During the pandemic, major airlines advertised the elimination of change fees. Yet basic-economy change fees quietly returned, creating a discrepancy between marketing promises and actual policy. Consumers grew confused by the conflicting messages. Generative AI systems, which ingest these policy pages to build knowledge bases, inherited that confusion. When an LLM answers a user’s query about airline fees, it relies on the text available. If that text is ambiguous, inconsistent, or buried in legal boilerplate, the AI outputs an error. This is not a temporary glitch; it is a structural failure in how airline policy content is currently written. If policy language isn’t structured for machine-readability, AI will keep misquoting fees, eroding consumer trust in automated travel assistants.
The DOT Rule That Demands Machine-Readable Policy Language
The U.S. Department of Transportation’s final rule, published as 89 FR 34620, fundamentally shifts how airline policy content must be structured. Effective July 1, 2024, the rule requires carriers and ticket agents to disclose passenger-specific or itinerary-specific fees for critical ancillary services at the first point a fare and schedule are provided. This applies directly to transporting checked and carry-on baggage, as well as changing or canceling reservations. The mandate is specific: fees cannot be hidden behind a hyperlink or buried in generic terms. They must be clear, conspicuous, and tied to the individual traveler’s itinerary before purchase.
This regulatory move addresses a documented failure. During the pandemic, inconsistent basic-economy change fees created confusion for consumers who saw advertised waivers that did not apply to their specific ticket type. That ambiguity is precisely what confuses generative AI answers today. When an AI system extracts data from a policy page that relies on broad ranges or buried links, it inherits the ambiguity, leading to inaccurate citations in search results. The DOT rule effectively mandates structured, machine-readable content by demanding that the specific fee for the specific fare type be visible upfront.
The contrast between the old and new disclosure models is stark. The previous system allowed vague ranges and hyperlinked details, forcing consumers—and algorithms—to dig for the actual number. The new model requires immediate, specific, and unlinked disclosure. This shift turns policy pages from static legal documents into structured data sources that can be accurately parsed by AI search optimization engines. For brands, this means your policy text is no longer just for human compliance; it is the primary data layer that determines accuracy in the AI era.
| Disclosure Element | Old Model (Pre-2024) | New Model (Post-July 2024) |
|---|---|---|
| Fee Specificity | Ranges or generic averages | Itinerary- and passenger-specific |
| Visibility | Buried in hyperlinks or fine print | Upfront at first fare display |
| Baggage Details | Often omitted until checkout | Weight/dimension limits shown pre-purchase |
| Change/Cancel | General policy summary | Specific restrictions and refund types per ticket |
Where AI Goes Wrong: The Carrier-vs-Ticket-Agent Disclosure Split
The DOT’s final rule draws a sharp line between what airlines must do and what ticket agents must do, a distinction that often gets lost in AI training data. Under the rule, carriers are required to provide accurate, current fee data to any entity obligated to disclose it. However, the rule explicitly defers the classification of metasearch sites to a later rulemaking. This creates a structural data gap. If a metasearch platform is not legally a ticket agent, it isn’t compelled to receive or display this data. When generative AI answers pull from a fragmented web where carriers, agents, and metasearch sites operate with different compliance postures, the model inherits this inconsistency. The result is a misattribution of fees: the AI cites a rule that doesn’t apply to the specific booking channel.
The Gap in Fare-Specific Data
Even when data is present, its structure often fails the test. Policy pages rarely separate fee rules by passenger status or fare type. A generic “change fee” policy might list a range of amounts, without specifying which fare class triggers which amount. For a human, this ambiguity is annoying; for a machine learning model, it is a signal to guess. If the policy text does not explicitly link “Basic Economy” to “Non-changeable,” the AI engine must infer the relationship from context. This inference is where errors originate. The model may cite the average fee or the highest possible fee, violating the DOT’s requirement for itinerary-specific disclosure.
Structuring for Machine Precision
To reduce this risk, airlines must restructure policy text to mirror the logic required by the regulation. Instead of a single, monolithic policy block, each fare type should have its own dedicated, labeled section.
| Fare Type | Change Policy | Cancellation | Refund Type |
|---|---|---|---|
| Basic Economy | Not allowed | Not allowed | None |
| Main Cabin | Fee applies | Fee applies | Voucher or Cash |
| Business | Free | Free | Cash |
This explicit, tabular format provides a clear anchor for extraction algorithms. It removes the need for the model to interpret prose. By aligning airline content strategy with this structural clarity, we help ensure that the data AI sees is not just compliant for humans, but also precise enough for machine consumption. The goal is to eliminate the guesswork, forcing the source text to be as specific as the regulation demands.
Making Baggage and Change Policy Language AI-Citable
The DOT final rule mandates that baggage weight and dimension allowances be disclosed before ticket purchase, not hidden in e-ticket confirmations. This applies equally to change and cancellation policies, which must specify restrictions, refund forms, and fare differentials upfront. For AI search optimization, this shift from buried legal text to front-end data changes how we structure airline policy content.
Designing for Machine Extraction
AI engines prioritize content that appears immediately adjacent to fare displays. Prose buried in terms-and-conditions pages carries lower authority in generative AI answers than structured summaries placed at the point of sale. To improve baggage policy clarity, write policy summaries as self-contained blocks. Each block should state the rule, the fee, and the eligibility without referencing other sections.
For example, a standard policy block should read:
Basic Economy – [Airline]: Checked bags: 23kg (51 lbs), 158cm (62 in) linear. Change fee: Not applicable. Cancellation: Not allowed. Refund: None.
This format allows algorithms to extract specific data points for the correct fare type without ambiguity.
From Ambiguity to Precision
Legacy policy text often relies on ranges or general statements, such as “fees vary by route and baggage size.” This forces AI systems to guess or provide a wide, unhelpful range. The AI-citable rewrite replaces ranges with specific, itinerary-aligned facts.
| Legacy Policy Block | AI-Citable Rewrite |
|---|---|
| “Change fees vary by route and class.” | “Change fee: $200 (Economy, domestic). Free (Business, all routes).” |
| “Baggage limits apply. See terms for dimensions.” | “Carry-on limit: 22kg, 56 x 45 x 25 cm. Checked bag limit: 23kg, 158 cm linear.” |
By eliminating legal boilerplate and isolating data, we create a clear path for accurate citation. This approach ensures that when a user asks about specific restrictions, the system retrieves the exact value tied to their itinerary, reducing the risk of hallucinated fees. Clear, structured airline content strategy is no longer just a compliance requirement; it is a technical necessity for visibility in AI-driven search results.
Structured Data and Schema: The Missing Layer
The DOT rule does not mandate schema markup, but its disclosure requirements align perfectly with structured data standards. Specifically, the requirement for itinerary-specific fees maps directly onto the priceSpecification property within a Service or Offer schema. This technical alignment turns regulatory text into machine-extractable facts.
Mapping Disclosure to Schema Types
Airlines should implement JSON-LD blocks that mirror the passenger-specific logic required by the rule. When a user searches a basic economy fare, the structured data should explicitly state the baggage fee for that specific itinerary, rather than linking to a general policy page. This approach ensures that generative AI answers pull the correct price without guessing. By embedding the data in the same context where the fare is displayed, we reduce the risk of hallucinations where an AI might cite a business-class fee for an economy ticket.
A Practical Example
Consider a simple schema block for a checked-bag fee. Instead of a generic policy URL, the markup specifies the exact cost and constraints for the current search result. This precision is what allows AI engines to verify facts before citing them.
{
"@context": "https://schema.org/",
"@type": "Service",
"name": "Basic Economy Flight: JFK to LAX",
"offers": {
"@type": "Offer",
"priceSpecification": {
"@type": "UnitPriceSpecification",
"price": 35,
"priceCurrency": "USD",
"name": "First Checked Bag Fee"
},
"termsOfService": "Basic economy tickets are non-changeable and non-refundable."
}
}
This structure removes ambiguity. It gives AI systems a clear, citable source of truth, ensuring that baggage policy clarity is maintained even in automated responses.
The DOT rule signals a shift: structured, machine-readable policy content is no longer just a compliance requirement—it’s the foundation for accurate generative AI answers. As airlines refine their airline content strategy, consider whether your team still treats policy pages as static text, or as the structured data problem they have become.
