Businesses in ChatGPT: a readiness guide for claiming your profile and product feed
OpenAI launched Businesses in ChatGPT in beta on November 10, 2026. Businesses can claim a profile, supply a product feed, and let a Business Agent answer customer questions from that data. Here is how to get ready while it is still early.

In this article
Short answer
Businesses in ChatGPT is a beta program OpenAI launched on November 10, 2026. A business claims a business profile, supplies a product feed, and a Business Agent answers customer questions using that profile and feed. Readiness means three things: your profile facts are accurate, your product data is structured and current, and someone in your business owns keeping both up to date.
Key takeaways
- Businesses in ChatGPT launched in beta on November 10, 2026, so a business that sets up now is early, and early setup is cheap compared with fixing wrong information later.
- The Business Agent answers from the data you supply. Accurate profile facts and a clean product feed are the whole game.
- The agent answers factual questions well when the source data is complete. It goes thin, or wrong, when data is missing, stale or vague.
- A product feed is only as good as the data hygiene behind it. Names, descriptions, prices and availability need an owner and an update routine.
- Readiness is a short checklist, not a project: claim the profile, complete every field, clean the feed, test the answers, and review on a schedule.
On November 10, 2026, OpenAI launched Businesses in ChatGPT in beta. The idea is simple enough to explain in one breath. A business claims a profile inside ChatGPT, supplies a feed of its products, and a Business Agent answers customer questions using that information. A customer can ask what you sell, what it costs, when you are open, and how to reach you, and get an answer drawn from data you provided, instead of a guess drawn from the open web.
That last part is the point of the whole program. ChatGPT already answers questions about businesses today, and when it has no reliable source, it fills the gaps as best it can. Sometimes it fills them wrong. A claimed profile and a clean product feed give it a source it can trust: yours.
This post is a readiness guide. It covers what the program is and why setting up early is worth a small amount of time. It then works through claiming and completing the profile, getting a product feed into shape, and where the Business Agent answers well and where it goes wrong. No hype, and no promises about how big this becomes. It is a beta. The honest case for acting now is that the setup is cheap insurance, and the work it asks for is work your business data probably needs anyway.
What Businesses in ChatGPT actually is
Strip away the launch language and the program has three parts.
A business profile you claim. This is the record of who your business is: its name, what it does, where it is, when it is open, and how a customer gets in touch. Claiming it means the record comes from you, not from whatever the model has picked up elsewhere.
A product feed you supply. This is your product data in a structured form: the products themselves, described clearly, with prices and availability. Structured matters. A feed is data a system can read field by field, not a page of marketing copy it has to interpret.
A Business Agent that answers from both. When a customer asks a question, the agent answers using the profile and the feed. It is not browsing your website and forming an opinion. It is reading the information you handed it and answering in conversation.
That is the whole mechanism, and it explains everything else in this post. The agent has no knowledge of your business beyond what you supply and what it can already find. Your readiness is therefore a data question, not a technology question. If the facts are right and current, the answers are right. If they are not, no setting fixes that.
Why set it up now, while it is early
The honest starting point: demand here is new, and nobody can measure yet how many customers will ask ChatGPT about a business like yours. Anyone who quotes you adoption numbers for a program that launched in beta on November 10, 2026 is guessing. So the case for setting up now is not that a wave of customers is definitely coming. The case is smaller and stronger than that.
The setup is cheap. Claiming a profile and completing it is writing work, not building work. Preparing a feed is cleaning data you already hold. For most businesses this is hours of effort, not a project with a budget line.
Early is quiet. While a program is in beta, few businesses in your market will have claimed a profile. There is no queue, no competition for the listing, and no penalty for being careful and slow. You can complete the profile properly, check it twice, and be done before it matters.
The work pays off even if the program changes. Beta programs change. Features get renamed, merged or dropped. But an accurate, written-down set of business facts and a clean, current product list are useful everywhere: your website, your listings, your own support team. If you do this work for ChatGPT and the program shifts, you still own better data than you had before.
Wrong information is the real risk, and it exists today. Customers already ask chatbots about businesses. An agent answering from incomplete information it found somewhere else can quote a wrong price or old hours right now, without you ever supplying a thing. A claimed profile is how you put your version, the correct version, in front of it.
Set up now, then, not because the future is certain, but because the cost is low, the downside of messy data is real, and the cleanup is overdue in most businesses anyway.
Claim and complete your business profile
The profile is the set of facts the Business Agent will quote most often, so treat every field as something a customer will hear read back to them. Complete it in this order.
Name. Use the exact name your customers know you by, spelled the way it appears on your shopfront, your invoices and your website. Not a legal entity name nobody recognises, and not a keyword-stuffed version. The agent will repeat this name constantly; make sure it is the right one.
Description. Say plainly what the business does, who it is for, and what it sells. Write it the way you would describe the business to a new neighbour: short sentences, no jargon, no superlatives you cannot support. A vague description produces vague answers. If your description could belong to any business in your trade, it will help the agent distinguish yours from none of them.
Locations. List every location a customer can actually visit or buy from, with the address in full. If you serve an area rather than hosting visitors, say so clearly. A customer who is told to visit an address that is really a warehouse, or a head office that takes no walk-ins, has been misled by your data, not by the agent.
Hours. Opening hours, including how weekends and holidays differ if they do. Hours are the classic stale fact: they change for a season, someone updates the sign on the door, and every other record keeps the old ones. Whatever hours you put in the profile, put a note in your calendar to check them when they next change.
Contact paths. How should a customer reach a human? A phone number that is answered, an email address that is monitored, a contact page that works. The agent will hand these out when a question goes beyond its facts, and that handoff only works if the path at the end of it is real. Test each one yourself before you list it.
Two habits make the profile trustworthy. First, one source of truth: decide where the correct version of these facts lives in your business, and make the profile match it, so updates happen in one place and flow to the other. Second, one owner: a named person who checks the profile when anything changes. Profiles do not go stale all at once. They go stale one unrecorded change at a time.
Get your product feed ready
If the profile is who you are, the feed is what you sell. It is also where most businesses will discover that their product data is in worse shape than they thought. That discovery is the valuable part.
A ready feed has four qualities.
Accurate names. Each product named the way customers ask for it, consistently. If the same product appears under two names in two systems, pick one. An agent matching a customer question to a feed entry needs names that line up with how people actually speak.
Descriptions that answer questions. A good feed description covers what the product is, who it suits, and the details a buyer checks before deciding: size, material, duration, what is included, whatever applies in your trade. Write for the question, not for the brochure. Marketing adjectives do not help an agent answer; facts do.
Current prices. The price in the feed must be the price a customer will actually pay, and it must change when the real price changes. Price is the field where staleness does the most damage, because a quoted price feels like a promise. If your prices vary too much to state plainly, say how pricing works rather than letting an old figure stand in for an answer.
Honest availability. In stock, out of stock, discontinued, seasonal: the feed should say which, and someone should update it when the answer changes. An agent that happily recommends a product you stopped selling has not failed. Your availability data has.
Underneath all four sits the same requirement: data hygiene. Most businesses hold product data in several places at once, a shop system, a spreadsheet, a website, someone's memory, and the copies disagree. Before you supply a feed, reconcile them. Choose the system that will be the source, fix its errors, and build the feed from it. Then set the routine that keeps it true: who updates prices, who marks availability, and how often someone checks the whole list rather than only the items that changed.
A small, accurate feed beats a large, doubtful one. If you cannot vouch for every field of every product today, start with the products you sell most, get those right, and extend the feed once the routine is working.
What the Business Agent answers well, and where it goes wrong
Knowing what to trust the agent with is the difference between a useful setup and a disappointing one. The pattern is consistent: it is strong on facts you supplied, and weak everywhere else.
What it answers well
Questions with a definite answer that lives in your data. What do you sell, and what does it cost? When are you open? Where are you? How does a customer reach you? What is your policy on the things your profile states plainly, such as bookings, delivery or returns, if you have written them down?
These are lookups, and the agent does lookups well. It finds the fact in the profile or the feed and states it in natural conversation, at any hour, without a queue. For a business whose staff answer the same factual questions every day, that alone is worth the setup time.
Where it gets thin or wrong
The failures follow the data, every time.
Missing data produces thin answers. If the feed has no description detail, the agent has nothing to work with, and the answer shrinks to a name and a price. It will not invent the missing detail reliably; it will either say little or reach for general knowledge about products like yours, which may not be true of yours.
Stale data produces confident wrong answers. This is the dangerous failure, because nothing in the answer sounds uncertain. An old price, last season's hours, a discontinued product: the agent states each one as fact, because as far as it knows, it is one. Garbage in, confident wrong answers out.
Vague data produces vague answers. A description that says a product is high quality and suitable for everyone gives the agent nothing to match a customer's real question against. The answer comes back equally general, and the customer leaves no wiser.
Questions beyond your data are beyond the agent. Advice that depends on a customer's situation, judgement calls, complaints, negotiations: these need a human. A well-set-up profile handles this honestly by giving the customer a working contact path, which is why the contact fields matter as much as the product fields.
The lesson is not that the agent is unreliable. It is that the agent is a mirror. It shows your customers exactly the quality of information you gave it, in a confident conversational voice. Readiness means making sure the mirror has something accurate to show.
The readiness checklist
Run through this in order. Each step is done when you can point at the result, not when you have thought about it.
| Step | Action | What done looks like |
|---|---|---|
| 1 | Claim the business profile | The profile is claimed by your business, in the exact name customers know |
| 2 | Complete the description | A plain description of what you do, who it is for, and what you sell |
| 3 | Check locations and hours | Every location listed in full, hours current, holiday differences noted |
| 4 | Test the contact paths | Every phone number, email and contact page in the profile tested by you and working |
| 5 | Reconcile your product data | One source system chosen, conflicting copies fixed, names consistent |
| 6 | Build the feed from that source | Names, descriptions, prices and availability accurate for your best-selling products first |
| 7 | Test the agent like a customer | You have asked it your ten most common customer questions and checked every answer against the truth |
| 8 | Name an owner and a review date | One person owns profile and feed updates, with a recurring date to recheck prices, hours and availability |
Step 7 deserves emphasis. Testing is where readiness becomes real. Write down the questions your customers actually ask, the ones your team hears weekly, and put them to the agent. Every wrong or thin answer points at a specific field in your profile or feed. Fix the field, not the answer, and test again. That loop, run once at setup and again whenever your data changes, is the whole maintenance job.
What the setup costs in time
There is no platform to buy here, and this post will not invent a price for a beta program whose terms can change. The real cost is staff time, and it splits into two parts.
The setup itself is a few hours of writing and checking for the profile, plus however long your product data takes to reconcile. That second figure depends entirely on the state of your data today. A business with one tidy product list will finish quickly. A business with three disagreeing spreadsheets has found its real task, and it was never really about ChatGPT.
The upkeep is smaller but permanent: an owner who updates the profile and feed when facts change, and a scheduled review, monthly is a sensible rhythm for prices and availability, to catch the changes nobody recorded. Budget it the way you budget keeping your website accurate, because it is the same job pointed at a new surface.
That is the readiness picture in full. Claim the profile. Complete it with facts you have checked. Feed it product data you would stand behind. Test what it says about you. Then keep it true. Businesses in ChatGPT is in beta, and it may grow, stall or change shape. A business that has done this work is ready in every version of that future, because its information is finally in order, and an agent, a website or a new hire can all answer from it correctly. If getting your product data and business systems into that shape is the part you need help with, that is the work RaftLabs does: AI agent development services built on data your business can trust.
Ask an AI
Get an instant summary of this post from your preferred AI assistant.
Common questions
- It is a beta program OpenAI launched on November 10, 2026. A business claims a business profile in ChatGPT, supplies a product feed with its product data, and a Business Agent answers customer questions using that profile and feed. Instead of guessing from the open web, the agent answers from information the business itself provided, which is why the quality of that information matters so much.
- Because the setup is mostly work you should do anyway: writing down your accurate business facts and cleaning up your product data. Doing it now, while the program is in beta and few businesses have claimed a profile, costs little time. Waiting means doing the same work later, in a rush, while an agent may already be answering questions about your business from incomplete information you did not supply.
- The facts a customer asks about, in a structured form: product names, clear descriptions, current prices, and whether each product is available. Every field should match what your website and your team would say today. If a price changed last month and the feed still carries the old one, the feed is wrong, and the agent will repeat the wrong price with confidence. Accuracy and freshness matter more than size.
- The Business Agent answers from the data it has. Wrong or stale data does not produce a cautious answer; it usually produces a confident wrong one, such as an old price, old opening hours, or a product you no longer sell. Missing data produces thin answers that say little and help nobody. The fix is not a better prompt. The fix is accurate source data, checked on a schedule.
- No. The agent answers factual questions from your profile and product feed: what you sell, what it costs, when you are open, how to reach you. It does not replace the judgement, problem solving and relationship work your team does, and it does not replace your website, which remains the full record of your business. Think of the agent as a new front desk for simple questions, one that only knows what you have told it.
