What are GEO and AEO?
Generative engine optimisation (GEO) is the work of helping AI assistants such as ChatGPT, Gemini, Claude and Perplexity find a business and describe it correctly when they write an answer. Answer engine optimisation (AEO) is the narrower craft of shaping a page so a search or AI system can lift a direct answer from it. Both sit on top of ordinary SEO, not beside it.
The labels are new; most of the work is not. An assistant asked "which resort in Lachung can I book directly?" has to find a page that answers the question, trust that the page is current and match it to the right business. Every one of those steps depends on things a well-run website should already have.
How does an AI assistant find the sources it cites?
Most assistants answer from two places: what the model learned during training, and what it retrieves from the web when the question is asked. Retrieval is where a business has the most influence, because it reads live pages that can be corrected today.
Each AI company documents its own crawlers. OpenAI says its search crawler, OAI-SearchBot, decides which sites can appear in ChatGPT search answers, while GPTBot collects training data. Anthropic lists ClaudeBot for training, Claude-SearchBot for search and Claude-User for pages fetched when a person asks. Google says that to appear as a link in AI Overviews or AI Mode, a page only has to be indexed and eligible to show with a snippet in Google Search; there is no special markup for it.
So the first test is mechanical: can these crawlers reach the pages at all? A robots.txt rule, a firewall or a CDN's bot protection can block them without anyone at the business knowing. We found exactly that on this website in November 2026. Our own robots.txt allowed AI crawlers, but a setting at the hosting edge refused GPTBot and ClaudeBot and quietly added its own rules above ours.
What makes a source worth citing?
When several pages could answer a question, AI systems tend to use the one that answers it plainly, comes from a recognisable source and agrees with other sources. That favours:
- a page that states the answer in its first sentences, under a heading phrased the way people ask
- facts that match across the website, Google Business Profile, maps, directories and social profiles
- structured data that names the business, its address, its services and its people, and says nothing the page does not
- evidence a reader can check: named clients, live work, dates and specific details rather than adjectives
An answer engine cannot check a claim the way a customer can, so it leans on agreement between sources. When we tested AI answers about Floret Resorts in October 2026, the Lachung resort appeared only through travel-agency listings, which gave it 80 rooms and a 4 PM check-in. The resort has 15 rooms and a 12:00 check-in. The assistant repeated the most visible source, not the correct one.
Why do verified facts matter more than volume?
It is tempting to publish more pages full of claims and hope some are picked up. That backfires. An assistant that repeats an unverified price, licence number or result spreads the error to everyone who asks, and the business cannot recall it.
Our rule is simple. If a phone number, service detail, location fact or proof point cannot be verified, it is marked as an open gap with an owner. It does not go into page copy or structured data until it is confirmed. Once confirmed, the same fact is used everywhere: on the page, in the contact details, on the Google profile and in the markup.
What does the checklist look like?
Before making any claim about AI visibility, we work through the same order we use for search, from the bottom of the Visibility Playbook up:
- Access: the business controls its domain, hosting, Google profile and listings, and search and AI crawlers are not blocked by robots.txt, a firewall or the CDN.
- Facts: name, address, phone, hours and offer details are confirmed once and copied everywhere.
- Pages: each room, itinerary, project or service has its own crawlable page that answers the questions behind the decision.
- Markup: Organization or LocalBusiness, Service, Article and FAQ structured data matches the visible content, with sameAs links to the official profiles.
- Proof: reviews, real photographs and named work that a person or a model can check.
- Measurement: a fixed set of questions asked in ChatGPT, Gemini, Claude and Perplexity on a schedule, with the answers and cited sources recorded.
How do you measure AI visibility?
No single tool reports how a business appears across every assistant. What works is a repeatable question set: ten to twenty questions a real customer might ask, such as "which travel agents in Kalyani handle visas" or "who builds hotel websites in Darjeeling", asked the same way each month. Record whether the business is named, whether the facts are right and which pages are cited.
Treat any single answer as a sample, not a ranking. Responses vary between sessions and change as models are updated. Server logs and CDN analytics add a second view: which AI crawlers visited, how often and which pages they fetched.
Where this fits in the audit
AI-answer visibility is part of Layer 09 of the Visibility Playbook, reach and amplification, because it multiplies whatever sits below it. The Digital Footprint Audit checks crawler access, verified facts, offer pages, structured data and listings first. When those are sound, AI answers have something dependable to repeat.