The Airlines

Four airlines that move most of America do not get the same answer from the engines that now field "which airline is good." AI frames Delta as the premium, reliable one. It frames Southwest as the friendly low-cost carrier — still trailed by a holiday operational collapse. United is introduced by an old viral incident; American by scale and the absence of a story.

The pattern is the franchise thesis in its purest form: AI does not surface your average day. It surfaces your worst one.

Reputation modeled across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews — 40+ reputation-intent prompts per carrier across identity, trust, track record, controversy, comparison, and decision intent. Multiple passes; recurring findings only. Cross-checked against current reporting, critical and favorable. Directional estimates — not a precision instrument.

Valence of the dominant framing each engine surfaces first.

The finding. Only Delta leads clean. The other three are paired, on first surface, with a frozen negative.

"The first sentence is the reputation. Almost no one reads past it."

Delta answers open with "major US airline known for reliability and premium service." Southwest answers open with "low-cost carrier known for customer-friendly policies" — then, fast, the holiday-meltdown reference. United answers reach the old viral incident inside the first answer, often unprompted. American answers open with "the largest airline in the world" — scale as the identity, because little else is.

Airline retrieval bases are dominated by event-driven press: crisis coverage, viral incidents, and delay-and-cancellation reporting. Delta's base carries more favorable quality coverage to offset the events. Southwest's and United's are anchored by single, heavily-reported crisis events the engines retrieve far more readily than any recovery. American's base is thin on positive distinctiveness and heavy on operational and labor friction.

Across all four, the same omission: the ordinary, successful operating day. Safety records, on-time recoveries, route and service improvements, and the sheer scale of incident-free travel are systematically under-surfaced. AI retrieves the meltdown and skips the millions of uneventful flights that followed it.

Moderately consistent — all engines converge on the same frozen events, with valence varying at the margin.

The agreement is the problem: when every engine retrieves the same crisis, the negative is not contested — and an uncontested negative is the hardest kind to move.

Delta leads — the only carrier with a positive frame strong enough to outweigh its event archive. Southwest edges United; both are defined by a single frozen crisis, and Southwest retains more residual goodwill. American trails — not because its worst day was worse, but because it has the least positive narrative to set against any day. In the airline category, AI-mediated reputation is decided by whether a carrier has built enough good story to outweigh its worst headline.

Southwest and United hold the widest gaps — both are, by most current measures, materially better operations than the frozen incident the engines lead with. Delta's gap runs the other way: the AI narrative is slightly better than the current experience, a gap that closes itself unless managed. American's gap is the absence of a story — there is little intended narrative for the delivered one to fall short of.

The category lesson: a single crisis, well-documented, becomes the permanent answer unless a deliberate positive record is built to outweigh it. For Southwest and United, the work is to make the recovery and the current operation as citable as the crisis already is — not by arguing with the incident, but by burying its share of the retrieval base under verifiable, current evidence. For American, the work is more fundamental: there is no positive narrative to surface, so one must be built. For Delta, the work is to close a gap running the other way, before the experience falls visibly short of the reputation.