How South African restaurant brands can take advantage of AI search
Limited specialist SA restaurant publishers — international food-delivery review sites take the citation share.
SA-AEO-Bench v1.2 — pre-registered on OSF before data collection. 19,020 successful AI responses across GPT-5, Claude Sonnet 4.5 and Gemini 2.5 Pro. 232,254 cited URLs classified against a published source taxonomy. 100 South African brands, 12 industries. Re-run quarterly. Methodology and analysis code public at osf.io/w4az2.
This page is one industry’s view of that dataset. The per-brand cut is in the Scorecard, the full per-industry tables in the Q2 Index Report.
The pattern.
When the question is about eating out in SA, AI looks abroad first. TripAdvisor reviews, Google Maps tagging and global travel-publication round-ups carry disproportionate weight on Cape Town, Joburg and Durban dining queries. The engines reach for international sources before SA-specialist ones because the global corpus is simply larger. Eat Out partially closes the gap — it is the SA-specialist authority every engine eventually reaches for — but its coverage tiers (Top 10, Top 100, Hot 100) shape some queries cleanly and leave others to the global default. Getaway, thingstodo.co.za and BusinessTech round out the citation pool on lifestyle and category-news questions.
Delivery splits the answer in two. Mr D Food and Uber Eats are the two brands AI engines reach for on "best food delivery in SA" by reflex. The split between them is closer than most marketing teams assume, and the per-engine treatment matters more than overall share. Gemini favours Mr D on Hellopeter signal because Uber Eats SA carries higher complaint volume on driver and refund threads. ChatGPT favours Uber Eats on global brand corpus. Claude hedges. Perplexity privileges whatever ran last in BusinessTech or MyBroadband. Same query, four genuinely different answer sets.
The chain-versus-independent split decides everything else. Famous Brands chains — Steers, Wimpy, Mugg & Bean, Debonairs, Fishaways — show up confidently on category queries because they generate consistent BusinessTech and lifestyle-press coverage. KFC SA, McDonald's SA, Nando's and Spur sit in the same default tier. Fine dining wins on Eat Out and on global travel-publication coverage — La Colombe, FYN, Test Kitchen are the names AI reaches for on "best restaurant in Cape Town" almost universally. Independent neighbourhood restaurants show up only when the prompt is suburb-specific, and they win or lose based on whether they have made it into an Eat Out tier or a Getaway round-up. The per-engine breakdown is in the SA AI Visibility Index Report. The brand-specific picture is in the free Scorecard.
Where AI gets its info.
Three to five SA sites carry most of the weight in restaurants. The names matter. The counts are reserved for the Index Report.
Eat Out is the closest thing SA has to a single restaurant authority, and AI engines treat it as the SA-specific reference on fine-dining and "best restaurant" queries. Inclusion in an Eat Out tier (Top 10, Top 100, Hot 100) lifts a restaurant materially across all four engines. A restaurant absent from Eat Out starts every quality-led answer behind the restaurants that made the tier.
Getaway carries the lifestyle and travel-led framing of SA dining — wine routes, regional food trails, hotel restaurants, weekend-away picks. The engines reach for it on tourism-shaped queries. A restaurant featured in Getaway round-ups shows up in AI answers about Cape Town, Stellenbosch, Franschhoek and the Garden Route disproportionately to its size.
thingstodo.co.za carries broad category round-ups — best burgers in Joburg, best pizza in Cape Town, best sushi in Durban — that AI engines reach for on category-specific dining queries. A restaurant featured here on a category list lands in answers where Eat Out tiers do not reach. The freshness layer matters: lists older than six months decay out of Perplexity and Gemini quickly.
The Q2 Index Report names every SA domain we measured for restaurants — the full source taxonomy classified by role (default authority, freshness layer, complaint signal, long-tail specialist) with per-engine citation share for each one. The long-tail is where the cheap wins live. Download the Q2 Index →
For the source map specific to your brand — which domains currently shape AI’s answer about you — run the free Scorecard.
How the engines differ.
Each AI reaches for restaurants sources differently. One strategy almost never works across all four. The per-engine picture for your brand is in the Scorecard.
ChatGPT defaults to global review platforms on SA dining queries — TripAdvisor and global travel publications dominate the corpus it learned from. La Colombe, FYN and Test Kitchen show up by reflex on Cape Town fine-dining queries. Famous Brands chains show up on category queries. Independents and neighbourhood favourites show up only when prompted by suburb or category. Worst engine for SA-specialist visibility.
Claude weighs Eat Out coverage more carefully than the other engines and is the most likely to name a SA-specialist restaurant ahead of a TripAdvisor-anchored international tourist favourite. Best engine for "best SA restaurant" intent when the user wants a local answer. Weakest for transactional "where do I eat tonight" intent.
Gemini pulls Hellopeter complaints, Google Maps review density and Reddit threads the other engines ignore. It elevates neighbourhood restaurants more confidently on suburb-specific queries and is the most dangerous engine for a delivery brand with active complaint volume. The engine that shapes Google AI Overviews — which is what most SA orderers and diners actually see.
Recent thingstodo.co.za round-ups, Getaway features, BusinessTech chain-news stories — Perplexity catches them first. A restaurant featured in the last fortnight reads as the current authority. A restaurant that has gone quiet decays out of this engine first. The Eat Out annual award cycle reshapes Perplexity fine-dining answers across SA every year.
Two ways restaurants brands play this.
Brands in this category split into two camps. Which camp you are in tells you what AI-visibility risks you actually carry.
The chain default
Famous Brands chains, KFC SA, McDonald's SA, Nando's, Spur. These brands have built AI visibility through a steady drumbeat of BusinessTech and lifestyle-press coverage — pricing changes, menu launches, franchise stories — that trains the engines to treat them as the default category answer. Hard to dislodge from "best burger chain in SA" or "best pizza chain in SA" intent. The exposure is on the delivery side, where Hellopeter signal moves Gemini against chains with persistent service-complaint volume, and on the independent side, where the chains lose suburb-specific neighbourhood-favourite queries.
The fine-dining anchors
La Colombe, FYN, Test Kitchen, Restaurant Mosaic, Wolfgat. These restaurants have anchored SA in global English-language food coverage — World's 50 Best, Michelin-adjacent international press, Eat Out's top tier. AI engines reach for them on "best restaurant in SA" by reflex across all four engines, and the global corpus depth makes the anchor extremely durable. The risk is silence: a quarter of light editorial and Eat Out tier movement can shift the ordering, and Perplexity in particular drops fine-dining names that have not generated coverage in the freshness window.
What changed this quarter.
Questions to take to your next meeting.
Each one is specific. Each one is answerable. We answer them for your brand in the Scorecard and for the category in the Q2 Index Report.
- 01When AI engines answer "best restaurant in Cape Town" or "best restaurant in Joburg", does your name get into the answer or stay in the qualifying paragraph?
- 02On the Mr D vs Uber Eats question, which engine favours your brand and which one is currently losing it for you?
- 03Where does your brand sit between ChatGPT (TripAdvisor-default, global corpus) and Gemini (Hellopeter-aware, Maps-sensitive)? The gap is almost never small.
- 04Which Hellopeter delivery or service thread is currently shaping Gemini's answer about your restaurant or delivery brand? Have you ever read it?
The Q2 Index Report includes the full diagnostic question set for restaurants — every meeting-ready question with the answer cross-referenced to the per-engine measurement. Plus the SA brands currently ahead, behind, or improving fastest on each one. Download the Q2 Index →
For the answer specific to your brand — where you sit on each question, on each engine — run the free Scorecard.
What to do about it.
Three categories of move lift restaurants AI visibility. Which specific tactic is right for your brand depends on where you stand today — that conversation lives in the 30-minute walkthrough.
Audit your placement across the five SA restaurant sites.
Eat Out, Getaway, thingstodo.co.za, BusinessTech and Hellopeter form the citation core. A brand cleanly placed on the first four and quiet on the fifth reads very differently to AI than one with delivery or service-complaint volume. The Scorecard names where your brand stands on each, and which placements have decayed since last quarter.
Quantify the global-versus-local default before you spend a rand.
TripAdvisor and global travel publications swamp SA-specialist coverage on ChatGPT in particular. Awareness spend will not fix the default — only fresh SA-specialist coverage on Eat Out, Getaway and the lifestyle press will. The Index Report quantifies the gap across the sector. The Scorecard quantifies it for your brand.
Re-check what AI thinks your restaurant or chain actually is.
AI engines confuse chain locations, sibling brands and Famous Brands subsidiaries more often than the sector tracks. They also confuse SA independents with foreign namesakes on common restaurant names. Wrong entity, wrong answer. The brand-specific confusions for your nameplate are in the Scorecard, with the impact of each one.
Brand-level scorecards for restaurants land in the next Index Report. Reserve your copy →
Common questions,
answered straight.
Why does ChatGPT keep recommending TripAdvisor-style answers for SA restaurant queries?+
ChatGPT learned from a body of global travel-and-food coverage where TripAdvisor, Yelp and international food blogs dwarf any SA-specialist source. The model defaults to the corpus it knows best. Claude behaves more carefully because it weights Eat Out coverage more heavily. Gemini partially closes the gap through Google Maps review density and Reddit signal. The full per-engine breakdown is in the Q2 SA AI Visibility Index Report.
Does the Mr D vs Uber Eats answer actually differ across engines?+
Yes — and the difference is bigger than the marketing teams assume. Gemini favours Mr D because it reads Hellopeter complaint signal against Uber Eats SA. ChatGPT favours Uber Eats because the global brand corpus is larger. Claude hedges, often naming both with caveats. Perplexity follows whichever brand last appeared in BusinessTech or MyBroadband. The per-engine treatment of the question is in the Index Report.
How much does an Eat Out tier actually move AI answers about my restaurant?+
Materially, across all four engines, but most strongly in Claude and Perplexity. Eat Out is the only SA-specialist source the engines treat as a quality authority, and a tier inclusion lifts a restaurant in "best restaurant in [city]" answers immediately. A drop out of a tier falls just as fast. The Scorecard names where your restaurant sits in the engines' current Eat Out-anchored set.
How fast does AI-visibility data go stale in SA restaurants?+
Faster than most consumer sectors on category queries (best burger, best pizza, best sushi) and slower on fine-dining ones. Perplexity drops you within roughly two months of a quiet round-up cycle. Gemini holds a bit longer. ChatGPT and Claude decay over many months on the structural questions. Around the annual Eat Out cycle all four refresh aggressively. The freshness state for your specific brand is in the Scorecard.
Where do you get this data?+
We run roughly four to six thousand structured prompts per quarter across ChatGPT, Claude, Gemini and Perplexity. Five replications per prompt. Latin-square order-reversal to control for position bias. Methodology pre-registered on OSF. See the full methodology for details.
Keep going.
Where SA brands need to get cited
The cross-industry SA-domain priority list for AI visibility.
Zero-click search impact on SA retailers
How AI search is reshaping referral traffic in SA consumer categories.
Methodology — SA AEO Bench
Pre-registered measurement design and source taxonomy.
Restaurants is the category.
You are the question.
The free Scorecard runs the same restaurants measurement against your specific brand. You get the per-engine gap, the SA sites currently shaping your AI answer, and the three moves that matter most for you.
- Dataset: sa-aeo-bench-v1
- Snapshot: 2026-05-19 · 17 days old · budget 270 days
- Pre-registration: osf.io/w4az2
- 14,826 responses · 188,877 citations · 100 brands · 10 industries
- POPIA-compliant · Information Officer registered
- Data residency: SA + EU only
- Methodology: pre-registered before data collection · audit trail per record
- Models measured: OpenAI GPT-5 · Anthropic Claude Sonnet 4.5 · Google Gemini 2.5 Pro