AI Visibility in South Africa: how SA brands show up in ChatGPT, Claude, Gemini and Perplexity.
You suspect your brand is not being named. You have asked ChatGPT a few questions, watched a competitor get the spotlight, and now you want to know whether it is a real pattern or a one-off roll of the dice. It is a real pattern. The same handful of SA brands keep getting named across every engine — Capitec in banking, MTN in telecom, Discovery in medical aid, Showmax in streaming. The brands sitting outside that handful are doing real business but losing the AI consideration set every day.
This page is the buyer-journey gateway. It tells you what AI visibility actually is, what it looks like for SA brands right now, the seven most-common patterns that pull a brand out of the answer, and the action ladder a CMO can walk down on Monday morning. It is built on the SA-AEO-Bench — pre-registered on OSF, replicated quarterly, 19,020 successful AI responses across the frontier engines, 232,254 cited URLs classified, 100 South African brands across 12 industries.
Why AI visibility matters right now: it is the new shape of being considered.
If you are not named, you are not considered. That is the whole argument in one sentence. Google used to give a customer ten blue links and let them pick. ChatGPT, Claude, Gemini and Perplexity give a customer three or four named brands inside a paragraph and skip the rest. The customer does not scroll. The customer reads the answer. The customer either sees your brand inside it or they do not.
That single change rewires the top of every funnel in South Africa. The medical-aid switcher who used to compare six schemes now reads two paragraphs naming three. The fibre shopper who used to evaluate every ISP in their suburb now sees MTN, Vodacom and one other. The CFO comparing acquiring banks for a new subsidiary now gets the same three names from every engine. The brands inside those answers compound. The brands outside them shrink, quietly, every quarter.
South Africa is not the United States. The SA corpus that ChatGPT, Claude and Gemini read is a fraction of the size of the US corpus, which means a small number of SA publishers carry outsized weight in how engines answer SA questions. BusinessTech, Moneyweb, MyBroadband, Fin24, IOL, TimesLIVE, the trade publications for each category, and the complaint layer — that is the citation core. If you are well covered there, AI knows you. If you are not, AI is making you up from fragments. Both outcomes are measurable.
Take banking. Capitec is named in nearly every AI answer about SA retail banking. FNB shows up in most. Standard Bank, Absa and Nedbank trade the next two spots depending on the question shape. Tymebank lands in challenger-tier answers reliably. Bidvest Bank — a perfectly real banking licence with a perfectly real customer base — barely registers at all. The gap is not about quality. The gap is about whether AI has been fed enough of the right SA-source signal to think of the brand when the question shows up. That gap is what AI visibility measures and what the free Scorecard surfaces for your brand.
The same shape repeats across every category we measure. In telecom, MTN and Vodacom dominate organic mentions and the fibre/ISP tier sits in shadow. In medical aid, Discovery is the default and the smaller schemes flicker. In streaming, Showmax holds the spot and everything else slides into a "vs Netflix" frame. In short-term insurance — the most SA-protected category we measure — the visibility battle is between SA brands, not against foreign challengers. The category-level snapshots live in the Q2 Index Report and the per-industry pages linked further down.
The CFO conversation is changing too. Two years ago AI visibility was a CMO curiosity. This quarter it is in the budget conversation in every brand we speak to. The reason is simple: the cost of being outside the AI consideration set is real, measurable and compounding. The brands taking it seriously now are the ones who will be inside the answer in twelve months. The brands waiting will spend longer and pay more to catch up. The research dossier has the data and the framing for that conversation.
The rest of this page is the working brief. Three layers of visibility. Twelve categories snapped to the SA-AEO-Bench reading. Seven patterns that pull a brand out of the answer. An action ladder that starts with a free Scorecard and ends with a board report. Read it. Forward it. Or skip straight to the Scorecard if your brand is the question.
AI visibility is three layers.
Every conversation about AI visibility collapses if you do not separate three things: whether AI names you, what AI says about you, and how that picture changes between engines. Confuse the three and you end up arguing about the wrong problem. Separate them and the brief writes itself.
Whether AI mentions you.
The first layer is binary. You ask the engine a generic question your category answers — "best medical aid in South Africa", "fibre provider Cape Town", "Black Friday SA e-commerce" — and you watch whether your brand gets named. Across replications. Across engines. Without the prompt cueing your name first. This is organic mention rate, and it is the cleanest read of whether AI thinks of you on its own.
The pattern in SA is brutal. A small number of brands per category run the table — Capitec in banking, MTN in telecom, Discovery Health in medical aid, Takealot in e-commerce. The rest of the field clusters into a middle pack with patchy visibility. Then there is the forgotten tier — real brands with real customers that AI almost never names. The shape repeats every quarter.
What the Scorecard measures here is your blind mention rate, per engine, against the same prompt set we use for the SA-AEO-Bench. You see whether you are in the dominant tier, the middle pack, or the forgotten tier — and you see the gap in plain numbers. The first time a CMO sees this read, it is usually a quiet meeting. The second time, it is a planning meeting.
The category snapshot lives in the Q2 Index Report. The brand-specific read is the Scorecard.
What AI says about you.
Getting named is not the same as getting named well. The second layer is narrative — what AI actually says when your brand shows up in the answer. Is the framing positive, neutral or hedged? Is the engine reaching for your own editorial (which you control), independent editorial (which you can influence), or the complaint layer (which can punish you)? The mix of own, editorial and complaint sources behind your name shapes the sentence that the customer reads.
The SA-specific signal here is Hellopeter. When a SA user asks a brand-evaluation question — "is this brand any good", "what are the issues with", "should I avoid" — Gemini and ChatGPT lean into the complaint layer hard. A brand with a thin complaint signal and a credible response posture comes out clean. A brand with a noisy complaint signal and silent management comes out scarred. The same brand can read very differently on a positive query versus a negative query, and the gap between those two reads is one of the most useful diagnostics on the whole Scorecard.
Take a SA short-term insurer. On a positive prompt — "best car insurance in South Africa" — the engines reach for the comparison sites, the financial press, the insurer's own editorial. The framing is favourable. On a negative prompt — "problems with car insurance claims South Africa" — those same engines reach for Hellopeter, MyBroadband forum threads and a handful of news pieces about industry-wide complaints. The framing tilts. Your brand can be in a tidy positive frame and a punishing negative frame at the same time. Both reads are the AI answer your customer sees.
This is the gap a Scorecard surfaces and the conversation a 30-minute walkthrough turns into a brief.
How that picture spreads between engines.
The third layer is per-engine spread. ChatGPT, Claude, Gemini and Perplexity read different SA sources, weight them differently, and answer the same question with different brand names. A brand that owns ChatGPT can be invisible in Gemini. A brand killing it on Perplexity may have decayed out of Claude entirely. There is no single AI-visibility play that works across all four — there is a posture per engine, and you need a read on each.
The most consistent split we measure is the Reddit signal. Gemini cites Reddit hundreds of times on SA queries. ChatGPT and Claude cite it zero times. If your audience is searching on Google — and Google AI Overviews is powered by Gemini — your forum exposure matters. If your audience is in ChatGPT, your forum activity does not move the answer at all. The SA-source share also varies wildly by category — over 70% in short-term insurance, around 3% in safari lodges — which means the same "be in the SA citation core" advice means very different things in different industries.
Per-engine spread is the layer most often missed by tools built on a single engine. A read that only tells you about ChatGPT is telling you about one quarter of the AI audience. The Scorecard runs across ChatGPT, Claude and Gemini and the Index Report adds the cross-engine context for your category. Treat the spread as a separate measurement — because it is.
See the per-engine cut for your brand in the Scorecard. See the per-engine cut for your category in the Q2 Index Report.
What SA brands actually look like in AI right now: how the twelve categories show up right now.
The SA-AEO-Bench covers twelve industries. Each one has its own pattern — its own dominant brands, its own SA source mix, its own per-engine quirks. A quick read on each is below. Click through to the category page for the full reading, including the SA publishers carrying the weight, the per-engine posture and the brands inside the measurement.
The category reads are the context. The Scorecard is the brand-specific cut against that context. Read the category to know the field. Run the Scorecard to know your position on it.
Capitec, FNB and Standard Bank get named almost every time. Bidvest Bank — perfectly real, perfectly licensed — barely registers. The gap between top-tier and forgotten is wider in banking than in any other SA category.
MTN and Vodacom dominate organic mentions. The fibre and ISP tier — Afrihost, MWeb, Webafrica, Vox — shows up almost only when the prompt names them first. A clean signal of how thin the SA telecom long-tail coverage actually is.
Discovery Health is the default in every engine. Bonitas and Momentum show up regularly. The smaller schemes — Sizwe Hosmed, Genesis, Fedhealth — appear or vanish depending on the engine and the question shape.
Toyota, Volkswagen and Ford SA own the heritage lane in ChatGPT. Haval, Chery and the Chinese-import wave show up faster in Gemini and Perplexity. The per-engine spread is the widest of any SA category.
Shoprite, Checkers, Pick n Pay and Woolworths trade positions depending on the question. Spar lags despite store count. The smaller chains — OK Foods, Boxer — are barely visible to AI on a blind prompt.
Takealot owns the SA mention. The hard problem is the international default — Amazon shows up in answers to SA shopping questions more often than any SA retailer expected. The category leaks citation share to global frames.
Showmax holds the SA spot. The rest of the category sits in a "vs Netflix" frame more than as standalone SA brands. The smaller services — eVOD, Vidi, BoxOffice — exist for AI only when their name is in the prompt.
The most SA-protected category in the bench. OUTsurance, Discovery Insure, Santam and the comparison sites dominate. International competition is near-zero. The visibility battle is between SA brands, not against foreign challengers.
Pam Golding, Seeff and Property24 carry the agency-and-portal layer. The bond-origination tier — ooba and the rest — is almost invisible on blind prompts and almost universal on named prompts. A textbook prompt-cued category.
The most internationally-routed SA category. Limited specialist SA restaurant publishers means international food-delivery review sites take the citation share. Local chains have to fight harder for every AI mention.
The branded SA groups — Tsogo Sun, Sun International, Protea — show up reliably. Independent properties live and die on TripAdvisor and Booking.com. A category where international-platform visibility maps almost directly to AI visibility.
The flagship lodges — Singita, andBeyond, Londolozi — are well covered in international travel media. SA-source share is the lowest in the entire bench. The whole category is being narrated to AI by foreign publishers.
The 100 SA brands measured cover the biggest names in each category — from Discovery, Capitec, MTN and Vodacom to Shoprite and the rest of the dominant tier — through the middle pack and into the forgotten tier. Every brand page in the brand directory shows where it sits in its category, how it reads across the four engines, and what is currently shaping its AI answer.
The seven most-common AI visibility problems: every brand we Scorecard sits inside one of these seven patterns.
The patterns are not surprising once you have measured a few hundred brands. They are surprising the first time you see your own. Each one carries a different fix, different timeline and different priority engine. The order below is the rough frequency we see in the SA bench.
International default.
You ask a SA shopping question. ChatGPT names Amazon before Takealot. The engine is reaching for the bigger global signal and skipping the SA brand. The fix is patient — you have to feed the SA citation core with enough new coverage that the engine updates its default. It does not happen in a week.
This is what your Scorecard surfaces.
Hellopeter dragnet.
You ask a brand-evaluation question. The engine reaches into the SA complaint layer and pulls Hellopeter, with the loudest threads at the top. A brand with thin complaint volume reads cleanly. A brand with high volume and silent management reads badly — and the same negative thread can shape three different engine answers.
This is what your Scorecard surfaces.
Wrong-entity collision.
Your SA brand shares a name with an unrelated foreign company and the engine confuses the two. Your answers get polluted with someone else's news, products and reviews. The fix is entity discipline — schema, knowledge-graph hygiene, consistent disambiguation across every authoritative SA source.
This is what your Scorecard surfaces.
Per-engine inconsistency.
You own ChatGPT and you are invisible in Gemini. Or the other way around. The same prompt returns different brands depending on which engine answers it — and your audience is split across all four. A single-engine strategy leaves three quarters of the AI audience to your competitors.
This is what your Scorecard surfaces.
Source concentration risk.
Your AI visibility is almost entirely driven by one or two SA publishers. If those publishers go quiet, change editorial focus, or stop covering your category, your AI visibility collapses with them. A healthy citation map has a spread of sources across the SA core, not a single load-bearing publisher.
This is what your Scorecard surfaces.
Freshness decay.
You used to be the obvious answer. Then your press cadence dropped. Six months later Perplexity reads you as "the old one" and Gemini has half-forgotten you exist. Freshness decay punishes incumbents who think the category is settled. The fix is a steady rhythm of newsworthy coverage in the SA-source layer.
This is what your Scorecard surfaces.
Forum invisibility.
Gemini reads Reddit and MyBroadband forums in its hundreds of citations on SA queries. ChatGPT and Claude reach for those forums almost never. A brand active on the forums is visible to Google AI Overviews and quiet in ChatGPT. A brand absent from the forums is the reverse. The split is structural and known.
This is what your Scorecard surfaces.
Which of the seven you are sitting inside is the first answer the Scorecard returns. Most brands carry two or three at once. The order in which to fix them — and which to leave for later — is the conversation in the 30-minute walkthrough.
What to do about your AI visibility: the four steps your team takes from here.
The work is not theoretical. It is four concrete steps, in order. Skipping the first one to chase a tactic from the third is the most common mistake we see — a brand decides it has a Hellopeter problem and starts a Hellopeter project before it has confirmed Hellopeter is even what the engines are reaching for in its category. Diagnose first. Plan second. Execute third. Each step below lands on the same outcome — a clearer read on where your brand stands and what the right next move is.
Diagnose with the free Scorecard.
The cheapest, fastest read on where your brand sits. Run your brand through the free Scorecard and you come back with your per-engine mention rate, the SA sources currently shaping your AI answer, and which of the seven patterns you are sitting inside. No commitment, no card, no sales call. The Scorecard is engineered to give you a real answer in a single sitting.
A CMO who has not seen their own Scorecard is operating on the assumption that ChatGPT thinks of them. Most of the time the assumption is wrong, and the meeting after the Scorecard is the planning meeting.
Audit your placement on the SA citation core.
The Scorecard tells you which SA sources are shaping your AI answer. The audit asks the harder question — which ones should be, and are not. The SA citation core for most brands is some combination of BusinessTech, Moneyweb, MyBroadband, Hellopeter and the industry-specific publishers (Cars.co.za and AutoTrader for automotive, Property24 and Private Property for real estate, Medshield and the medical-aid press for healthcare). A thin presence on the right publishers is worth more than a thick presence on the wrong ones.
The category-specific list of the right publishers is in the Q2 Index Report and on each of the industry pages.
Get the Q2 Index Report for category context.
Your Scorecard is your read. The Q2 Index Report is the field around you. You learn how the dominant brands in your category got to where they are, where the challengers are gaining ground, what is changing quarter-on-quarter, and what the per-engine posture looks like for the whole field. This is the document you forward to the CFO before the budget conversation. It frames the spend against a category context, not a one-brand curiosity.
The Index Report ships quarterly. The Q2 reading is live now. Q3 lands in three months.
Book the 30-minute walkthrough for the brand brief.
The walkthrough is where the Scorecard and the Index Report turn into a brand-specific brief. Thirty minutes with the research team, your Scorecard open on screen, the category context loaded, and a working answer to the questions — what are the three patterns I am sitting inside, which one do I fix first, what does fixing it look like in practice, how do I show progress to the board. There is no follow-up sales loop. The work either lands or it does not. Book the walkthrough when you want the action plan and not just the read.
The brands we have run through this loop come out the other side with a 90-day plan and a quarterly cadence. That is the operating rhythm.
The questions CMOs ask, answered straight.
What is AI visibility?+
AI visibility is whether your brand gets named when a customer asks an AI assistant — ChatGPT, Claude, Gemini or Perplexity — a question your category answers. If your brand is named, you are in the consideration set. If your brand is not named, you are not being evaluated. AI visibility measures three things at once: whether AI mentions you, what AI says about you, and how that picture changes between engines. It is the AI-era equivalent of the question "do I show up on the first page of Google" — but with three or four answer slots instead of ten, and four engines instead of one.
How do I check if AI cites my brand?+
The fastest read is the free Scorecard at citedbrands.co.za/scorecard. We run your brand against the same prompts the SA-AEO-Bench uses for your category, across ChatGPT, Claude and Gemini, and return your per-engine mention rate, the SA sources currently shaping your AI answer, and the three patterns most likely to lift your visibility. You can also test it yourself by asking each engine the questions your customers ask — but a one-off prompt is a single roll of the dice. AI answers vary between sessions. You need replication and a benchmark to read the signal cleanly.
How is AI visibility different from SEO?+
SEO is about being listed. AI visibility is about being named. Search engines return a page of ten ranked links and let the user decide. AI engines return a short answer with three or four brands named in the sentence and skip the rest. SEO rewards keyword coverage and link equity. AI visibility rewards being the obvious answer to a question — which in South Africa means being well covered by the SA citation core (BusinessTech, Moneyweb, MyBroadband, Hellopeter, the industry-specific publishers) and being framed correctly when AI summarises that coverage. The skills overlap, the tactics do not.
Does AI visibility translate to revenue?+
It already does, and the share is growing every quarter. AI assistants are now the first stop for a meaningful slice of high-intent SA buying queries — medical aid choice, banking switch, fibre provider, holiday lodge, e-commerce comparison. A brand that gets named in the AI answer enters the consideration set. A brand that does not is invisible to that buyer. We cannot promise a revenue multiple — that depends on category, conversion path and pricing — but the directional finding from the SA-AEO-Bench is unambiguous: AI-cited brands win share of consideration, and share of consideration leads buyers down the funnel.
How fast does AI visibility change?+
Faster than SEO, slower than paid. A brand that has been ignored by ChatGPT for two quarters does not flip in two weeks. The engines need a meaningful body of fresh SA-source coverage before they update their default answer. Perplexity moves fastest — fresh news lands in its answers within days. Gemini follows. ChatGPT and Claude are the slowest to update, which is good news for incumbents and bad news for challengers. The Q2 Index Report tracks the rate of change per industry. The Scorecard is the read on where your brand sits today.
Can I measure AI visibility myself?+
Yes, partially. You can ask each engine the questions your customers ask and see whether you get named. That gives you a directional read. What you cannot do casually is replicate prompts to wash out variance, counterbalance for ordering bias, classify the citations behind each answer, and compare your numbers against a published SA peer set. That is what the SA-AEO-Bench gives the Scorecard — the benchmark to compare against. You can DIY the question. You cannot DIY the answer with statistical confidence. That is why the free Scorecard exists.
Which AI engine matters most in South Africa?+
It depends on your category and your audience. Google AI Overviews — powered by Gemini — reaches the most SA users by sheer search-share, so any brand with a high-funnel discovery problem cannot afford to be invisible there. ChatGPT dominates considered-purchase research and the professional knowledge-worker audience. Claude leans into thoughtful comparison work and is the engine most likely to flag a complaint signal alongside a recommendation. Perplexity owns the news-shaped query. If you have to pick one to chase, pick the engine your audience uses. For most SA brands, you do not get to pick — you need to show up in all four.
Does my Hellopeter rating matter for AI?+
For Gemini and ChatGPT, yes — significantly. Both engines reach into the SA complaint layer when a user asks a brand-evaluation question, and Hellopeter is the largest layer they touch. A brand with a thin complaint volume and a credible response posture reads cleanly. A brand with high complaint volume and silent management reads badly. Claude pulls Hellopeter less frequently but weights it heavily when it does. The Scorecard surfaces your current Hellopeter exposure across each engine and the gap between your positive-query framing and your negative-query framing — which is often wider than CMOs expect.
How is AI visibility different in South Africa from the US or UK?+
Smaller corpus, sharper signal. The SA AI corpus is a fraction of the size of the US or UK corpus, so a handful of SA publishers carry disproportionate weight in how engines answer SA questions. That is good news — the citation core is short, which means the tactical priorities are clear. It is also a vulnerability — if your brand is absent from BusinessTech, Moneyweb, MyBroadband, Hellopeter and your industry-specific publisher, you are absent from AI. The other difference is the complaint signal. SA AI answers lean harder on Hellopeter than US answers lean on equivalent platforms, which makes complaint management a much bigger AI-visibility lever locally.
What is the SA-AEO-Bench?+
The SA-AEO-Bench is the dataset behind everything on this page — the public South African benchmark that measures how ChatGPT, Claude and Gemini answer real SA buying questions across 100 brands and 12 industries. It is pre-registered on OSF (osf.io/w4az2), replicated quarterly, and the basis for both the per-brand Scorecard and the quarterly Index Report. Read the methodology at citedbrands.co.za/methodology.
How often does AI visibility need to be re-measured?+
Quarterly is the floor. AI engines retrain, the SA news cycle moves, and your competitors are also working on their visibility. A Scorecard read this quarter is a snapshot of where you stand now — useful for the board, the budget conversation and the next 90 days of work. The Index Report ships quarterly so the category context updates with you. If your brand is in active reputation work — a crisis, a launch, a leadership change — monthly is more appropriate.
What does a Cited Brands Scorecard actually deliver?+
Three things. First, your per-engine mention rate across ChatGPT, Claude and Gemini on the prompts your category cares about — so you know whether AI thinks of you at all when a customer asks a generic question. Second, the citation map — which SA domains are currently shaping the AI answer about you, and where the gaps are. Third, a prioritised action shortlist — the three patterns most likely to lift your visibility in the next quarter, ordered by leverage. The free Scorecard delivers the read. The 30-minute walkthrough turns it into a brief.
The next stops on the AI visibility journey.
Eight more pages worth reading. The two sibling pillars cover the head-term context. The blog pieces cover the tactical edges. The methodology page is the receipt for everything above.
AI Search in South Africa
The category-education head term. What AI search is, how it works in SA, why it matters.
AEO in South Africa
The practitioner head term. Answer Engine Optimisation — the playbook for getting cited in AI answers.
Where SA brands need to get cited in AI search
Six months into the bench — the SA citation map and the Hellopeter scatter that made us stop scrolling.
What is AEO? Answer Engine Optimisation explained
Short, dense, exact-match. The SA-context definition that AI assistants can lift cleanly.
Methodology
How the SA-AEO-Bench is built. Pre-registered on OSF, full protocol, replication code public.
The research dossier
The full reading from the SA-AEO-Bench — insights, briefs and downloadable artefacts.
AI visibility knowledge base
Tactical pieces, working definitions and case studies for the practitioner side of the brief.
The glossary
Plain-language definitions for every term used on this page and across the platform.
AI visibility is the category. Your brand is the question.
The free Scorecard runs the same SA-AEO-Bench measurement against your specific brand. Per-engine gap. SA sources shaping your AI answer. The three moves that matter most for you next quarter.
Or book a 30-minute walkthrough when you want the action plan, not just the read.