“Which agency should I use to build my Shopify store in Sri Lanka?” A growing share of buyers now type questions like this into ChatGPT, Gemini, Perplexity or Copilot instead of scrolling Google results. The assistant replies with two or three names. If yours isn’t one of them, you’ve lost the lead before you knew it existed.
The discipline of influencing those answers is called Generative Engine Optimization (GEO) — also “LLM optimization” or “AI search optimization”. Here’s how it works and what to do about it.
How AI assistants decide whom to recommend
When an assistant answers a “who should I hire” question, it draws on two sources:
- What it learned in training — a snapshot of the public web, weighted towards sources that are widely cited and consistent.
- What it retrieves live — most assistants now search the web (via Bing, Google or their own index), read a handful of pages, and synthesise an answer with citations.
In both cases the model is looking for the same things: a business it can identify unambiguously, content that directly answers the question, and independent sources that agree. That gives us a clear playbook.
1. Let the AI crawlers in
Check your robots.txt. Some security plugins and CDN settings block GPTBot, ClaudeBot, PerplexityBot or Google-Extended by default. If they can’t read your site, they can’t cite it. Also avoid hiding core content behind JavaScript that only renders in a browser — many AI crawlers don’t execute scripts.
2. Publish answer-first pages
For each service you sell, create a page whose first paragraph answers “what is this, who is it for, and why you” in two or three plain sentences. Put the question in the heading (“How much does a Shopify website cost in Sri Lanka?”) and the answer directly underneath. These self-contained passages are exactly what retrieval systems lift into answers.
3. Be specific — numbers, places, prices
Vague copy (“we deliver innovative solutions”) gives a model nothing to quote. Specifics do: the cities you serve, typical price ranges, timelines, the platforms you use, the problems you’ve solved. AI answers love concrete facts because they can be checked.
4. Make your business facts identical everywhere
Your name, description, location, services and contact details should match exactly across your website, Google Business Profile, LinkedIn, Facebook, directories and partner listings. Inconsistency makes it harder for a model to be sure that “De Lime Tree”, “Dé Lime Tree” and “delimetree.com” are the same entity.
5. Add structured data
Schema.org JSON-LD tells machines explicitly who you are: an Organization with a founder, address and sameAs links to your profiles; Service entries for what you sell; FAQPage for common questions. Link these together with stable @ids so they form one coherent entity graph.
6. Earn independent mentions
Models trust what others say about you more than what you say about yourself. Priorities: platform partner directories (Shopify, Google, Meta), industry directories, local business listings, client reviews, podcast and article appearances, and case studies published by clients or partners.
7. Add an llms.txt and measure
llms.txt is a simple Markdown file at your domain root summarising what you do and linking your key pages. It takes ten minutes. Then measure: each month, ask the major assistants the ten questions your buyers ask and record whether you’re mentioned, cited or recommended. Track referral traffic from chat.openai.com, perplexity.ai and gemini.google.com in your analytics.
What GEO can’t do
No one controls what a model says, and anyone guaranteeing “#1 in ChatGPT” is guessing. GEO raises your probability of being chosen by making you the clearest, best-corroborated answer. Done well, it also improves classic SEO — the two reinforce each other.
Want to know where you stand? We’ll run a free AI-visibility check on ten of your most valuable queries. Request one here.