Your website is a brochure. Your customers need an answer.

The most expensive website I ever saw a small business buy was a sixteen-thousand-dollar redesign — agency-built, beautifully designed, animated transitions, custom photography, the whole treatment. Six months later the owner couldn’t tell me whether it had brought him a single inquiry. He just knew that he liked looking at it more than he’d liked looking at the old one.

This is the gap I want to talk about. The work an actually performant 2026 website needs to do is mostly invisible to the person commissioning it. And as a result, business owners keep buying brochures when what they need is a customer engine.

What a 2026 buyer journey actually looks like

Strip away the marketing language and the buyer journey for a typical service business now runs like this:

The buyer has a problem. They open Google or ChatGPT and ask a question — sometimes a category question, sometimes a specific one. An assistant or an AI Overview gives them a paragraph back. That paragraph names two or three businesses. The buyer clicks the most appealing name, lands on a website, scans for one minute, and either calls the number or closes the tab.

Notice what didn’t happen. The buyer didn’t browse to your portfolio page. They didn’t scroll through your beautiful hero animation. They didn’t watch your founder’s intro video. They didn’t read your about page. They had one minute, on one page — usually the page the AI sent them to, not your homepage — to decide whether you were the business they were going to call.

The website’s job, in that one-minute window, is to confirm what the AI already told the buyer about you. Reliability. Real reviews. Clear pricing range. Easy way to contact you. The end.

A website built for that job looks nothing like the design-led portfolios most agencies sell. It loads fast. It surfaces the trust signals on the first screen. It puts the phone number where a thumb can find it. It tells the AI crawler, in machine-readable structured data, exactly what business it represents, where, with what hours, what services, and what reviews. None of that is glamorous. All of it is what wins.

The hidden tax of a “designed” site

A site optimized for the agency portfolio carries a hidden tax that most owners don’t see until it hurts them.

The first part of the tax is speed. Custom hero animations, full-screen video backgrounds, multi-megabyte image sliders — every one of these adds seconds to first paint. A site that takes more than two seconds to load loses roughly half of its mobile visitors before they see anything. That number isn’t from me; it’s from Google’s own published research on bounce thresholds. Pretty sites bounce at twice the rate of fast sites.

The second part of the tax is machine readability. AI crawlers — and I mean the actual crawlers from OpenAI, Anthropic, Google, Perplexity, all of them — don’t render JavaScript the way humans do. They fetch the HTML and look for structured signals. Schema markup. Semantic landmarks. llms.txt. Open Graph. Microdata. A site built on a heavy page builder typically has very little of this, because page builders prioritize the visual canvas over the markup beneath it. Result: the AI cites your competitor’s plainer site instead of yours.

The third part is maintenance. A custom-designed agency site needs a designer or developer in the loop every time anything changes. Want to update your phone number? Wait for the agency. Want to add a new service? Submit a ticket. Two years in, the site is a museum piece — beautiful, brittle, expensive to touch.

What a customer engine actually looks like

I’m going to describe the inverse, because the contrast is the clearest way to teach this.

A customer engine — the kind of site I build with clients — is fast above all else. Subsecond first paint on mobile is the floor, not the ceiling. Every page is hand-coded HTML and CSS served from edge servers, no plugin layer, no theme system. The fonts are self-hosted, the images are sized for the viewport before they’re served, and there is no client-side JavaScript that does anything important above the fold.

Every page carries schema markup that names the business, its category, its address, its hours, its services, and its reviews — in machine-readable JSON-LD that lives invisibly in the page’s head. The schema is validated against Google’s Rich Results test and Schema.org’s structured-data tool before the page goes live. AI crawlers fetch the schema and learn the business; humans never see it.

Every page also has FAQPage schema where it earns it — meaning if a section answers a common question, that section is tagged so AI assistants can lift the question-and-answer pair into their responses. This is how the same page that ranks for a buyer-intent keyword in Google also gets cited by ChatGPT and Perplexity when a buyer asks the same question conversationally.

The phone number is in the header, on every page, click-to-call on mobile. The contact form is two clicks deep, no more. The trust signals — review counts, license badges, years in business — sit at the top of the page, not buried in a footer ribbon.

There are no carousels. There is no auto-playing hero video. The animations are restrained to the point that most visitors don’t consciously notice them. Everything that doesn’t directly help a buyer decide to call has been cut.

The shift in how to evaluate a website

The way to know whether a website is performing isn’t whether you like looking at it. It’s three numbers:

First, how fast does it load on a real cellular connection in the worst part of your service area? If it’s over two seconds, the buyer is already gone before they see your logo.

Second, how many inquiries does it produce per thousand sessions? Anything under fifteen on a service business is a sign that the site is failing at conversion. Twenty-five is decent. Forty is what a well-built engine produces.

Third, how often is the site cited by AI assistants when a buyer in your category asks a relevant question? Zero is the answer for most sites. The work to take that number from zero to a recognizable share is what separates a website from a customer engine.

None of these numbers are produced by how good the site looks. The brochure problem is that owners evaluate websites the way they’d evaluate a magazine spread, when the buyer is evaluating them the way they’d evaluate a phone book — quickly, instrumentally, with one question: is this the business I should call.

What I tell owners

When an owner asks me whether they should redesign their site, I ask them three things back.

Has your traffic grown year over year? If yes, the site is doing more work than they realize; redesigning is risk for unclear reward.

Are you getting inquiries from people who say “I asked ChatGPT and your name came up”? If no, the site is invisible to half the buyer population; the redesign has to address machine-readability or it’s wasted spend.

Do you like looking at your current site? If the answer to this is the only thing the owner is sure of, I tell them to keep what they have and put the redesign budget into content and AI search work instead. The brochure isn’t the problem. The lack of an engine is.

---

Luke LaFave is the founder of LaFave Consulting. The studio builds custom websites engineered to be cited by AI assistants — not to win design awards.