The website your clients' AI assistant reads

There's another Kristina Agustin. She lives in California, she's done nothing wrong, and until earlier this year, if you asked an AI assistant about my name, she often came up ahead of me. My family collected screenshots of my outrage. I decided to treat it as a challenge to a digital duel, one she never knew she was in.

So I turned the tools I use for clients on my own business. I wrote down, in one place, who Southern Sky AI is: the founding story, the credentials, the positioning, the plain facts. I rebuilt my About page as the single page a machine can resolve my identity from, with everything else on the site pointing back to it. I brought every professional profile and directory listing into line with the same description, correcting years of small inconsistencies I'd stopped noticing. And underneath the words, I added the technical layer: structured data a machine can parse without guessing, a file at the root of the site pointing AI crawlers to the pages that matter, access for those crawlers by name, and headings phrased as the questions people ask, with the answer sitting directly underneath.

The results arrived within days. Google began returning clear, accurate descriptions of Southern Sky AI, and the assistants followed. I never handed any model a script; what they say is their own synthesis of what they can now read. Ask Gemini about my name today and I come up first. If the order ever flips back, tell me.

What moved those results is a discipline called answer engine optimisation, AEO for short: making your business the source the AI assistants quote when someone asks a question you're qualified to answer.

Traditional SEO was a ranking game. A search returned a list of results, people clicked through and compared, and the contest was for page one. Now AI answers sit on top of roughly a quarter of Google searches, and where they appear, the clicks that follow have fallen by around a third. That's before you count the people who skip the search box altogether and ask ChatGPT.

When a charter client, an owner's representative or a fleet manager asks an assistant to recommend a surveyor, a management company or a training provider, what comes back is one synthesised answer, sometimes a short list. There's no page two.

So the question your website answers has widened. It was always whether a person browsing could find you and trust you. Now it's also whether the machine reading your site can find you, parse you and cite you. Most maritime websites were built for one visitor, a person with a browser, comparing and clicking, and the good ones did that well. A second reader has arrived alongside that person, and it reads differently.

My own site was the first rebuild, and the duel was the proof.

The largest so far is a regional industry member association carrying close to fifteen years of published content, a deep archive sitting on a site built for an earlier era of the web. The association brought that archive into a rebuild, and Southern Sky AI delivered it as a platform of 2,076 indexable pages: 1,645 migrated articles and a 147-listing member and partner directory, with structured data on every page and a crawler guide file at the root. The site now carries over a million requests a month, is cited as a source in ChatGPT, and is read daily by every major AI engine. In one recent four-day window, ChatGPT and Claude between them consulted it live close to 1,500 times, and one in ten of ChatGPT's fetches was a member's individual directory listing. When someone asks about that corner of the industry now, those pages are what the machine is reading.

A third engagement ran a three-stage AEO programme across two entity sites and four credibility categories, and it surfaced the find that justifies the whole discipline: an override at the infrastructure level, silently blocking AI crawlers even though every setting on the site itself was correct. The owner could have checked every recommended box and never known the machines were being turned away a layer higher up. We identified it, resolved it, and the presence is now readable and recommendable by AI systems at every layer. Access gets verified at every layer, and verified again after anything changes.

Four things, whether it's a new build or a renovation of the site you have.

You don't need to read code to take a first sounding. Open your site and consider:

• -Does the first screen say plainly what you do and who you serve?
• -Does each service have its own page, or do several share one?
• -If a stranger read only your About page, could they say who you are, what you're qualified in and where you operate?
• -Are your headings the questions clients ask, with the answer directly underneath?
• -Do the descriptions of your business match across your site, your LinkedIn profile and the directories that list you?
• -Do your pages open within a couple of seconds on a phone?

Then ask an assistant two questions: "Who is [your name]?" and "What do you know about [your business]?" Whatever comes back is your starting position, in the machine's own words.

AEO-informed website builds and renovations are now part of how I deliver AI deployment: new sites designed from the first page plan to be read and cited by the systems answering your clients' questions, and reworks of existing sites that keep what serves your visitors and add what the machines need to read you. Websites are one build among several in that work; what a finished AI build looks like walks through four builds from the Log Book.

If website work is anywhere in your plans this year, or you'd like to know what the assistants can see on your site today, start with a conversation. Ask them about your own name first, and bring me what they say. I'd like to hear it.

And if a namesake comes up ahead of you, you'll know how I felt. The order can change; mine did.