---
title: "What a 1984 Airline Screen Rule Would Ask of ChatGPT's Unpaid Answers"
canonical: "https://current.tinyfish.ai/issue/55/market-pulse/article/65029/what-a-1984-airline-screen-rule-would-ask"
publication: "TinyFish"
publisher: "TinyFish"
language: "en-US"
guide: "https://current.tinyfish.ai/llms.txt"
format: "Market Moves"
author: "Sevda Polat"
author_url: "https://current.tinyfish.ai/team/200"
issue: 55
section: "market-pulse"
section_name: "Market Pulse"
published: 2026-11-13
description: "A 1984 rule banning owner-biased airline screens suggests what ChatGPT's unpaid business answers might need: published, checkable weights."
word_count: 826
source_links: 12
---

# What a 1984 Airline Screen Rule Would Ask of ChatGPT's Unpaid Answers

*By [Sevda Polat](<https://current.tinyfish.ai/team/200/markdown>)*

*November 13, 2026*

[Read on the website](<https://current.tinyfish.ai/issue/55/market-pulse/article/65029/what-a-1984-airline-screen-rule-would-ask>)

Forty-two years ago tomorrow, a federal rule took effect that implied a simple test for whether a list of flights was fair. Change the airline's name on a flight and leave everything else alone, and the flight should stay where it was on the screen.

The screen belonged to a travel agent. In the early 1980s, agents booked through reservation systems owned by the largest airlines. Those systems showed competitors' flights alongside the owner's, but the owner set the order. Agents booked from the top. One system, System One, looked at five days of bookings and found that 93 percent were made from the first screen, a figure the GAO [cited in 1995](https://www.gao.gov/assets/rced-95-99.pdf).

The owners had a real argument: they had paid to build infrastructure their rivals now depended on. American Airlines' Robert Crandall complained in [1983](https://www.washingtonpost.com/archive/business/1983/06/26/flight-booking-controversey-lands-on-hill/daef2f66-7f41-4267-9a26-24c702720295/) about criticism from carriers "that did not choose to make similar investments."

The Civil Aeronautics Board, which then regulated airlines, answered with a rule effective November 14, 1984. It barred ordering the main display by factors "[directly or indirectly relating to carrier identity](https://archives.federalregister.gov/issue_slice/1984/8/15/32533-32564.pdf)." Departure time and trip length were allowed if applied equally to every airline, and owners had to disclose their criteria and how much weight each carried. The rule did not require a label. The agent already knew whose flight it was, and a label would have left the owner's flight at the top of the screen.

On Tuesday, OpenAI [opened a beta](https://openai.com/index/introducing-businesses-chatgpt/) that lets businesses claim profiles inside ChatGPT. Eligible US ecommerce and online-service businesses are the first to get their own agents there. Conversations with those agents are labeled and kept separate from ChatGPT's own answers, and businesses can buy labeled ads that send users to them. OpenAI says a claimed profile [does not guarantee higher placement](https://openai.com/business/businesses-in-chatgpt/) and that ads [do not influence answers](https://openai.com/index/our-approach-to-advertising-and-expanding-access/).

Labels make sense for a company whose product is trust in its answers, and they cover both the paid route and the conversation with a business's own agent. The ordinary answer carries no label: the hotels ChatGPT names when nobody has paid to appear. That answer may already have decided which hotel a customer approaches before she ever talks to a hotel's agent. It is the step the 1984 rule was written for.

Because OpenAI does not sell the rooms it recommends, the conflict runs between businesses that pay and businesses that don't, rather than between a platform and its competitors. A conversational answer may also name a single business, with no screen to scroll and no visible runner-up. A customer cannot notice an omission she never sees, so a published, testable standard may matter more here than it did to travel agents, who could at least page down.

The airline history also suggests where advantage goes once ranking is policed. By 1992, the GAO reported that screen bias had been "[virtually eliminated](https://www.gao.gov/assets/rced-92-130.pdf)." Owner flights still took fewer keystrokes to book, though, and their confirmations came back faster, so some of the edge seems to have moved from screen position to how hard a flight was to buy. The GAO noted that vendors and rival airlines disputed how much those differences mattered.

In ChatGPT, the comparable questions are about effort. How many exchanges does it take to reach a merchant with no advertised agent? Does that merchant's booking go through as smoothly as a sponsor's? Some gaps will have honest causes, like cleaner inventory data or a booking connection that works.

OpenAI's promises point the same way as the 1984 rule, and the company [publishes some shopping factors](https://help.openai.com/en/articles/11128490-shopping-with-chatgpt-search), among them availability, price, quality, and whether a merchant makes or mainly sells the item. What the 1984 rule added was weights, disclosed on request so a competing airline could test them against the order it actually saw. For a business that relies on ChatGPT for introductions, the open question is who outside OpenAI could see it if payment ever began to change the path.

## Things to follow up on...

- **When neutrality had gaps:** DOT's [1992 replacement rule](https://archives.federalregister.gov/issue_slice/1992/9/22/43774-43837.pdf) extended the neutrality requirements to every integrated display and added programmers' algorithm specifications to what had to be disclosed, a sign of how far the original 1984 rule had left room to work around it.

- **Why the rules expired:** In its [2004 sunset decision](https://www.govinfo.gov/content/pkg/FR-2004-01-07/html/03-32338.htm), DOT let most display rules lapse because US airlines had divested their reservation systems and were selling more through their own websites, even though it still found market power in the systems.

- **Europe's single-answer clause:** The [Digital Markets Act](https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?from=en&qid=1673870123858&uri=CELEX%3A32022R1925) bars designated gatekeepers from self-preferencing in ranking and explicitly counts virtual assistants and even a single communicated result, though it applies only to firms designated as gatekeepers.

- **Agents as new gateways:** The OECD's November 12 paper on [competition and agentic AI](https://www.oecd.org/en/publications/competition-and-agentic-ai_d1a3ee15-en/full-report/component-3.html) weighs better comparison shopping against the risk that seller payments or vertical integration bias which suppliers an agent puts in front of a customer.

## More

- [Market Pulse — Issue 55](<https://current.tinyfish.ai/issue/55/market-pulse/markdown>)
- [Issue 55](<https://current.tinyfish.ai/issue/55/markdown>)
- [Publication guide](<https://current.tinyfish.ai/llms.txt>)