AI Knows the Word. Does It Know the Meaning?
By Layton Han, Chief Executive Officer · October 22, 2026
AI has become remarkably good at language. As it moves from answering questions to booking travel, the gap between knowing a word and understanding its meaning can turn a wrong interpretation into a wrong transaction.
The Tourist with a Phrasebook
Picture a traveler who has memorized a phrasebook in a language they've never spoken. They can ask, "How much is this?" perfectly. But when the shopkeeper responds with a detailed explanation, they nod politely, understanding almost none of it.
They know the words. They don't understand the meaning.
That's a useful way to think about one of AI's biggest challenges today. AI has become remarkably good at language, recognizing patterns and generating answers that sound knowledgeable. But knowing how words are used isn't always the same as understanding what they mean in a particular context.
Until now, the consequence of misunderstanding has often been a wrong answer. As AI moves from answering questions to making decisions and booking travel, a wrong interpretation can become a wrong transaction.
The distinction between knowing words and understanding meaning is no longer just an interesting question about AI. It's becoming a critical business issue.
Words, Details, and Meaning
Consider the word "suite."
| Layer | Example | What it tells AI |
|---|---|---|
| The Word | Suite | What the product is called |
| The Details | 600 sq ft, king bed, sofa bed | What the product includes |
| The Meaning | Separate sleeping spaces for a family; meeting space for a business traveler; a special experience for a couple | Why the product matters to a particular traveler |
The word is the same. The details vary. But the meaning depends on who is asking and what they need.
The word is only the label. Meaning is what the word amounts to once you add context: what the traveler needs, and what this particular supplier actually offers.
A hotel knows what its suite offers. A traveler knows what they need. AI's job is to connect the two.
When AI lacks specific information about a property, it may fall back on general assumptions about what a "suite" typically includes. AI learns from patterns in text, not from experience, so it tends to settle on the most common meaning of a word. But what's typical isn't necessarily what a particular hotel offers, or what a traveler needs.
And when AI moves from recommending to booking, a plausible assumption can become a costly mistake.
Why Travel Has More at Stake
Unlike many consumer products, travel is purchased before it's experienced. A product that doesn't meet expectations can often be returned. A hotel stay, flight, or vacation cannot be so easily undone.
Travel products are also highly differentiated. A "suite" means something different across properties. A "premium" airline fare can include entirely different benefits depending on the carrier. Even "free cancellation" comes with conditions that vary by supplier, rate, and date.
When AI cannot accurately connect traveler intent with supplier-specific meaning, three risks emerge:
1. Differentiation Gets Lost
When AI cannot recognize what makes a property, room, or fare distinctive, it falls back on attributes that are easier to compare: price, location, ratings, and availability.
Premium features and unique experiences risk being reduced to generic inventory. Products that are meaningfully different begin to look interchangeable.
Worse, a property can be left out entirely. If AI can't tell whether a hotel matches what a traveler asked for, it may never surface it.
2. AI Can Make Promises You Didn't Make
Without accurate, supplier-specific context, AI may fill information gaps using third-party descriptions, reviews, or generalized assumptions.
A room may be described as having a separate bedroom when it doesn't. A loyalty benefit may be presented as guaranteed when it's subject to availability. A cancellation policy may be interpreted more generously than the actual terms allow.
The traveler doesn't necessarily know where the incorrect information originated. They simply expect the supplier to deliver.
3. The Supplier Bears the Cost
When expectations don't match reality, the hotel or airline often inherits the problem.
Disappointed travelers, additional service requests, refunds, and weakened loyalty relationships can follow.
And as AI takes on more travel decisions, these mismatches could become more frequent and costly.
The larger risk is commoditization (interchangeable products competing mainly on price), not because travel products lack differentiation, but because AI cannot reliably understand and communicate what makes them different.
From Product Descriptions to Meaning
One practical example is the difference between a traditional product description and a contextual FAQ.
Traditional Room Description:
"Executive Suite, 600 sq ft, one king bed, sleeper sofa."
Contextual FAQ:
Q: Can adults and young children sleep in separate rooms?
A: Yes. This suite includes a separate bedroom with a closing door and a living area with a sofa bed, allowing adults and children to sleep in separate spaces.
Both contain useful information. But the second connects specific product attributes to a traveler's actual needs.
It helps AI understand not just what the room includes, but why those features matter.
This principle extends beyond FAQs to room configurations, cancellation policies, loyalty benefits, amenities, and service offerings.
The opportunity isn't simply to publish more content. It's to make the attributes, conditions, and distinctions of each product accessible and understandable to AI.
What makes a product understandable to AI isn't only what it includes, but what needs it fulfills, who it's right for, and where its limits begin.
The Next Challenge: Protocols and Context
Much of the industry's attention is focused on building the infrastructure that allows AI agents to interact with businesses and execute transactions.
Emerging frameworks such as Poppy, a draft standard developed by Sierra and Meta, focus on how an agent identifies itself and what the customer permits it to do. Other protocols address how agents can interact with commerce systems and complete transactions.
These are important developments. They answer a different question from the one this issue is about: whether the agent understands what it's booking. We call that context: what the traveler needs and what the supplier actually offers, which together give a word its meaning.
The Agentic Commerce Stack
| Layer | Question it answers |
|---|---|
| Transaction & Delegation | Can the agent securely make the booking? |
| Context | Does the agent understand what it's booking and why it's right for the traveler? |
Protocols handle the handshake. Context helps ensure the transaction delivers on the traveler's actual intent.
Looking Ahead
AI is rapidly becoming capable of searching, comparing, and booking travel on behalf of consumers. But completing a transaction isn't the same as making the right decision.
That requires more than words and product descriptions. It requires context that makes each supplier's products, distinctions, and promises understandable to AI.
In our view, the winners in an AI-first travel industry won't necessarily be those with the most content. They'll be those whose products and unique value AI can understand, differentiate, and act on.
