Can ChatGPT find new businesses? What AI can and can’t do
Auspector ·
Partly: with web search, ChatGPT can find new businesses that were written about somewhere its searches reach — local news, a business’s own website, an announcement — and help you research and approach them, but it does not produce a complete or dated list, gives no count, and can answer the same question differently twice. For scale: on October 23, 2026 alone, at least 58,686 records were added to our newly opened businesses section.
How does ChatGPT look for businesses?
When a question would benefit from current information, ChatGPT can search the web — on its own, or because you asked it to. According to OpenAI’s help centre, it typically rewrites your request into one or more targeted search queries, sends them to its search partners, reads the results and may follow up with narrower queries. For local questions it can use your approximate location, estimated from your IP address, and if memory is on, saved details can shape the queries too (OpenAI · Searching the web with ChatGPT, read on October 24, 2026).
So the answer to “Can you give me a list of businesses that opened this week in Austin?” is built from whatever those few searches return. What is not published, or cannot be reached, is not in it. OpenAI notes that the model may be unable to read a site because of technical problems, a paywall or the site’s robots.txt settings (OpenAI · Does ChatGPT tell the truth?), and sites that opt out of OpenAI’s search are not shown in ChatGPT’s search answers (OpenAI · OAI-SearchBot). Other assistants with web search work along similar lines: Claude’s help centre, for example, describes Claude calling a search tool when a question needs current information and citing the pages it used (Anthropic · Enable and use web search).
What it does well
- Openings that were announced. A press release, a local news piece about a new restaurant, a “we’re open” page on the business’s own site: if a search can reach it, the assistant can find it and give you the link.
- Researching one business. Before a call: what it sells, who it serves, what its website says about the opening, what its menu or service list looks like. Ask for sources and open them.
- Drafting. An opening line, call notes, a short follow-up — quick to produce and easy to adapt to your offer.
- Working on a list you already have. Sorting, grouping by district, flagging rows to check. Copy-paste prompts for this are in AI prompts for prospecting.
Where it falls short as a lead source
- No coverage guarantee. A business that opened without being written about anywhere a search reaches will not appear — and the answer cannot tell you what it missed.
- No count. An answer built from search results is a sample of what those searches turned up. It cannot tell you how many businesses opened in a city this week, or what share of them it found.
- Not repeatable. OpenAI’s developer documentation describes its models as non-deterministic by default: the same request can produce a different output (OpenAI · Reproducible outputs). Location and memory feed into the search queries as well, so asking twice — or a colleague asking the same thing — can give a different list.
- It can be wrong. OpenAI’s help centre says search results and citations “can be incomplete, outdated, or incorrect”, and lists made-up facts and references to non-existent sources among the errors a model can make. A name, address or phone number in a chat answer has to be checked against its source.
- No reliable date. The date on a news page is when it was published, not necessarily when the business opened — and a “coming soon” story may describe a place that has not opened yet. OpenAI itself advises checking when a cited page was published or updated.
How big is the gap?
Those 58,686 records were a single day’s arrivals. In the 30 days to October 23, 2026, 2,183,676 records entered our newly opened businesses section across the 246 territories in the database. A chat answer built from a few searches is a different kind of thing: useful for one city and one category, but neither complete nor countable. A list at that scale needs a database that is refreshed every day, keeps new records apart from older ones and dates them by when they arrived (how fresh the data is).
| Task | AI assistant with web search | Dated new-business list |
|---|---|---|
| Which businesses opened in a city this week? | Some: those written about where its searches reach | Every record that entered the list, filtered by arrival date |
| How many? | No reliable count | A count, before you pay |
| The same answer tomorrow? | Not guaranteed: answers vary | Same filter, same records; new ones added daily, records older than 30 days leave |
| Phone and address | May be outdated or wrong: check the source | Fields on the record; still verify before calling |
| What does the business do, what is on its site? | Strong: reads its pages and cites them | Category and website on the record only |
| Opening line, call notes | Strong, given your context | No |
A sensible division of labour
- Coverage from a dated list. Filter new businesses by country, city, category and arrival date, and check the count before you open anything — search and counts are free on every Auspector plan. The steps are in building a lead list from new businesses.
- Prioritise with the assistant. Export the shortlist as CSV or XLSX and ask the assistant to rank it against your criteria, group it by district or draft openers. Tell it to use only the columns in the file and to write “unknown” rather than guess.
- Research the top of the list. For the ten or twenty records you will actually contact, let the assistant search the web for each one, and open the sources it cites.
- Verify before you contact. Check that the business is open and the number is right. Rules on contacting businesses without consent differ by country, and the obligation sits with the sender. A person, not a tool, sends each message.
Other ways to find new businesses — company registries, local permits, walking the area — are compared in how to find newly opened businesses.