United Airlines has changed its contract of carriage so AI agents can no longer make, change, or service a ticket, the first major airline to say no in its contract. Host Paris Childress also covers Reddit's new AI review pages, a Sistrix drop in Google visibility for big UK news sites, Google AI Overview tweaks, a slipped MRC measurement timeline, an Invoca tracking gap, and Brave Search's growth.
United Airlines has changed its contract of carriage so AI agents can no longer make, change, or service a ticket, the first major airline to say no in its contract. Delta and American told Fast Company they are not doing the same for now. Paris Childress says agentic commerce depends on sellers letting agents complete purchases, and he is not sure how long United can hold out once verification standards are in place.
The same briefing covers Reddit's new AI review pages, a Sistrix drop in Google visibility for big UK news sites, Google AI Overview tweaks, a slipped Media Rating Council measurement timeline, an Invoca tracking gap, and Brave Search's growth.
United changed the contract on Wednesday. AI agents can no longer make, change, or service a ticket, or buy other United services or products. United defines an AI agent broadly, including assistants inside search, social, and messaging apps and AI browsers. Its press release, as quoted by One Mile At A Time, gives two reasons: sharing logins with AI assistants can expose sensitive information, and AI agents could break FAA and DOT rules. Delta supports AI assistants that can properly confirm who the passenger is. American says some AI tools can already buy its tickets, it is evaluating, and it keeps the right to block automated traffic. One X post relayed on Techmeme says agents can still show United flights and prices but cannot book. Paris flags that as coming from a single unverified source, and the contract text does not address it.
United is the first major airline to say no in its contract.
Paris is perplexed that other airlines are still allowing AI agents to book tickets. He thinks it is a matter of time before verification standards and agreed protocols are in place, because people want their agents to book travel, and he is not sure how long United can hold out.
On November 10, Reddit started testing Reddit Reviews, AI-organized review pages for about 10,000 frequently discussed products in electronics, health and beauty, home and garden, and fashion and apparel, built from user conversations. Each page summarizes what people think, counts positive, mixed, and negative conversations, and links back to the original threads. Reddit says businesses cannot pay to be featured. The test reaches US users on iOS, Android, and the web over the next few weeks, according to PPC Land. Images, prices, and purchase links appear where brand and retailer catalogs supply them. Reddit says half of US shoppers come to Reddit to check AI recommendations, a figure from its own Path to Purchase 2026 survey.
Paris notes that Reddit is already one of the sites AI answers cite the most, so tidy review pages could become an even easier source for AI tools to cite. Brands should watch how their products are being reviewed there. He also warns that AI summaries can flatten the nuance in those threads, and brands may face summary errors they cannot correct and cannot pay to get around.
Press Gazette, using Sistrix data, reports that more than two-thirds of the biggest UK news sites lost Google search visibility over the past year. BBC News went from 41.8 to 20.35. The New York Times went from 58 to 38. The Guardian went from 236 to 216. Substack rose from 13 to nearly 28. Money Saving Expert improved from 81.82 to 94, and The Times rose from 12.5 to 17.78. The Daily Mail's numbers are not comparable because it moved domains in April 2026. Sistrix's visibility score counts only regular organic results, not AI Overviews, Top Stories, or the Google Discover feed. Barry Adams of Polemic Digital says the shifts are likely due to Google showing a wider range of sites. Top Stories carousels now appear inside some AI Overviews, so if the UK's Competition and Markets Authority forces an opt-out from AI Overviews, publishers who use it could also lose Top Stories visibility.
Paris asks whether Substack's gain means Google is starting to favor individual writers over institutions, and what that would mean for brand content.
Two changes showed up on November 11. Brodie Clark spotted Google testing the count of cited sites at the bottom of the answer instead of above the citations. Hovering over the source count still shows the full list. Glenn Gabe noticed definition answers now open fully with no Show More click. Barry Schwartz says the fully expanded view applies to all definition searches, and he points to Google's Robbie Stein saying in August that AI Overviews expand only where Google systems think it is most useful. Paris's take is that these look like minor tweaks, but they all push in the same direction: more AI Overview real estate and fewer reasons to click.
Paris closes with a lightning round: Google's John Mueller saying lots of crawled and discovered, currently not indexed pages usually means Google is not convinced about the site's quality, and shrinking the site does not fix that; moving to a CDN can cause a surge in Google crawling until the cache fills, and blanket bot blocking can also block Googlebot; Supermetrics has added ChatGPT ads as the seventh platform in its AI campaign management tool; and early ChatGPT Business Hub data from one business showed 690 mentions and 24 website clicks over 28 days, with some signups put on a wait list on launch day.
Paris: Hi, everybody. Welcome back to The GEO Show, brought to you by GEOforge, full self-driving for AI visibility. I'm your host, Paris Childress, and today is November 12th. We are headlining today's episode with news from United Airlines. United Airlines bars AI agents from buying or servicing its tickets.
Paris: United changed its contract of carriage on Wednesday. AI agents can no longer make, change, or service a ticket. United says sharing account logins with third-party AI assistants can expose sensitive data. Delta and American told Fast Company they aren't doing the same for now. Pretty interesting here.
Paris: The new clause covers making, changing, or servicing tickets and buying other United services or products. United defines an AI agent broadly. The definition includes assistants inside search, social, and messaging apps and AI browsers. United's press release, as quoted by One Mile At A Time, gives two reasons. Number one, sharing logins with AI assistants can expose sensitive information. And two, AI agents could break FAA and DOT rules. For example, the check-in confirmations passengers must give. Delta supports AI assistants that can properly confirm who the passenger is. American says some AI tools can already buy its tickets and it's evaluating. It keeps the right to block automated traffic.
Paris: One X post relayed on Techmeme says agents can still show United flights and prices but can't book. The contract text doesn't address this. It comes from a single source and is unverified. So agentic commerce depends on sellers letting agents complete purchases. United is the first major airline to say no in its contract. Airlines are now split between blocking agents and allowing them if they can prove who the passenger is.
Paris: I'm a bit perplexed by this announcement from United, given that other airlines are still allowing AI agents to book tickets, and I do think it's a matter of time before this really gets cleared. I think once we have more standards and agreed protocols in place for people to be able to verify and validate their agents, this trend is only gonna continue because people really want their agents to go out and do stuff like booking flights, booking travel. So I'm not sure how long United Airlines can really hold out on this.
Paris: Next up, Reddit launches AI review pages built from user threads with no paid placement. On November 10th, Reddit started testing Reddit Reviews. These are AI-organized review pages for about 10,000 frequently discussed products built from user conversations. Each page summarizes what people think and links back to the original threads. Reddit says businesses can't pay to be featured. The test covers about 10,000 products in electronics, health and beauty, home and garden, and fashion and apparel. It reaches US users on iOS, Android, and the web over the next few weeks, according to PPC Land. Each page has an AI summary and counts of positive, mixed, and negative conversations. If a product has no review page, Reddit's AI search answers instead. Reddit says businesses can't pay to be featured. Images, prices, and purchase links appear where brand and retailer product catalogs supply them. Reddit says half of US shoppers come to Reddit to check AI recommendations. That figure is from Reddit's own Path to Purchase 2026 survey.
Paris: Really interesting move by Reddit, and as you all probably know, Reddit is one of the sites that AI answers cite the most. So tidy, summarized Reddit review pages could become an even easier source for AI tools to cite when people ask about products. This is a really interesting move by Reddit moving into reviews. I think it makes a lot of sense given that they can harness AI here to summarize all of the sentiment around a particular brand or product. And now this means that if your product is getting reviewed by Reddit, it's important to be aware of that and to manage those reviews and manage that feedback as much as possible. Also, AI summaries with opinions can flatten the nuance that's hiding in those Reddit threads. So brands here may face summary errors that they can't correct and they can't pay to get around.
Paris: The next story is from Sistrix. Over two-thirds of big UK news sites lost Google visibility over the past year. Press Gazette, using Sistrix data, reports that more than two-thirds of the biggest UK news sites lost Google search visibility over the past year. BBC News roughly halved and Substack doubled. Sistrix's visibility score counts only regular organic results, not AI Overviews, Top Stories, or the Google Discover feed. So these are the numbers in terms of visibility, and this is a visibility index, so this is a score that represents a broad index.
Paris: BBC News went from 41.8 to 20.35. The New York Times from 58 down to 38. The Guardian from 236 to 216, so The Guardian didn't lose too much, interestingly. Substack is a big gain, a big winner, 13 up to nearly 28. Money Saving Expert also improved its visibility from 81.82 to 94, and The Times also increased a little bit from 12.5 to 17.78. The Daily Mail's numbers aren't comparable because it moved domains in April 2026. Barry Adams of Polemic Digital says the shifts are likely due to Google showing a wider range of sites in its results. Top Stories carousels now appear inside some of AI Overviews. So if the UK's Competition and Markets Authority, the CMA, forces an opt-out from AI Overviews, publishers who use it could also lose Top Stories visibility. So even before counting AI Overviews, big publishers' share of regular search results is shrinking. Opting out of AI Overviews may now also cost News carousel visibility.
Paris: These results are quite interesting, especially Substack. If Substack is the big winner here, does that mean that Google is starting to favor more individual writers over institutions? And what does that mean for brand content?
Paris: Next story, more tweaks from Google AI Overviews. The source count moves to the bottom. Definitions are now shown in full. Two AI Overview changes showed up on November 11th. Brodie Clark spotted Google testing the count of cited sites at the bottom of the answer instead of above the citations. Glenn Gabe noticed definition answers now open fully with no Show More click. So that means more real estate and more pixels being given to AI Overviews. In the test, hovering over the source count shows the full list of citations. Barry Schwartz says the fully expanded view applies to all definition searches. He points to Google's Robbie Stein saying in August that AI Overviews expand only where Google systems think it is most useful. So where sources appear decides whether users see them. Moving them lower and expanding answers both give users fewer reasons to click. So these changes individually seem very small, and they seem like minor tweaks, but they're all pushing in the same direction, which is expanding the real estate of AI Overviews and taking away more clicks.
Paris: Next story. Media Rating Council sets a timeline for its first AI measurement standards. The Media Rating Council, MRC, is the US industry body that accredits ad measurement. It says draft standards for AI in media measurement will arrive in two phases. Phase one goes out for public comment in Q1 2027, and phase two follows by mid-2027. This timeline came in an MRC LinkedIn post on November 10th, according to PPC Land. In July 2026, the MRC issued 13 pages of guidance that applied its existing standards to AI. That guidance said the work would finish throughout 2026 and early 2027, so the timeline has slipped a bit. Areas the standards will cover include, first, measuring ads inside AI chat answers. Second, traffic from AI agents. Third, zero-click measurement. And last, labeling AI-generated content. So AI search, AI Overviews, and ChatGPT are not named specifically. So advertisers who are now buying ChatGPT and AI Mode ads will now have no accredited measurement. The earliest standard is a draft in 2027. But to put this in perspective, we really need to ask who defines AI visibility metrics first. Is it the MRC or the platforms like OpenAI's new Business Hub or the vendors? I think it's gonna be more defined by the platforms and the vendors, especially given OpenAI's new rollout of the Business Hub.
Paris: Next story. Invoca survey says 74% of marketers using AI agents can't track agent-handled leads to a sale. Invoca surveyed 451 US consumer brand marketing leaders. Of the 323 using AI agents, 74% can't fully track a lead that an agent handled from first click to a sale. Only 19% can reliably feed conversation results back to their ad platforms. The survey was run by Gather from the 14th of September to October 5th among marketing leaders at US consumer-facing companies with 100-plus employees. The margin of error is about 4.6 points. Among those using AI agents, here's the data: 74% can't fully track agent-handled leads. 26% can link agent conversations to the campaign or the ad, and 19% can send outcomes back to ad platforms. 86% say they can spot leads that started in AI answer tools such as ChatGPT or Google's AI Mode. So AI answers and on-site AI agents now each handle part of the path to purchase, and the tracking gaps really add up now. And of course, you can't optimize what you can't measure, so there continues to be a big blind spot here in attributing lead generation from AI agents, and this is still an outstanding problem, I think.
Paris: The next story is about Brave Search. Brave Search, which is a search engine now that's getting more prominent, says that their queries are up 56% in a year, and about a third of those get an AI summary. Brave says its search engine handled 2.7 billion queries in October 2026, up 56% from 1.73 billion a year earlier. Users grew from 101 million to over 130 million. About a third of its 50 million-plus daily queries trigger an AI summary. Brave had about 85.5 million queries on its busiest day on August 31st, and its ad revenue has grown 2.85 X year on year. Brave sells access to its search index to several AI companies, and that arrangement is part of The New York Times' lawsuit against Brave, which we covered on November 3rd. Several AI answer tools pull results from Brave's index, so its growth matters more for AI visibility than its consumer share might suggest. So is Brave really the most important search engine index that nobody optimizes for? Perhaps it is.
Paris: Okay, let's wrap up with some lightning round stories. First, Google's John Mueller on indexing says lots of crawled and discovered currently not indexed pages usually means Google isn't convinced about the site's quality. So just making the site smaller doesn't fix that. Next, Google on moving to a CDN. The move can cause a surge in Google's crawling and the origin server takes the load until the CDN's copies fill up. Blanket bot blocking can also block Googlebot. Google advises allowing verified crawlers and filling the cache in advance. Next, Supermetrics adds ChatGPT ads as the seventh platform in its AI campaign management tool. And last, ChatGPT Business Hub early data. One business's dashboard showed 690 mentions and 24 website clicks over 28 days. Some signups were put on a wait list on launch day. All right, that's a wrap for today's episode. Thank you all for tuning in and see you on the next one.