The Great AI Agent Lockout: What United’s Ban Means for Agentic Commerce

The promise of agentic AI is that you tell an assistant to “book me the cheapest nonstop to Chicago on Friday” and it handles the rest. The trouble is that the airline, retailer or bank on the other side never agreed to deal with your assistant. On November 11, United Airlines made that explicit. It revised its Contract of Carriage to bar the use of AI agents to make, modify or service a ticket, and reserved the right to cancel such bookings without notice.

Our earlier briefing covers the specific rule changes. This analysis looks at the broader pattern across news coverage and search data, and at what it means for anyone building, buying or defending against AI agents.

The story behind the headlines

The United decision was covered widely: The New York Times, The Wall Street Journal (which reported it first), Reuters, CNBC, CBS News, The Washington Post, USA Today, TechCrunch, Quartz and PhocusWire. Taken together with other recent stories, the coverage shows a market in transition.

• United is not alone, but it is out front. Forbes reported that United stands alone among major US carriers in writing a ban into its terms, with rivals so far leaving the risk with customers. Business Insider published a running list of companies banning AI agents. The Verge reported eBay is restricting third-party agents. eBay’s user agreement already barred autonomous “buy for me” agents from February 2026, according to earlier trade coverage. Amazon has reserved the right to cut off agent access at its discretion.
• The stated reasons are security and compliance. United cited the risk of customers sharing credentials with third-party assistants, unintended purchases and changes, misuse of MileagePlus miles and regulatory steps that require a human. Check-in, for example, requires a human to confirm that bags contain no prohibited items. Coverage also quoted CEO Scott Kirby saying he does not want another middleman.
• Agents are hard to detect. PPC Land reported on a Resemble AI test in which bot-detection tools caught only 3.1% to 14.8% of disguised AI agents. Cloudflare’s CEO said most websites are now struggling with AI agent traffic. A contract term is easier to write than to enforce.
• Agents still make mistakes. Fortune covered a study in which AI shopping agents buying TVs and laptops got it wrong nearly half the time. Platforms are not only protecting margins. They are also protecting themselves from errors and the chargebacks and complaints that follow.
• The commercial stakes are large. Reuters reported that AI shopping agents could threaten Amazon’s and Walmart’s advertising businesses, because an agent does not see sponsored listings the way a human scrolling a results page does. Adobe forecast 130% growth in AI-driven shopping traffic. Google launched a “Universal Cart”. Walmart is building infrastructure for agentic commerce. Boston Consulting Group asked whether consumer-goods companies are ready.

The same week, TechCrunch reported that Meta’s Muse assistant can now place calls to US businesses for paying users. Agent capabilities are expanding even as more platforms restrict them.

What Google Trends shows

The data shows:

• Travel is where people feel this. “AI travel booking” more than tripled on November 13, from an index of 5–7 to 21, as consumer coverage of the United ban spread. “Flights” and “hotels” were among the top related queries for “agentic commerce”. Travel is one of the use cases people most want agents to handle.
• Consumer assistants drive curiosity. The fastest-rising “AI agent” queries were about Meta’s Muse. Interest is coming from consumer assistants that will increasingly run into platform restrictions, not from enterprise tools.
• Industry is searching for standards. “Agentic commerce protocol” was the top related query for “agentic commerce”, followed by searches naming Stripe, Google and Visa. Practitioners are looking for an approved, authenticated route for agents, which is exactly what United’s ban lacks.
• Interest is global, led by Asia. China, Hong Kong, Singapore, Vietnam, South Korea and Taiwan led interest in “AI agent”, with the UK and UAE also high.

Note on the data: every series we pulled showed a shared spike around September 20–21, and some rising queries look unrelated (for example, recipe and tourism searches). We left out results that are clearly noise and compare short windows directly.

Why platforms are pushing back

The economics explain most of it. Platforms that own the customer relationship earn money in three ways that agents threaten:

1. Merchandising and ads. A human sees seat upgrades, bundles, sponsored listings and loyalty offers. An agent asked for “cheapest nonstop” skips them.
2. Data and loyalty. Every direct session feeds personalization and loyalty programs. An intermediary agent keeps that data for itself.
3. Risk. Errors, fraud, credential sharing and regulated steps all land on the platform’s support desk and compliance team.

Airlines have dealt with intermediaries before. Global distribution systems, online travel agencies and metasearch each forced a renegotiation of who owns the customer. AI agents are the next round. United’s bet is that it is better to set terms early than to inherit a new middleman by default.

The likely end state: accredited agents

Outright bans are unlikely to last as the only answer. Customers will want agents, competitors will accommodate them and detection is unreliable. The more probable outcome is a tiered model:

• Blocked: anonymous browser automation using shared passwords.
• Tolerated: agents that research and compare, then hand off to the customer to complete the purchase. Coverage suggests United still allows this.
• Accredited: agents that authenticate through an approved protocol, use scoped and revocable payment tokens, identify themselves and accept liability terms, possibly with a commercial agreement. eBay has signalled openness to partnerships on its own terms.

Checklist for technology teams

If you build or deploy agents

• Never ask users for raw account passwords. Use OAuth, delegated tokens or official APIs wherever they exist.
• Keep a machine-readable registry of each target site’s terms of service and contract rules for agents, and check it before acting.
• Require explicit human confirmation for payments, loyalty-point redemptions and any step a regulator expects a human to perform.
• Have your agent identify itself honestly in its user-agent string and request headers. Disguising agents to evade detection invites contract claims and possibly fraud claims.

If you run a platform agents want to use

• Decide on a policy before an incident forces one: block, tolerate, or accredit.
• Update your terms, but don’t rely on them alone. Pair them with bot management, rate limits and anomaly detection on high-risk actions.
• Offer a sanctioned route, such as an API or a protocol integration, so legitimate agents have a reason to identify themselves.
• Track agent-driven support cases and chargebacks separately. That data will inform your policy decisions.

What to watch next

• Whether other US and international carriers follow United or compete by welcoming agents.
• How United enforces the rule against agents arriving through travel agencies and aggregators.
• Progress on agent payment and identity standards from payment networks, processors and AI platforms.
• Regulators’ views. Consumer-protection agencies may ask whether blanket bans unfairly limit consumer choice, or whether agents harm consumers.

Frequently asked questions

Can I still use AI to search for United flights?
Coverage indicates agents can still display schedules and fares. The restriction targets making, changing or servicing tickets and buying other products.

Does it matter if I supervise the agent?
As reported, United’s definition applies whether or not a person supervises or approves the agent’s actions.

Is this only about airlines?
No. Marketplaces, retailers and other platforms are setting similar rules. Travel is simply where regulated steps and high ticket prices make the stakes clearest.

Sources and further reading

• United Airlines statement and revised Contract of Carriage, November 11, 2026; coverage by The Wall Street Journal, The New York Times, Reuters, CNBC, CBS News, TechCrunch, Forbes and PhocusWire.
• Business Insider: “Not so fast, bot: Companies are banning AI agents. Here’s the list.” November 12, 2026; The Verge on eBay restricting third-party AI agents.
• eMarketer: eBay’s agent ban signals the next battle in agentic commerce; TechCrunch: The next hurdle for AI agents—getting websites to let them in.
• Reuters on AI shopping agents and retail ads; Fortune on agent shopping errors; PPC Land on Resemble AI’s agent-detection test.
• Google Trends: “AI travel booking”, past 90 days.