Generative Engine Optimisation (GEO): how Caribbean businesses get cited by AI

What GEO is, and how it differs from SEO

Search engine optimisation aims to win a position in a ranked list. Generative engine optimisation aims to win a mention inside a written answer. The two overlap heavily — both reward fast, crawlable, well-structured, genuinely useful pages — but the winning conditions differ. A ranked list rewards authority and relevance for a query. A generated answer rewards passages that can be retrieved, verified and safely quoted.

In practice this means an AI engine is looking for something it can lift: a clear definition, a specific figure, a named location, a dated fact, a structured list. Pages written as marketing atmosphere with no extractable claims are invisible to that process even when they rank well.

• SEO optimises for position; GEO optimises for citation.
• SEO rewards pages; GEO rewards passages within pages.
• SEO tolerates vagueness; GEO requires specific, checkable facts.
• Both depend on crawl access, speed, structure and trust signals.

Why this matters more in the Caribbean than in larger markets

Caribbean search volumes are small, which cuts both ways. Low volume means fewer competing pages, so a well-structured site can become the default source an AI engine reaches for when someone asks about villa rentals in Tortola, company formation in the BVI or dive operators in Grenada. But low volume also means the training and retrieval corpus is thin: if your business has not published clear facts about what it does and where, the engine will answer from a directory listing, an out-of-date aggregator or a competitor.

The visitor economy compounds this. Most enquiries begin off-island, weeks before travel, from people with no local knowledge and a strong habit of asking an assistant rather than scrolling results. Whoever the assistant names gets the enquiry.

The practical GEO checklist

Most of GEO is unglamorous engineering and disciplined writing. The list below is the order we work through it on client sites.

• Allow AI crawlers explicitly in robots.txt — GPTBot, OAI-SearchBot, ClaudeBot, Claude-SearchBot, Claude-User, PerplexityBot, Google-Extended and Bingbot.
• Publish an llms.txt at the site root summarising the organisation, services and key facts in plain machine-readable text.
• Answer first: open every page with a direct, quotable one-paragraph answer before any narrative.
• Mark up entities with schema.org — Organization, LocalBusiness, Service, FAQPage, Article and BreadcrumbList, linked by stable @id references.
• Keep name, address, phone and service descriptions identical everywhere they appear, on-site and off.
• Write real FAQs in the words customers use, with answers long enough to stand alone when quoted.
• Date and maintain content; engines discount pages with no freshness signal.
• Keep pages fast and server-rendered so retrieval does not depend on JavaScript execution.

How to measure GEO

You cannot rank-track an answer engine the way you track Google positions, so measurement is a mix of direct testing and server evidence. Ask the major assistants the questions your customers ask, monthly, and record whether you are named and whether the facts are right. Check server logs and analytics for referrals from chat.openai.com, perplexity.ai and copilot.microsoft.com, and for crawler hits from the AI user agents. Watch for the pattern that matters commercially: fewer, higher-intent visits that convert better than generic organic traffic.

Expect the same time horizon as SEO. Technical and structural fixes register within weeks; consistent citation across multiple engines usually takes three to six months of maintained, factual content.