What is retrieval-augmented generation (RAG)?
Retrieval-augmented generation (RAG) is an architecture pattern used in AI systems where, before generating a response, the system retrieves relevant information from an external source (typically the live web or a database) and incorporates that information into its answer.
The “retrieval” step is a search: the system queries an index of web content to find pages relevant to the user’s question. The “augmented generation” step is the AI composing its response using both its training knowledge and the retrieved content.
What are the two types of AI search tools?
The major AI tools split into two types that behave very differently: training-data-only systems like base ChatGPT, and RAG-enabled systems like Perplexity and browsing-enabled ChatGPT.
Training-data-only systems (like base ChatGPT): These tools know only what they learned during training. If your hotel was not in their training data, or was described poorly, no amount of recent website updates will change what they say about you. Changes only take effect after the next training cycle, which can be months away.
RAG-enabled systems (like Perplexity and browsing-enabled ChatGPT): These tools fetch live web content before answering. If someone asks Perplexity for “boutique hotels in Edinburgh,” Perplexity searches the web for relevant pages, reads them, and incorporates what it finds into its response. Your website, your Google Business Profile, and recent review content can all influence what it says about you right now.
What does RAG mean for hotels?
It means optimising for both types of system at once: long-term editorial presence for training-data systems, and a crawlable, current website for RAG-based ones.
For training-data-based systems: Build long-term editorial presence through TripAdvisor reviews, travel publication coverage, and consistent entity signals across platforms. This takes months to accumulate but compounds over time.
For RAG-based systems: Make your website easy to crawl and read. Add an llms.txt file, implement Hotel schema markup, ensure your property description and amenities are in plain HTML text (not loaded by JavaScript), and keep your content current.
A hotel with a well-structured website with up-to-date content can appear in Perplexity recommendations relatively quickly, even without deep historical editorial coverage. This is the fastest path to AI visibility for a new or recently rebranded property.
Why does website readability matter more than it used to?
With RAG-enabled AI systems increasingly powering travel queries, a hotel website that is difficult for AI to read is a hotel that is invisible to a growing segment of recommendation tools. JavaScript-heavy pages that load content dynamically, or sites with no structured data, are at a growing disadvantage.