Who AI actually recommends in South Africa.
The Gr8 Index is the first South African AI-visibility benchmark measured to a published standard. We asked the four leading AI assistants — ChatGPT, Claude, Gemini and Perplexity — the questions real buyers ask, eight times each, and recorded how often every brand was named. This is the result, shipped with its questions, its run count and its margin of error. Nothing here is a single lucky reading.
What the Gr8 Index is
The Gr8 Index is an independent analysis of information generated by publicly accessible AI assistants. It records how often each named South African brand was mentioned when the four leading assistants — ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google) and Perplexity — were asked the questions real buyers ask, logged out and with no personalisation, on the dates stated. It is an analysis of independently sourced, reproducible information — not a set of allegations by GR8ER2 about the institutions named.
It draws on four categories of source, each referenced in full under Sources & references: the four AI assistants' own public interfaces and APIs, queried under stated conditions; the GR8ER2 AI Visibility Standard and the Gr8 Index method of record, which govern how every figure is produced; the public regulatory registers used to confirm that each institution named is correctly identified and categorised; and GR8ER2's own measurement, published on the same rule as every other brand.
"Named" and "recommended" here mean only that an assistant mentioned the brand in answer to a buyer's question — a measure of visibility, not an endorsement of the brand by the assistant or by GR8ER2, and not a judgement of product quality.
This publication has been prepared from publicly available information — the outputs of publicly accessible AI assistants, measured under the method described and referenced herein. Its sources are identified wherever practicable, and every figure is reconstructable from logs retained by GR8ER2 under that method.
The figures are a point-in-time measurement of how often each brand was named on the dates given. AI outputs vary with time, phrasing, location and personalisation, and may differ from what any individual user sees. The figures are not GR8ER2's assessment of any institution's products, services, financial standing or conduct, are not a ranking of quality, and are not financial, legal or purchasing advice. Any analysis or commentary represents GR8ER2's interpretation of the measured information and should be read in that context. Brand and institution names are used only to identify the entities measured; no affiliation or endorsement is implied.
What these numbers are — and what they are not
Every figure on this page is a measurement of what AI assistants said — ChatGPT, Claude, Gemini and Perplexity — when asked buyer questions about South Africa, logged out, with no personalisation, across 1–2 October 2026. It records how often each brand was named by the engines. It is not GR8ER2's opinion of any brand, and it is not a ranking of product quality. Each score carries its margin of error; where two brands' ranges overlap, we show them as a tie, never as a false gap.
The Gr8 Score is a single number from 0 to 100, built from three measured components and always reported with all three shown, so anyone can re-weight them:
- Mention Rate (MR) — how often the brand is named at all.
- Share of Voice (SoV) — of the brands named together, how much of the attention is this one's, weighted by where it appears in the answer.
- Citation Rate (CR) — how often the brand's own website is cited as a linked source. N/A means no citation to its domain was found anywhere in the data — not a measured zero.
The composite is 0.40·MR + 0.35·SoV + 0.25·CR — a published, fixed weighting, frozen for this version. Every score below is High-confidence: four engines, eight runs each. The full list of questions, the full brand list, and the brands that fell just below the line are published in the Standard and the method of record.
Retail banking
When South Africans ask an AI assistant which bank to use, these are the banks it names, and how often. The institutions listed are registered banks on the Prudential Authority register.[7]
FNB and Standard Bank follow. Below them, five banks sit in a statistical tie: the engines do not clearly separate them.
| # | Brand | Gr8 Score | 95% interval | MR | SoV | CR |
|---|---|---|---|---|---|---|
| 1 | Capitec | 57 | 54.3–59.2 | 95.6% | 25.6% | 38.2% |
| 2 | First National Bank | 45 | 42.5–47.0 | 87.7% | 16.4% | 15.8% |
| 3 | Standard Bank | 39 | 37.0–41.0 | 73.0% | 10.6% | 24.4% |
| 4 | Nedbank | 33 | 30.9–35.0 | 71.0% | 6.2% | 9.7% |
| 5 | Absa | 32 | 28.4–36.4 | 59.6% | 7.7% | 23.5% |
| 6 | Discovery Bank | 30 | 27.6–32.4 | 57.7% | 6.8% | 18.2% |
| 7 | TymeBank | 30 | 26.0–33.2 | 62.3% | 10.4% | 4.2% |
| 8 | Bank Zero | 24 | 20.4–26.9 | 50.0% | 6.1% | 6.1% |
| 9 | GoTyme Bank | 13 | 9.3–16.4 | 25.4% | 4.0% | 5.2% |
| 10 | Pick n Pay | 12 | 7.6–17.0 | 28.1% | 3.0% | N/A |
| 11 | Boxer | 12 | 7.4–16.5 | 27.9% | 2.2% | N/A |
| 12 | Investec | 7 | 5.1–9.2 | 16.0% | 1.0% | 1.5% |
Statistical ties (overlapping intervals): Nedbank = Absa = Discovery Bank = TymeBank = Bank Zero · GoTyme Bank = Pick n Pay = Boxer = Investec. Pick n Pay and Boxer are retailers offering banking services, surfaced by the engines and included under the category rule. CR is N/A where no citation to a brand's own domain was found.
Medical aid / health insurance
The schemes AI assistants name when South Africans ask which medical aid to join. The institutions listed are schemes registered with the Council for Medical Schemes.[8]
Bonitas and Momentum Health follow clearly; the mid-field tightens into ties from fourth.
| # | Brand | Gr8 Score | 95% interval | MR | SoV | CR |
|---|---|---|---|---|---|---|
| 1 | Discovery Health | 62 | 59.9–64.5 | 96.7% | 37.0% | 42.3% |
| 2 | Bonitas | 52 | 49.7–54.1 | 93.4% | 21.1% | 28.7% |
| 3 | Momentum Health | 43 | 40.5–46.0 | 80.5% | 13.6% | 25.2% |
| 4 | Bestmed | 35 | 31.1–38.8 | 70.2% | 11.2% | 11.6% |
| 5 | Fedhealth | 30 | 27.6–32.3 | 64.7% | 7.5% | 5.7% |
| 6 | Medihelp | 21 | 19.1–23.6 | 44.7% | 4.0% | 8.5% |
| 7 | KeyHealth | 11 | 6.5–15.2 | 17.8% | 1.2% | 13.2% |
| 8 | Medshield | 10 | 7.9–12.8 | 22.4% | 1.5% | 3.5% |
| 9 | GEMS | 8 | 5.6–10.7 | 16.2% | 2.1% | 3.7% |
| 10 | Profmed | 5 | 3.6–6.6 | 11.8% | 0.7% | 0.6% |
Statistical ties: Bestmed = Fedhealth · KeyHealth = Medshield = GEMS = Profmed.
Short-term insurance (car & household)
The insurers AI assistants name when South Africans ask about car and home cover. The institutions listed are licensed insurers on the Prudential Authority register (banks appear where they are licensed to offer short-term insurance, under one consistent category rule).[9]
A long tie forms behind it: from King Price down to Budget Insurance, the engines do not clearly separate the field.
| # | Brand | Gr8 Score | 95% interval | MR | SoV | CR |
|---|---|---|---|---|---|---|
| 1 | OUTsurance | 58 | 56.0–59.4 | 93.9% | 24.7% | 46.0% |
| 2 | King Price | 39 | 35.9–42.2 | 76.7% | 11.7% | 17.1% |
| 3 | Santam | 37 | 33.8–41.1 | 70.8% | 14.6% | 16.0% |
| 4 | Discovery Insure | 31 | 27.9–33.9 | 66.0% | 8.7% | 5.7% |
| 5 | MiWay | 30 | 26.4–32.8 | 64.2% | 6.6% | 6.6% |
| 6 | Naked Insurance | 28 | 23.3–32.7 | 59.6% | 8.6% | 4.8% |
| 7 | Auto & General | 21 | 18.3–24.3 | 39.9% | 5.6% | 13.6% |
| 8 | Pineapple | 21 | 15.9–25.8 | 33.3% | 3.8% | 24.8% |
| 9 | Budget Insurance | 19 | 15.9–22.4 | 42.3% | 4.2% | 3.1% |
| 10 | Momentum Insure | 13 | 10.7–15.6 | 25.4% | 2.1% | 9.2% |
| 11 | Old Mutual Insure | 13 | 11.0–14.5 | 27.4% | 2.8% | 3.3% |
| 12 | Hollard | 12 | 9.5–14.3 | 26.1% | 2.1% | 2.9% |
| 13 | Dialdirect | 11 | 8.0–13.1 | 24.8% | 1.4% | 0.4% |
| 14 | First for Women | 10 | 8.5–12.3 | 23.0% | 1.5% | 2.8% |
| 15 | Virseker | 4 | 3.0–4.9 | 9.2% | 0.8% | N/A |
| 16 | Absa | 3 | 2.1–4.4 | 7.0% | 0.5% | 1.1% |
| 17 | Prime South Africa | 1 | 0.4–1.6 | 2.2% | 0.3% | 0.0% |
Statistical ties: King Price = Santam = Discovery Insure = MiWay = Naked = Auto & General = Pineapple = Budget Insurance · Momentum Insure = Old Mutual Insure = Hollard = Dialdirect = First for Women · Virseker = Absa. Banks appear here where they genuinely offer short-term insurance to SA consumers, under one consistent category rule.
AI-visibility / digital-marketing / SEO agencies
Our own category. We measure it to exactly the same rule as the others — and we report where GR8ER2 itself landed, below.
Our readingBehind the top two, the field is long and flat: twenty-six agencies sit within one tie band. In our interpretation that spread is itself a finding — AI has no settled answer to "best agency" in South Africa yet.
| # | Brand | Gr8 Score | 95% interval | MR | SoV | CR |
|---|---|---|---|---|---|---|
| 1 | Woww | 56 | 48.6–62.6 | 74.5% | 22.0% | 72.4% |
| 2 | MO Agency | 52 | 46.0–58.7 | 69.3% | 23.0% | 66.2% |
| 3 | Ruby Digital | 22 | 14.1–29.0 | 32.2% | 6.6% | 25.4% |
| 4 | Algorithm Agency | 20 | 14.3–26.3 | 29.2% | 5.2% | 27.2% |
| 5 | Easy Rank | 20 | 12.6–27.4 | 27.0% | 2.8% | 32.9% |
| 6 | Syte | 19 | 9.1–28.3 | 26.3% | 5.8% | 24.4% |
| 7 | Design Zeen | 18 | 13.1–23.3 | 21.7% | 3.8% | 32.9% |
| 8 | Honey Whale | 17 | 13.6–19.9 | 25.6% | 5.2% | 18.9% |
| 9 | Fast Forward Marketing | 12 | 7.5–17.4 | 14.5% | 1.2% | 24.8% |
| 10 | BlueMagnet | 12 | 9.4–14.5 | 20.2% | 2.6% | 11.9% |
| 11 | Macrocosm | 12 | 7.6–16.2 | 22.6% | 2.0% | 8.6% |
| 12 | iBoost | 11 | 6.7–15.9 | 13.1% | 0.9% | 23.2% |
| 13 | BrandAxis | 11 | 7.6–14.2 | 23.9% | 3.4% | 0.7% |
| 14 | Le Roux Digital | 8 | 4.7–11.6 | 10.3% | 3.9% | 10.7% |
| 15 | Juicy Designs | 7 | 4.8–8.9 | 7.7% | 1.0% | 13.6% |
| 16 | Black & Yellow | 6 | 4.5–8.2 | 7.3% | 1.4% | 11.8% |
| 17 | Vetro Media | 6 | 3.0–9.3 | 8.6% | 1.2% | 9.0% |
| 18 | Prebo Digital | 5 | 3.4–7.3 | 9.2% | 0.7% | 5.9% |
| 19 | Gridweb | 5 | 2.2–8.5 | 7.7% | 0.6% | 8.3% |
| 20 | Click2Flow | 5 | 3.2–6.7 | 6.8% | 0.8% | 7.7% |
| 21 | doubleBaRRiL | 5 | 2.6–7.0 | 7.5% | 2.2% | 4.0% |
| 22 | Launchcraft | 5 | 2.4–6.7 | 8.6% | 0.6% | 3.7% |
| 23 | Techsys | 4 | 2.2–5.0 | 6.1% | 0.4% | 4.0% |
| 24 | SEOPros | 3 | 2.5–4.2 | 5.3% | 0.8% | 3.7% |
| 25 | Rogerwilco | 2 | 1.6–3.3 | 5.2% | 0.8% | 0.6% |
| 26 | Simply Ecommerce | 2 | 0.8–3.9 | 3.1% | 0.3% | 3.9% |
| 27 | Top Click Media | 2 | 1.0–2.3 | 2.8% | 0.5% | 1.5% |
| 28 | Incubeta | 2 | 0.9–2.1 | 3.5% | 0.3% | N/A |
Woww and MO Agency form the top tie (T1). Ranks 3–28 sit within a single wide tie band (T2): the engines do not reliably separate them, so we do not pretend to. The full below-threshold tail is published in the method of record.
Where GR8ER2 lands in its own Index
The Standard requires GR8ER2 to measure itself every cycle, to the same rule as any brand — including when the result is bad. It is. We publish it anyway, because a benchmark whose author exempts itself is not a benchmark.
Named in 0 of 544 open-question answers. When South African buyers ask AI for an agency, GR8ER2 is not yet named. This is the honest headline — and our starting line.
When asked directly "what is GR8ER2", it is correctly identified 19.8% of the time (human-coded from all 96 branded answers). Branded Gr8 Score: 20.
Our readingThe engine split is the real story. Asked by name, only one of the four engines reliably identified GR8ER2:
ChatGPT 54.2% · Perplexity 25.0% · Claude 0.0% · Gemini 0.0%
Two engines have never heard of us. This is exactly the gap we help clients close — now measured, publicly, on our own name. Next cycle, under identical protocol, we publish the movement.
GR8ER2's branded figure was coded by hand because the automated check mis-read it: the dominant failure was misidentification, not absence — two engines confidently described unrelated things with similar names. The human coding (19.8%, sensitivity 18–23%) is the authority for this figure. Raw branded answer text is not published.
How the Gr8 Index is measured
Everything here follows the GR8ER2 AI Visibility Standard. In short:
- Four engines, eight runs each. ChatGPT, Claude, Gemini and Perplexity, each queried through its official API with web search on, each question asked eight times — the independent academic floor for a stable reading. Engines are reported separately, never silently blended.
- A published question set. For each category, 17 shared, brand-neutral buyer questions drive the league table. The full set is published and frozen for the cycle.
- An engine-chosen brand list. We did not pick who is in each category. The questions themselves surfaced the field; any brand named often enough, on enough engines, is included by a published rule — and the brands that fell just short are published too.
- Every number traceable. No figure is invented; non-appearances and zeros are published, not hidden. Each score ships with its margin of error.
Our own score, including when it's zero. The error bar on every number. The exact run count. The full question list and brand list. Most AI-visibility numbers in this market are a single reading with none of these — a snapshot wearing a measurement's clothes.
Disclosed limits of this cycle
- Google AI Overviews / AI Mode are not in v1. They have no official API; they arrive in v1.1 via a licensed data channel, published as a dated panel change rather than blended into these numbers.
- Gemini could not be region-pinned by parameter — it was pinned by in-prompt framing only, and this is logged per answer. The other three engines accept a South-Africa location setting.
- English only this cycle; other South African languages are a planned expansion.
- Scope: four categories. More follow in later cycles.
Where every figure comes from
The Index follows one consistent referencing approach. Every measured figure derives from the AI-assistant outputs at [1]–[4], produced and scored under the governing documents at [5]–[6]; the institutions named in each category are real entities on the public regulatory registers at [7]–[9]. Each figure is reconstructable from the logs retained under the method of record.
- OpenAI — ChatGPT. Responses measured via the OpenAI API (model gpt-5.4-mini, web search enabled, location South Africa), 1–2 October 2026. openai.com
- Anthropic — Claude. Responses measured via the Anthropic API (model claude-sonnet-5-5, web search enabled, location South Africa), 1–2 October 2026. anthropic.com
- Google — Gemini. Responses measured via the Google Generative Language API (model gemini-3.8-flash, Google Search grounding; region set by in-prompt framing, as disclosed in the method), 1–2 October 2026. ai.google.dev
- Perplexity. Responses measured via the Perplexity API (model sonar, web search enabled, location South Africa), 1–2 October 2026. perplexity.ai
- GR8ER2 AI Visibility Standard, v1.1 (GR8ER2, 2026). The governing method: scoring, confidence intervals, tie rules and reporting rules. gr8er2.com/standard
- The Gr8 Index — method of record, v1 (GR8ER2, 2026). The full prompt corpus, the engine-chosen brand-universe rule, run conditions, per-engine detail and the full below-threshold tail for this cycle. gr8er2.com/index/method
- Prudential Authority, South African Reserve Bank — SA registered banks and representative offices. Used to confirm the banks named are registered institutions. resbank.co.za
- Council for Medical Schemes — registered medical schemes. Used to confirm the schemes named are registered under the Medical Schemes Act 131 of 1998. medicalschemes.co.za
- Prudential Authority, South African Reserve Bank — list of registered insurers. Used to confirm the short-term insurers named are licensed insurers. resbank.co.za (PDF)
The AI-visibility / digital-agency category has no single statutory register; its field was derived entirely by the engine-chosen brand rule in [6], and the full below-threshold tail is published there. Regulatory registers are used only to confirm the identity and category of the institutions named; they are not the source of any score, and the Index does not draw on, endorse or reflect those regulators.