What are Businesses in ChatGPT and Business Agents, and who can use them?

OpenAI's announcement, dated November 10, 2026, says Businesses in ChatGPT is in beta. Eligible businesses located where ChatGPT is available can claim a profile through Business Hub and see the number of mentions and website clicks. Business Agents are available to eligible ecommerce and online service businesses in the United States. The pages we read state no price, and OpenAI's sign-up page says "Claiming does not guarantee higher placement or sales." OpenAI's Terms say a listing may be displayed whether or not you claim it.

Can I still use AI to plan a trip with United?

United's help page says you "can still use AI to research and plan trips" as long as the tool "doesn't access united.com or the United app". It restricts unauthorized third-party AI agents or assistants that search, book, buy or manage on united.com or the app for you without United's permission. The page prints no date; its FAQ says agent bookings "made prior to November 11, 2026 will not be canceled". United's news release and revised Contract of Carriage, both dated November 11, 2026, also address AI agents; we did not interpret the contract.

Is Gemini 4 Carbon's price going up on December 1?

Google's pricing page (Standard tab, paid tier) lists $2.00 per million input tokens and $10.00 per million output tokens through November 30, 2026, then $4.00 and $20.00 starting December 1, 2026. That is double each list price, our arithmetic on two rows and not a forecast of any bill. Google's changelog, dated November 10, 2026, labels the model generally available, and the pricing page says it is not available on the Free tier. Google's Nov 10 blog adds the footnote "Applies to the Carbon model intended for public access only", which does not say what it excludes.

What does Google's Decisions API cost?

Google's blog, dated November 11, 2026, says Gemini Decision 1 is priced at $0.10 per 1M input tokens and $0 for output tokens, with a no-cost tier in Google AI Studio for prototyping. The 150-250 ms latency figure is Google's, for inputs up to 16K tokens. The pages we read print no stage label.

Who can open apps built with Grok, according to Claroc?

Claroc says that of 37,571 Grok-built apps still online when it checked, 35,578 (94.7 percent) opened for anyone, with no sign-in. The figures are Claroc's own. The company the post calls SpaceXSI, as Claroc relays it, says apps are private until their creator shares one and that publishing takes several steps. Claroc tested changing data only on its own two apps, and we opened none of the apps.

Signal / Noise Nov 11: ChatGPT, United, Gemini Price

Five small glass-and-metal objects on one square dark platform with a glowing red edge: a storefront with a striped awning, a glass airplane in front of a closed gate, a blank price tag beside a calendar page, a chip with a red speed gauge arc above it, and a row of three glass app tiles with open doors. Red glow accents each object.

The short version

OpenAI says businesses can claim a profile in ChatGPT, in beta. United's help page, news release and revised Contract of Carriage address third-party AI agents and assistants. Google lists Gemini 4 Carbon at $2.00 input / $10.00 output per million tokens through November 30, then $4.00 / $20.00. Also: Google's Decisions API at $0.10 per million input tokens, and Claroc's report on Grok-built apps.

In plain words

Think of this edition as five notices pinned to one board. One is a company listing inside ChatGPT, and one is an airline's help page about AI assistants. One is a price that changes on a set date, one is a new Google tool that answers a list of questions with probabilities, and one is a vendor's own report, from a company that sells a related check, on apps people built with an AI assistant. For each we say in one plain sentence what it is, what the source's own pages print, and what we could not check.

OpenAI says a company can now claim its own profile inside ChatGPT and see how often ChatGPT mentions it. United Airlines has a help page about which AI assistants may act for its customers. In our reading, both items ask the same question: who controls the link between a customer's AI assistant and the business?

This edition has three Signal items and two Watch items. Signal means you can act on it this week. Watch means worth knowing; one cheap check at most. The main sources are dated November 9 to 11, 2026; United's help page and several documentation pages print no date, as the Sources list marks. Every number belongs to whoever published it. The "What it means for you" and "Our take" lines are ours. For the previous edition, see the November 10 edition.

Signal: What are Businesses in ChatGPT, and can a company claim a profile?

OpenAI says businesses can now claim a profile inside ChatGPT, keep their details and product lists current, and see how often ChatGPT mentions them and how many website clicks they get. Its announcement is dated November 10, 2026, and says "We're introducing Businesses in ChatGPT in beta".

Who can claim one, and what do you see?

OpenAI's page says any eligible business "located where ChatGPT is available can claim their business in ChatGPT" through Business Hub, OpenAI's tool for managing a profile. Once verified, a company can "update key details and add self-service product feeds with pricing and availability". A product feed is a list of what you sell with prices and availability. OpenAI's separate sign-up page adds that eligibility "depends on factors including industry, location, and our verification and integrity checks".

For results, the announcement says Business Hub shows "the number of mentions and website clicks a business receives". On privacy, OpenAI says: "We don't share users' conversations or Business Agent chats with businesses."

What are Business Agents, and where are they available?

A Business Agent is a chat assistant that a business sets up inside ChatGPT to answer questions about what it sells. The announcement says Business Agents are "now available to eligible ecommerce and online service businesses in the United States." The page prints "beta" for Businesses in ChatGPT and prints no stage label for Business Agents. Businesses can also "sponsor their Business Agent" and use it "as the destination for ChatGPT Ads". ChatGPT Ads is OpenAI's own product name, and this is OpenAI's own announcement.

Does claiming help you show up, and what do the Terms say?

OpenAI's sign-up page answers the first part directly: "Claiming does not guarantee higher placement or sales." OpenAI's help article says claiming "doesn't guarantee increased visibility or give you direct control over ChatGPT's independent recommendations".

The help article says you "verify your connection to the business and be approved", then "complete identity verification with Persona" (an identity-check service) to publish changes, view insights or create an agent; the blue check "appears after your Persona verification is approved". The Businesses in ChatGPT Terms, last updated November 9, 2026, say, in short:

OpenAI may display a listing "regardless of whether you claim, manage, or cease managing it", and managing it "does not give you exclusive control over its contents" (section 4.1).
The license for basic business details you submit, which lets OpenAI show them, is "perpetual and irrevocable" (4.2).
A business that publishes an agent is "responsible for the Business Agent" (5.2).
OpenAI promises no "particular placement, ranking, visibility or level of distribution" (8).

The six OpenAI pages we read (listed under Sources) do not state a price for Business Hub, Business Agents or ChatGPT Ads, a date for markets outside the United States, or accuracy figures for Business Agent answers. Neither the Business Agents Policy nor the Businesses in ChatGPT policy lists the eligible categories. We did not read OpenAI's ads page or sign in to Business Hub.

What it means for you: ask your web person whether your business fits OpenAI's eligibility text, what it would cost (the pages we read state no price) and what the Terms ask of you. This is not legal advice.

Our take: Signal. The one check worth doing this week is whether your business is eligible and who would own the profile.

Signal: What do United's help page and contract say about third-party AI assistants?

United's help page says an AI assistant that searches, books or buys for a traveler on united.com or in the United app, without United's permission, is "unauthorized". It prints no date; we saw it on November 11, 2026. United also published a news release and a Contract of Carriage (the document that sets the rules for United tickets), both dated November 11, 2026. CBS News reported on the page that day; we used it only to find the page.

What exactly is restricted?

The page's central sentence says United is "taking steps to restrict access" on its own platforms and "indirectly via travel agencies and aggregators", "prohibiting use by unauthorized third-party AI agents or assistants". The FAQ counts a tool as unauthorized when it "searches, books, buys, manages or collects information" on united.com or the app for you and "United hasn't given permission". You "can still use AI to research and plan trips" provided the tool "doesn't access united.com or the United app", and "Assistive technology for customers with disabilities isn't affected." Asked whether United works with third-party AI agents, the page answers "Not yet".

What do the three documents say about timing?

The help page's only date is in its FAQ: agent bookings "made prior to November 11, 2026 will not be canceled". The news release says the work "includes updates to our booking, ticketing and distribution policies". The contract, "revised November 11, 2026", defines "AI Agent". Under Rule 5 (Cancellation of Reservations) it lists "reservations made, modified or serviced by AI Agents" among the "improper reservations" that United "will cancel without notice". Under Rule 6 (Tickets), its "Prohibited Practices" include "The use of AI Agents to make, modify or service a Ticket". None of the three prints an effective date for the restriction.

The help page does not state how United detects an agent, how many agent bookings exist, or how an assistant could become approved; it says it wants "standardized, secure technology integrations". The wording CBS quotes about sharing account credentials is in United's news release, not on the help page. For other sectors, see our piece on banks and AI agents and our Visa agentic commerce benchmark decode.

What it means for you: if an AI assistant books or buys for your staff, ask each supplier (airline, hotel, software vendor) in writing whether it allows that assistant, and read its help page. This is not legal advice.

Our take: Signal. United's help page draws its line at acting inside united.com or the app, not at using AI to plan. We did not interpret the contract.

Signal: Is Gemini 4 Carbon's price changing on December 1?

Yes, on Google's price list: Gemini 4 Carbon has one price through November 30, 2026 and a higher one starting December 1, 2026. A token is a chunk of text, often part of a word, and the prices below are per million tokens. This follows our October 28 edition, which reported limited access.

Google's Gemini API changelog, in an entry dated November 10, 2026, labels the model "generally available (GA)". It lists introductory pricing of "$2.00 per 1M input tokens and $10.00 per 1M output tokens" through November 30, 2026. On the pricing page (Standard tab, paid tier), per million tokens:

	Through Nov 30, 2026	Starting Dec 1, 2026
Input	$2.00	$4.00
Output (includes thinking tokens)	$10.00	$20.00

Google's November 10 blog post says the $4 and $20 prices apply after the introductory period and adds "Applies to the Carbon model intended for public access only"; we do not know what that wording excludes.

Our arithmetic: $4.00 / $2.00 = 2.0 and $20.00 / $10.00 = 2.0, so each printed list price per token doubles on December 1. The Batch, Flex and Priority tabs show the same doubling (our arithmetic). That compares rows on a price list; it is not a forecast of any bill, which depends on how many tokens you use. Google's page also says "Gemini 4 Carbon is not available on the Free tier." It does not state why the price is introductory, the time zone of the change, or how many customers use the model.

What it means for you: if a product you pay for runs on Gemini 4 Carbon, ask the vendor whether its own price changes on December 1. This is not advice to switch vendors.

Our take: Signal, with a date on it. The date matters if your usage runs on Carbon directly or through a vendor that bills you for it.

Watch: What is Google's Decisions API, and what does it cost?

That was one Google price tag; here is a second, for a different job. The Decisions API is a different kind of model from Carbon, so we do not compare the two. Google's blog post, dated November 11, 2026, says it takes one input and a list of questions, and returns for each a probability and a confidence score instead of writing text. The model behind it is called Gemini Decision 1.

In Google's example, a photo of a breakroom counter is checked with questions such as "Is there an uncleaned liquid spill on the counter?".

What does Google say about speed and price?

The blog says answers come back in "150-250 ms end-to-end latency" (ms is milliseconds, thousandths of a second); Google gives that figure for inputs up to 16K tokens. It allows "up to 1,000" questions per request. On price, the blog says Gemini Decision 1 is "priced at $0.10 per 1M input tokens and $0 for output tokens".

What stage does Google call it, and what is missing?

The pages we read print no stage label. The changelog says "Released". The guide's limitations line says the API is available on the Gemini Developer API "(v1beta) with Gemini Decision 1 only"; v1beta is a version tag in the code sample, and we do not read a stage into it.

The same line says it "doesn't support context caching, Batch API, or Flex or Priority inference" (optional discount and speed features). The context window (how much text the model can take in at once) is 262,144 tokens "across shared input and all questions", and each question adds its own instructions and criteria to the input. The blog says Gemini Decision 1 "leads across the suite, with the lowest calibration error" (how far its stated confidence is from how often it is right) and shows benchmark comparison images; the scores are Google's and we did not reproduce them. The pages give no accuracy figure for any customer task.

The guide itself notes a limit of the design: "Choice questions always distribute probability across your options, even when none fits." A different vendor described a similar kind of small decision model in our November 10 edition; we do not compare them.

What it means for you: when a vendor says its product uses AI to sort, score or route things, ask whether that step is a text-writing model or a decision model, and what it costs per input.

Our take: Watch. These pages give a price list and a design; how well it works on your data is not in them.

Watch: What does Claroc report about apps built with Grok?

Claroc, a company that says it "secures the apps people build with AI" and offers a related check, says in a post dated November 10, 2026 that most Grok-built apps it checked opened with no sign-in. Grok's Build mode writes an app from a description.

Of 37,571 such apps still online when Claroc checked (its check began on November 8), it says 35,578 (94.7 percent, Claroc's figure) opened for anyone. Of those open apps, it estimates about 1 in 10 held sensitive material and roughly 2% held material "serious enough to rate high or critical". It says it tested changing data only on its own two apps.

The company's side, as Claroc relays it: the company it calls SpaceXSI says apps are private until their creator shares one, employee accounts start as Only you, and publishing takes several steps. Claroc also says that Anyone with the link was already selected in the paid account it used, and that the company took down 24 staff-built apps but not other users' apps.

Claroc's page prints 43,551 addresses tried in one place and 43,601 names listed in another, so we use neither; it does not say whether these apps represent all Grok-built apps.

What it means for you: ask who can open each app your team publishes, and who chose that setting.

Our take: Watch. One company's report on its own research.

Do this week

Four quick checks for an owner or engineering lead; this is not legal or buying advice.

Ask your web person whether your business fits OpenAI's eligibility text for Businesses in ChatGPT, what it would cost (the pages we read state no price) and what the Terms ask of you.
If an AI assistant books or buys for your staff, ask each supplier (airline, hotel, software vendor) in writing whether it allows that assistant.
If a product you pay for runs on Gemini 4 Carbon, ask the vendor whether its own price changes on December 1.
List the apps your team has built with any AI app builder, and check who can open each one and who chose that setting.

Book an AI discovery call

How we know, and what we could not check

We read every source on November 11, 2026 and tested nothing. The OpenAI, United and Google documentation pages were read in a browser; the Google blog, Claroc and CBS pages as page text. Not read: OpenAI's ads page, Business Hub, The Verge's report on United, Google's rate-limits page, and the apps Claroc describes. For how we work with AI search and SEO, see our illustrative AI SEO demo (a demo dataset, not client results).

Signal / Noise (Nov 11, 2026): Businesses in ChatGPT, United's AI Agent Rules, Gemini 4 Carbon's Dec 1 Price, Google's Decisions API, a Grok Apps Report