AI search is the new front door to your business. Is your website built for it?

A growing share of the people who used to search Google, click a result, and land on a business's website now get their answer without ever leaving the search page — or without searching Google at all. ChatGPT, Perplexity, and Google's own AI Overviews read the web, synthesise an answer, and sometimes cite a source. If your site isn't one of the sources, you didn't lose a click. You lost the entire moment someone was deciding who to trust.

This has a name now — Generative Engine Optimization, or AI search optimization — and it's already attracting the same kind of vendors who spent a decade selling SEO "tricks" that mostly amounted to keyword stuffing. The honest version of this is less mysterious than the sales pitch suggests: AI answer engines are still reading your HTML, still weighing your structured data, and still favouring pages that answer a real question clearly. The fundamentals didn't change. What changed is that the reader on the other end is a model summarising your page for someone else, not a person scrolling ten blue links.

The single biggest lever, and the one most sites still get wrong, is structured data. When a page carries proper schema markup — telling a crawler explicitly "this is a business, here's its address; this is a frequently-asked question, here's the answer" — an AI system doesn't have to guess what a paragraph means. It can lift the answer directly. This site's own FAQ section is marked up with FAQPage schema for exactly that reason, and every page carries JSON-LD identifying what it actually is — a case study, a blog post, the business itself. That's not a trick. It's giving the reader, human or machine, an unambiguous answer instead of a puzzle to solve.

The second lever is answering the actual question, in the actual words a person would ask it — early, plainly, and once. AI summarisers reward pages that state a clear answer near the top and then support it, the same structure a good journalist uses. A page that buries its point under three paragraphs of scene-setting before it says anything concrete is exactly the kind of page these systems skip past in favour of a competitor's page that just answers the question. This is also, not coincidentally, what makes a page good for a human skimming on their phone — the two audiences want the same thing.

The third lever is one most businesses have never heard of: an `llms.txt` file, a plain-text summary of a site's key pages that some AI crawlers use the way search engines have long used a sitemap. It's new enough that its long-term impact is genuinely unproven — we say so plainly rather than oversell it — but it costs almost nothing to add and asks nothing dishonest of the content. Low cost, uncertain but plausible upside, no downside: that's a reasonable bet, not a guaranteed win.

What doesn't help, despite what's being sold: stuffing a page with the exact phrasing you hope an AI will repeat back verbatim. Fake review counts or invented statistics, which erode the trust signals these systems are explicitly trained to weigh. And a slow, poorly structured site with no schema at all, hoping a new "AI SEO package" will paper over fundamentals that were already broken for regular search. If a vendor's pitch doesn't mention structured data, page speed, or clear factual content, they're selling the appearance of a solution, not the thing itself.

The uncomfortable part for a lot of businesses: being well optimised for AI search sometimes means fewer visits to your actual website, because the answer satisfied the person where they asked it. That's a real tradeoff, not a myth to wave away. The counter-argument is that the visits you do get are from people who arrived already trusting the answer they were given — often already leaning toward you specifically, because your business was the one the AI named. Optimising for being the cited source, not just the clicked link, is a genuine shift in what "good" looks like, and worth building for deliberately rather than backing into by accident.

None of this replaces the basics — a fast site, honest content, real structured data, and a clear answer to the question your customer actually has. It just means those basics now have a second audience reading them: not only the person, but the system standing between that person and you. Build for both, and you're covered either way. Build only for the search engine crawler pattern from 2015, and you're optimising for an audience that's shrinking.