Hibiscus Media · AI Search Notes

How to optimise your website for AI search

Your next customer isn’t Googling you — they’re asking ChatGPT who to call. Here’s how to make sure the answer is your business, in plain language, including the parts most guides won’t tell you.

Last updated: 27 July 2026 · Andy Bamford

To optimise a website for AI search you need four things: AI crawlers allowed into your site, schema markup labelling your business facts, pages that answer real customer questions in plain text, and content fresh enough that machines treat it as current. This guide covers all four.

The way people find businesses has changed. Instead of typing “accountant cairns” into Google and scrolling through ten blue links, people now open ChatGPT, Claude or Gemini and ask: “Who’s a good accountant in Cairns?” The AI gives them one answer, maybe two or three names, and that’s the whole search. No page two. If your business is one of the names, you win the enquiry. If it isn’t, you were never in the race.

I build websites and AI systems for Australian businesses, and AI search optimisation is now part of every site I build. Here’s how it actually works.

Why does this matter now?

Because your customers have already moved. Nearly nine out of ten small businesses now use AI in some form, and their customers are the same people. Asking an AI is now a normal way to find a restaurant, a tradie or a consultant. The businesses getting recommended aren’t necessarily the best ones. They’re the ones whose websites AI can read.

Step 1 — Is your website blocking AI crawlers?

Quite possibly, and this is the step almost everyone misses. AI companies read websites using crawlers with names like GPTBot, ClaudeBot and PerplexityBot. Many security tools and Cloudflare settings block these bots by default, and nobody ever goes back to check. If they’re blocked, nothing else in this article matters — your content could be perfect and no AI will ever see it.

Check your robots.txt file for rules blocking these bots. If your site runs through Cloudflare, check the bot settings in your dashboard for anything blocking “AI scrapers and crawlers.” You want the legitimate AI crawlers allowed in, while still blocking the bulk scrapers you never wanted anyway.

Step 2 — What is schema, and why does AI need it?

Schema is invisible labelling written into your pages in a format every major search and AI company understands. Humans never see it. Machines rely on it. It tells them, with no guessing: this is a naturopathy clinic in Cairns, here’s the exact address, phone number and list of services.

Without schema, AI has to guess your facts from your marketing copy — and guessing is where wrong prices, confused locations and mixed-up businesses come from. With it, every page type carries its own label: the business card on the homepage, service labels with pricing on service pages, question-and-answer labels on FAQs, article labels with dates on blog posts.

Schema isn’t new, and that’s its strength. Google and Bing built the standard fifteen years ago, and the AI tools are built on top of the same infrastructure. One piece of work, both audiences: your Google results get richer and AI gets your facts right.

Step 3 — How do you write pages AI will quote?

Write your headings as the exact questions customers ask, then answer each one immediately underneath in about fifty words of plain sentences, with real numbers where you can. The brand story comes after the answer, not before. The story is for humans. The answer is for both.

The rules that follow from this:

Keep key content in plain text. Anything inside tabs, accordions or pop-ups may as well not exist — AI often can’t read content that needs a click to reveal. FAQs in click-to-expand boxes are the classic offender.

Put your prices on the page. “How much does it cost” is one of the most common questions people ask AI. The business whose page contains an actual price range is the one that gets quoted in the answer.

Show your dates. Add a “last updated” line to your important pages and genuinely refresh them every few months. A services page untouched since 2023 reads as stale to machines too.

Use real names in testimonials. A full name, a business name and a town are trust signals machines can verify against the rest of the web. “Sarah M.” from nowhere convinces no one.

Does llms.txt actually work?

You may have read about llms.txt, a proposed standard where you publish a plain-text summary of your site for AI to read. I install it on the sites I build, and I’ll be straight with you about it, because most articles aren’t.

So why install it at all? It costs nothing once automated. It’s a sensible bet on where the standard is heading. And it’s a surprisingly good audit tool — the file is a plain-text X-ray of everything your site tells machines, which is how I’ve caught stale phone numbers and leftover staging content on sites that looked perfectly fine in a browser.

What I won’t tell you is that llms.txt gets you recommended by AI. It doesn’t — the schema and page structure are what work today. Anyone selling the file as the main event is selling you last year’s blog post.

How do you test your AI visibility?

The same way your customers use it. Open ChatGPT, Claude or Gemini and ask three questions:

Run them before you change anything and screenshot the answers — that’s your baseline. Run them again every couple of months. You’re watching for whether you get mentioned, whether the facts are right, and who gets recommended when you don’t. The competitor who keeps getting cited is worth studying, because their site is doing something yours isn’t.

Common questions

How long does it take to show results?

The technical work takes a day or two on most sites. Page restructuring takes longer. AI tools that browse live can pick up changes within days; results that depend on model retraining take months. Test monthly and expect gradual movement, not overnight change.

Is this the same as SEO?

They overlap but aren’t the same. Good SEO helps, because AI tools lean on search infrastructure. But AI optimisation adds things traditional SEO ignores: crawler access for AI bots specifically, direct quotable answers, and fact consistency across every page.

Do I need to rebuild my website?

Usually not. Most sites need the access and schema work done once, plus restructuring of a handful of key pages — homepage, services, about, contact. A full rebuild is only on the table if the site is old, slow, or on a platform that blocks the technical work.

Can I do this myself?

The testing, absolutely — run the three prompts above today. The content rules too, if you’re comfortable editing your own pages. The crawler access and schema work is technical and easy to get wrong, which is where most businesses bring someone in.

Andy Bamford builds brands, websites and AI systems for Australian businesses — from a small island in Japan. Be the business AI recommends.