in the Age of AI report (April 2024): newsroom.gettyimages.com
- BrightLocal, Local Consumer Review Survey 2026: brightlocal.com
- Spiegel Research Center, Northwestern University, How Online Reviews Influence Sales: spiegel.medill.northwestern.edu
- Statcounter Global Stats, Desktop vs Mobile vs Tablet Market Share Australia (August 2026): gs.statcounter.com
- Econsultancy reporting Google/Ipsos click-to-call research (2013): econsultancy.com
- Nielsen Norman Group, Jakob’s Law (End of Web Design): nngroup.com
