How to Find Local Business Leads with ChatGPT and Claude (2026)

Why the list comes back thin

On its own, a chat assistant is not connected to Google Maps listings. When you ask it for local businesses, it answers from whatever it can find on the web in that moment, or from what it remembers, which may be out of date. That gives you a handful of names, sometimes with numbers, and no way to tell whether they are the whole market or a tenth of it.

For lead generation that is the wrong way round. The value of a list is that it is complete for the area you chose, so the questions you ask of it (who has no website, who has few reviews, which city is under-served) have real answers. The fix is to give the assistant the data rather than asking it to find the data.

There are two ways to do that today.

Option 1: connect Claude to Google Maps data

A connector lets Claude look things up in an outside service while you chat. The Maps Harvest connector is free, needs no Maps Harvest account, and takes about a minute to add:

1. In Claude, open Settings, then Connectors, and choose Add custom connector.
2. Name it Maps Harvest and paste https://www.mapsharvest.com/mcp as the URL.
3. Save, start a new chat, and ask about a business type in a city.

Claude then answers from Google Maps listings instead of guessing: the average rating, the typical number of reviews, the share of businesses without a website, the share showing a phone number, and up to ten example businesses. When we asked it for auto repair shops in Houston on October 14, 2026, half of the listings in the sample had no website, which is the kind of number that decides whether a web design campaign is worth running there.

ChatGPT: Maps Harvest has been submitted to the ChatGPT directory. Until it is listed, the same link works in any ChatGPT plan that lets you add a custom connector.

Questions to ask once it is connected

The connector answers questions about one business type in one city at a time, and can tell you which markets it covers. These work well:

• “How competitive are roofing contractors in Houston on Google Maps?”
• “Find auto repair shops in Houston that don't have a website.”
• “What rating and review count does a dentist in Los Angeles need to look established?”
• “Compare plumbers in London and Birmingham: ratings, reviews, and how many have no website.”
• “Which cities do you have data for hair salons?”

Comparisons are where it earns its place. Ask for two cities, or two trades in the same city, and let the assistant point out the gap: the city where a third of the businesses have no website, or the trade where a few dozen reviews already puts you in the top half.

Option 2: give ChatGPT or Claude your own lead file

This works in both assistants today, and it is the better option once you have chosen a market. Export a list from Maps Harvest as CSV or Excel, attach it to the chat, and ask questions about the real rows. A file of every dentist in a state is too long to scan by eye, but an assistant can sort, group and summarise it in seconds.

Two habits make the answers better. Keep the Google Maps URL column in the export, so every row the assistant mentions can be opened and checked. And ask for counts alongside any summary ("how many rows did you use?"), so you notice if it quietly worked on part of the file.

Where the assistant stops and the list starts

The connector is for deciding: which market, which trade, what a strong business looks like there. It answers with a sample of each market, phone numbers show only their first three digits, and it covers major cities in five countries.

When the decision is made, the list itself comes from a scrape: every business of that type across the cities, county or state you pick, with up to 23 fields each, filtered as it runs (minimum rating, review counts, website or no website, phone or no phone). Then the file goes back into the assistant for the sorting and the openers. Our guides on finding businesses without a website and building a cold-calling list cover that part step by step.

FAQ