In most industries, "get in early" is a motivational poster. In Mauritius right now, it is arithmetic.
When a generative engine like ChatGPT, Perplexity or Google's AI Overviews answers a question, it names a handful of businesses and stops. Not a page of results with a long tail of also-rans. Three or four names, with reasons. In a market like the UK or France, the fight to be one of those names involves thousands of competitors and a decade of accumulated content. In Mauritius, a category might hold a dozen serious players, and most of them have not published anything a model could quote if it wanted to.
That gap is the subject of this article. We have written a fuller research guide to generative engine optimisation in Mauritius covering the citation platforms and what a GEO partner does. This piece is about the part that keeps coming up in client conversations: what actually works, according to evidence rather than vendor blogs, and why the maths of a small market makes it work harder here.
The shift is measurable, so let's measure it
First, the honest picture of where AI search stands, because both the hype and the dismissal are wrong.
Google's AI Overviews now reach 2 billion monthly users across more than 200 countries, a figure Sundar Pichai gave on Alphabet's own earnings call. ChatGPT passed 800 million weekly users by OpenAI's October 2025 count, and it has offered full web search to everyone, no login required, since February 2025.
What that does to clicks is now well documented. Pew Research tracked real browsing behaviour across roughly 900 US adults and found that when an AI summary appeared, people clicked a result on 8% of visits, against 15% without one. Ahrefs measured a 34.5% drop in click-through rate for the top-ranking page when an AI Overview was present, across 300,000 keywords. And SparkToro's 2026 clickstream study found 68% of US Google searches now end without a click to the open web, up from 60% in 2024. Gartner's much-quoted 2024 prediction that traditional search volume would fall 25% by 2026 has not fully landed, but the direction it pointed at has.
Meanwhile the traffic that AI engines do send is small but growing violently: Similarweb counted 1.13 billion AI referrals to the top 1,000 websites in June 2025, up 357% year on year.
The practical summary: the recommendation increasingly happens inside the answer, whether or not the click ever happens. Your job is to be in the answer.
Why Mauritius specifically has an edge
Here is where the low-competition argument stops being a slogan and becomes a set of numbers.
The market is digitally ready but editorially empty. Mauritius has 1.01 million internet users, a 79.5% penetration rate, and mobile connections equal to 168% of the population, per DataReportal's Digital 2025 report. Consumers are online and mobile-first. The supply side has not kept up: the WTO's analysis of e-commerce in Mauritius notes that many smaller Mauritian businesses still run their entire web presence through Facebook rather than a site of their own. A Facebook page is largely invisible to a generative engine. The customers are reachable; the content competing to reach them barely exists.
SMEs are the economy, and SMEs have not started. SMEs make up over 90% of enterprises in Mauritius and around half of employment. Globally, GEO adoption among small businesses sits in single digits (NP Digital's survey of 350 marketing leaders put it at 8%; the figures are in our main GEO guide). Combine those and the conclusion writes itself: in almost every Mauritian category, nobody is doing this yet. The first serious mover is not fighting for share of the answer. They are defining it.
Generative engines are unusually kind to underdogs. This is the most under-appreciated finding in the founding academic paper on GEO (Aggarwal et al., presented at KDD 2024). When lower-ranked sites applied GEO methods, the paper measured the fifth-ranked site gaining up to 115% visibility while the top-ranked site lost around 30%. A generative engine reads your content and decides whether it is useful; it does not weigh a decade of backlinks the way classic rankings do. For a market full of small businesses that were never going to out-link a multinational, that is a genuine reset of the rules.
The bilingual advantage is real and almost free. Mauritians work in both English and French as a matter of routine. Those happen to be two of the largest question-asking languages fed into generative engines, and your two biggest buyer pools for tourism, property and services. A Mauritian business can publish credibly in both, from the same team, and be quotable to a model answering a question from London and one from Paris. Very few markets get that for free.
Your buyers are already asking machines. Mauritius welcomed 1,436,250 tourist arrivals in 2025, per Statistics Mauritius, with the fastest growth from India, Italy and Spain. Nearly all of those people researched from abroad, and the way they research is changing fast: Phocuswright found close to 40% of US travellers used generative AI to plan trips in 2025, an 11-point jump in a single year. A traveller in Mumbai asking ChatGPT for a quiet west-coast hotel is not browsing page two. There is no page two.
The secrets, in order of evidence
"Secrets" is a marketing word for "things that are published in research papers that nobody reads". Here is what the evidence actually supports, ranked roughly by how well it is proven.
1. Substantiate everything
The KDD 2024 paper tested nine ways of rewriting a page and measuring its visibility in generative answers across a 10,000-query benchmark. The winners were all forms of backing yourself up: adding relevant quotations, adding statistics, and citing credible sources. The best methods lifted visibility by 30 to 40% on the paper's primary metric. The loser was keyword stuffing, the oldest trick in SEO, which produced little to no improvement and actively hurt on some engines.
The mechanism matters more than the exact percentages, which came from 2023-era models. A language model assembling an answer needs material it can lift, attribute and trust. A paragraph carrying a named source and a specific number is more useful to it than a paragraph of confident assertion. If your service pages are three hundred words of adjectives, there is nothing there for a model to quote, and so no reason for it to mention you.
2. Write the way questions are asked
The a16z analysis of GEO makes a point most content strategies miss: prompts to ChatGPT average around 23 words, against roughly four for a classic search. People ask generative engines full questions with context, the way they would ask a knowledgeable friend. Content that wins is content shaped like a good answer: a direct response up front, clear structure, summaries, and enough depth behind it to survive follow-up questions. Google's own guidance for generative AI features says the same thing in fewer words: unique, genuinely useful, non-commodity content, fully indexable.
3. Be present where the models actually look
Your own site is necessary but not sufficient. Semrush's 13-week study of over 100 million citations and Profound's analysis of 680 million citations agree on the pattern: a handful of third-party platforms (Reddit, Wikipedia, YouTube, LinkedIn, review sites) carry an outsized share of what engines cite, and each engine has its own taste. Perplexity leans on Reddit. ChatGPT leans on Wikipedia. Google's AI Mode favours LinkedIn and YouTube.
Two cautions. First, these shares are volatile: Semrush watched ChatGPT's Reddit citation rate swing from roughly 60% to 10% within weeks, so never build on a single platform's current behaviour. Second, you cannot astroturf forums, and you should not try; it gets detected and deserves to. Participate honestly, under your real name, in your actual area of expertise.
4. Stay fresh, and tell the engines
Microsoft's Fabrice Canel, the principal product manager for Bing, confirmed that generative AI systems particularly value fresh content and recommended pushing updates through IndexNow. This matters more than it sounds for Mauritius: ChatGPT's search runs on Bing's index, so being well understood by Bing is quietly one of the highest-leverage technical moves available, and one almost no local business has made deliberately.
5. Put names and credentials on your content
Google's quality guidelines are built around E-E-A-T: experience, expertise, authoritativeness and trust, with trust explicitly called the most important of the four. For a generative engine trying to decide whether your claim about villa yields or offshore structuring is worth repeating, an article by a named person with a verifiable track record beats an anonymous "admin" post every time. In a small market this compounds: there are few enough credible named experts per category here that becoming one is an achievable project, not a lottery.
6. Keep the boring foundations, because they carry most of the weight
Google is blunt that standard SEO best practice is what feeds its AI features. Strong organic rankings buy you most of your Perplexity and AI Overviews visibility, since those engines draw heavily from the same sources as classic results. GEO is not a replacement for SEO. It is SEO judged by a new reader with different taste.
7. Know what not to buy
Two line items are being oversold across the industry right now, and you should hear it from an agency rather than discover it on an invoice.
llms.txt is a file proposed by Jeremy Howard in 2024 as a machine-readable map of your site. Google's documentation states plainly that it does not use llms.txt, and John Mueller has compared it to the long-dead keywords meta tag. It costs an afternoon and is a reasonable hedge on where AI agents are heading, so ship it, but pay nothing meaningful for it and never as a "citation strategy".
Schema markup for GEO sits in genuine dispute: Bing says structured data helps its language models understand content, while Google says it is not required for generative AI features. Keep your schema for rich results and for Bing's sake, but the safe rule is simpler: if you want a model to know something, put it in visible text on the page, not only in markup.
The tailwind nobody is using
There is a final, distinctly Mauritian reason to move now: the state is pushing in the same direction, and almost no business is riding it.
Mauritius was one of the first African countries to publish a national AI strategy, back in 2018. The Digital Mauritius 2030 plan targets lifting ICT to 10% of GDP, training 10,000 AI specialists and embedding AI in half of public services. The government has even launched a national survey of business AI adoption to measure progress. Across Africa more broadly, 64% of workers report having used AI in the past year, ahead of the global average.
The infrastructure, the policy and the audience are all moving. The content is not. That mismatch is the opportunity, and it will not stay open. Windows like this close the way they always do: slowly, then all at once, when the second and third movers in each category force everyone else to catch up from behind.
What to do with this
If you strip the acronym away, GEO asks a question that was always fair: is there anything published about your business worth repeating? The difference is that the thing deciding now reads everything, forgets nothing, and answers your buyer directly, in English and in French, from Port Louis to Paris.
For a Mauritian business the sequence is short. Find out what the models currently say about you and your category. Identify the questions your real buyers ask, in both languages. Publish substantiated, answer-shaped material against them, on your site and on the platforms the engines actually cite. Then keep it fresh and keep the receipts, because the first credible source in a small category tends to stay the source.
That audit is where we start with every client. Read about our approach to SEO for AI engines, dig into the full research guide to GEO in Mauritius, or book a strategy call and we will show you what the models say about you right now.
Sources. Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. Google Search Central, AI features and your website and Optimizing for generative AI features. Pew Research Center, AI summaries and click behaviour (July 2025). Ahrefs, AI Overviews reduce clicks (April 2025). SparkToro, 2026 zero-click study. Semrush, most cited domains in AI search (2025). Profound, AI platform citation patterns. a16z, How GEO rewrites the rules of search (May 2025). TechCrunch on AI Overviews reach and AI referral growth. Search Engine Land on Bing, schema and LLMs and Gartner's 25% prediction. DataReportal, Digital 2025: Mauritius. Statistics Mauritius, International travel and tourism, year 2025. MCCI, SME Competitiveness Outlook. WTO, E-commerce in Mauritius. Phocuswright, Search slips, AI surges. Digital Watch Observatory on the Mauritius AI Strategy and Digital Mauritius 2030. TechCabal, AI use among African workers (December 2025).