AI search optimisation (AEO and GEO) for South African businesses

What AI search optimisation is, and what it is not

AI search optimisation is the work of making your business clear, credible and easy to quote for the AI systems people now use to research suppliers. It sits next to traditional SEO rather than replacing it. The same pages that earn a place in Google’s results are usually the pages an AI assistant reads before it answers.

You will meet several labels for this work, and they overlap. We use them plainly so you can tell what a proposal really contains.

Answer engine optimisation (AEO)

AEO means shaping your content so that a direct question gets a direct, accurate answer from your site. That includes Google’s AI Overviews and featured answers, voice assistants and chat assistants. It rewards pages that state the answer early, in plain words, with the evidence close by.

Generative engine optimisation (GEO)

GEO is the same idea applied to generative tools such as ChatGPT, Gemini, Perplexity and Copilot. The aim is that when one of them builds an answer about your category, it describes your business correctly and, where it shows sources, links to you.

How it differs from SEO

SEO is about ranking a page in a list. AI search is about being understood well enough to be included in a written answer. In practice the foundations are shared: crawlable pages, clear structure, accurate facts, real expertise and trustworthy mentions elsewhere. What changes is how much weight falls on clarity, consistency and third-party corroboration.

What nobody can promise. The companies that run these assistants do not publish how they choose what to recommend, and the answers vary between tools, between users and from one day to the next. Any agency that guarantees a mention is selling certainty that does not exist.

How AI assistants decide which businesses to mention

Assistants answer in two broad ways. Some answers come from what a model absorbed during training. Others come from live retrieval, where the assistant searches the web, reads a handful of pages and summarises them. Many tools blend both.

That difference matters for what you can influence. You cannot edit a trained model, but you can make sure the pages a retrieval tool is likely to read are accessible, clear and consistent with everything else said about you. We explain the reasoning, and the limits of what is known, in our guide to how AI search chooses which businesses to recommend.

What tends to matter

• Access. If a crawler is blocked, or the key content only appears after a script runs, it may never be read.
• Clarity. Pages that say who you serve, what you do, where, and at what scale are easier to summarise correctly.
• Consistency. Your name, services, locations and claims should match across your site, directories and profiles. Conflicts make a summary less reliable.
• Corroboration. Mentions on sites other than your own, such as directories, publications and associations, give a system independent reasons to trust you.
• Substance. Original, specific, well-sourced content gives an assistant something worth quoting. Thin pages give it nothing.

None of this is secret, and most of it is ordinary good practice done carefully. The skill is in the order of work and in knowing which fixes matter for your business.

What we do

Our AI search work is concrete and checkable. Every item below ends in something you can open and see.

Check what the assistants can currently reach

We review robots rules, server responses, rendering and sitemaps to confirm that search and AI crawlers can fetch your key pages. You decide which crawlers you want to allow, and we explain the trade-offs, because some businesses have good reasons to restrict certain bots.

Make your pages easy to quote

We rewrite or restructure important pages so the answer comes first, claims are specific, and each page covers one question properly. Headings are written as the questions people actually ask. Facts such as pricing ranges, service areas and turnaround times are stated plainly where they apply.

Add accurate structured data

We add and validate schema markup for your organisation, services, articles and frequently asked questions, so machines can read the basic facts without guessing. We only mark up what is visible on the page.

Align your entity across the web

We audit how your business is described on your site, your Google Business Profile, directories and social profiles, then fix conflicts in names, addresses, founding dates and service descriptions. This is unglamorous and often the single most useful step.

Build corroboration

Where it fits your plan, we help you earn accurate listings, citations and mentions in places a buyer or a model would trust. We do not buy links or place fake reviews.

A note on llms.txt

You may be told that an llms.txt file will put you in AI answers. We treat it as optional housekeeping. No major assistant has shown that it relies on the file, so it never replaces the work above.

How we measure it without invented scores

We do not sell a proprietary “AI visibility score”. Scores from tools that ask a handful of prompts are noisy, and a number that moves for no clear reason helps nobody. We report things you can verify yourself.

• A prompt log. We agree a set of real questions a buyer in your market would ask, run them on a schedule across the main assistants, and record exactly what each one said about you and your competitors. You see the raw answers, not a score.
• AI referral traffic. We set up reporting in GA4 so visits that arrive from AI assistants are separated from other referrals.
• Search demand for your name. Rising branded searches and direct visits are a useful signal that more people know who you are.
• Enquiries and how they heard. We add a simple “how did you find us” question to your forms and track the answers over time.

Reports are in plain language and tie back to enquiries and revenue, not to activity. If something is not moving, we say so and explain what we will change.

What AI search optimisation costs

We publish our prices in Rand. AI search work is part of our SEO and content programmes rather than a separate retainer, because it depends on the same pages and the same foundations.

Retainers run for a minimum of three months so there is time to see movement, then month to month on 30 days’ notice. The full inclusions are on our pricing page. Prices are reviewed annually, and a quote we give you is held for 30 days.

Who this suits, and who it does not

AI search optimisation works best for established businesses that already know something specific and valuable: an advisory firm, a fintech, a training provider, a specialist services company. Those businesses have real expertise to be quoted, and a handful of pages that matter a great deal.

It is a poor fit if you want a guaranteed result, if your site has almost no content to build on, or if you expect a model to recommend you for work you cannot yet demonstrate. In those cases we will say so, and usually suggest starting with the SEO foundations in our SEO services or a short consultation.

What happens in the first 60 days

1. Week 1 to 2: audit. We check crawler access, rendering, structured data and how your business is currently described by the main assistants.
2. Week 3 to 4: agree the plan. You get a prioritised written plan, the prompt set we will track, and a clear view of which pages come first.
3. Week 5 to 8: fix and publish. We repair technical blockers, rewrite priority pages, add structured data and start on entity clean-up.
4. Day 60: first review. You see the prompt log, the changes made, early movement and what comes next.

If you would like to see what the work involves before committing, our free team AI readiness check and our white paper on search visibility in the age of AI assistants set out the thinking in detail.

Frequently asked questions