Getting found when the search box is a chatbot
For twenty years, "being findable" meant one thing: rank on Google. You wrote pages, earned links, chased keywords, and hoped to land above the fold on the search everyone did.
That search is changing under everyone's feet. A growing share of people now ask a question and read the answer an AI writes back, inside Google's own AI overviews, inside ChatGPT, inside whatever assistant is baked into their phone. They never see ten blue links. They see one paragraph, assembled from sources they don't click.
This has a name people are settling on, answer-engine optimisation or generative-engine optimisation, but the acronym matters less than the shift underneath it. The question is no longer only "do I rank?" It's "when an AI answers a question my business should own, is my business what it cites?"
For South African businesses this is not a far-off American problem. The assistants are already in the browsers and phones your customers use, and they're already answering "best bookkeeper in Stellenbosch" and "how do I claim VAT on fuel" without sending anyone to a website. Here's what actually moves the needle.
Write the answer, not the bait
AI models assemble answers from content that is clear, specific, and self-contained. A page that dances around a question to keep people scrolling is useless to a model. It can't extract a clean answer, so it reaches for a competitor who stated one plainly.
So state things plainly. If you want to be the source an assistant quotes on "what does receipt automation cost a small SA business," have a page that answers exactly that, in a sentence a model can lift without ambiguity. Lead with the answer. Put the specifics (numbers, conditions, the SA context) right next to it. The old SEO instinct to withhold and tease works directly against you here.
Be specific enough to be the obvious source
Generic content is invisible now, because a model already has generic content. What it doesn't have is your specificity: the actual VAT treatment of an unusual expense category, the real cost of manual capture for a business doing four hundred slips a month, the way a Sage integration behaves at month-end. Depth on a narrow, real question is what gets you cited. Breadth on a broad one gets you averaged into the background.
This is also why local specificity is an edge, not a limitation. "Reconciliation automation for South African accountants" is a lane a global tool won't bother to own precisely. You can.
The fundamentals didn't leave. They got promoted
None of this replaces the plumbing. If anything, the boring parts matter more, because they're how a model decides you're trustworthy enough to quote:
- Structure the page so a machine can read it. Clear headings, real HTML, a sensible hierarchy. Structured data (schema markup) for your business, products, and FAQs tells an engine exactly what it's looking at.
- Be fast and crawlable. If a model's retrieval step can't load your page, or gives up waiting, you're not in the running. Server-rendered content beats a page that only assembles itself after a heavy script runs.
- Earn the ordinary trust signals. Being mentioned, linked, and reviewed across the web still tells both Google and the models that you're a real, credible source. Citations are the new currency, and they're earned the slow way.
What we'd tell a client to do this quarter
Pick the three questions a good customer asks right before they'd buy from you. Write one genuinely excellent, specific, self-contained page for each: the answer first, the SA detail next to it, structured so a machine can parse it. Make sure those pages load fast and render their content without waiting on scripts. Then get mentioned, honestly, wherever your customers already look.
That's it. It's less mysterious than "AI optimisation" sounds. The engine changed; the job is the same job it always was, done more plainly: be the clearest, most specific, most trustworthy answer to a question worth owning.
If you'd rather have the site itself built that way from the start, fast, structured and written to be the answer, that's work we take on.