# Local business data for AI agents: sources and rules

> How an AI agent can list local businesses with ratings, reviews, websites and opening hours: map platforms, open data and business websites, their terms and freshness.

- URL: https://ikag.dev/blog/local-business-data-for-ai-agents
- Author: Ronit Ahuja, Founder, IKAG (https://ikag.dev/authors/ronit-ahuja)
- Published: 2026-10-17
- Updated: 2026-10-17
- Reading time: 5 min
- Publisher: IKAG

## Overview

An AI agent lists local businesses by calling a places source: a map platform’s API for ratings, review counts, websites and opening hours, open map data such as OpenStreetMap or Overture for records you may keep, and each business’s own website to confirm details that change. Every source has terms. Google’s forbid copying business names, addresses and reviews out of its services, and OpenStreetMap’s require credit and share-alike. A good answer gives the source and date of every value and flags what it couldn’t confirm.

## Key points

- No single source has everything: map platforms hold ratings and reviews, open map data holds records you can reuse, and a business’s own website is often the best check on its opening hours.
- Google Maps Platform’s terms forbid copying and saving business names, addresses or user reviews outside its services, but place IDs may be stored indefinitely.
- OpenStreetMap data is free to reuse under the Open Database License if you credit its contributors and share adapted data under the same licence. It does not collect reviews or customer ratings.
- A rating means little without its review count, source and date. Google says it doesn’t verify reviews but removes fake content it identifies, and UK law has banned fake reviews since 6 April 2025.
- A good answer gives each value’s source and date, marks businesses that are closed or unconfirmed, and lists what it couldn’t find.

Ask your AI for every independent gym in a city, with ratings, websites and opening hours, and it needs a places source. A model learned from data gathered before its training ended, and it was never a directory: businesses open, close and change their hours all the time. The agent has to look each business up, and a good answer says where each value came from.

## Where does local business data come from?

Four kinds of source, each good at different things.

**Map platforms.** Google’s Places API returns a place’s name, address, [rating, number of ratings, website, opening hours and business status](https://developers.google.com/maps/documentation/places/web-service/data-fields), plus [up to five reviews](https://developers.google.com/maps/documentation/places/web-service/reference/rest/v1/places). It needs a billing account and a key, and each request is [billed at the highest price tier of the fields you ask for](https://developers.google.com/maps/documentation/places/web-service/usage-and-billing): ratings, websites and hours sit in a higher tier than names and addresses, and reviews in a higher one still.

**Open map data.** [OpenStreetMap](https://www.openstreetmap.org/copyright) is built by a community of contributors and licensed under the Open Database License. It has standard tags for a place’s [opening hours](https://wiki.openstreetmap.org/wiki/Key:opening_hours) and [website](https://wiki.openstreetmap.org/wiki/Key:website), but by [its own rules](https://wiki.openstreetmap.org/wiki/Verifiability) it holds no reviews or customer ratings, because another mapper couldn’t check an opinion. The Overture Maps Foundation’s [places data](https://docs.overturemaps.org/guides/places/) holds tens of millions of places from sources including Meta, Microsoft and Foursquare, under permissive licences, released monthly. Overture says plainly that it contains duplicates, a high junk rate and many missing fields, and gives each place a confidence score to filter on.

**Official registers.** Some ratings are inspections, not opinions. In the UK, the Food Standards Agency publishes [food hygiene ratings](https://ratings.food.gov.uk/open-data) for restaurants, pubs, cafés, takeaways and food shops through a free API, each reflecting the standards found on the date of inspection.

**The businesses’ own websites.** A business’s site is often the most current word on its hours and services, and a site can describe them in a standard format: [schema.org’s LocalBusiness](https://schema.org/LocalBusiness) has fields for opening hours. Reading sites means following each site’s terms and its [robots.txt](https://www.rfc-editor.org/rfc/rfc9309.html).

| Source                    | Good for                                | Terms                                  | Freshness                               |
| ------------------------- | --------------------------------------- | -------------------------------------- | --------------------------------------- |
| Map platform API          | Ratings, review counts, websites, hours | Credit required; no copying out        | Shows if a place has closed             |
| OpenStreetMap             | Places, hours, websites                 | ODbL: credit and share-alike           | Can record when hours were last checked |
| Overture places           | A large list of places you can reuse    | Permissive open licences               | Monthly releases; a confidence score    |
| Food hygiene ratings (UK) | Official hygiene ratings                | The Food Standards Agency’s conditions | The date of inspection                  |
| The business’s website    | The latest hours and services           | Site terms and robots.txt              | As current as the business keeps it     |

## What do the terms allow?

This is where a request that ends “in a sheet” gets interesting. [Google Maps Platform’s terms](https://cloud.google.com/maps-platform/terms) say customers will not scrape or export its content for use outside its services, and give copying and saving “business names, addresses, or user reviews” as an example. They also rule out caching, except where Google allows it, and using Maps content to train AI models. Google’s [general terms](https://policies.google.com/terms) separately forbid automated access that ignores the robots.txt instructions on its pages.

What you may keep is narrow: Google’s [Places API policies](https://developers.google.com/maps/documentation/places/web-service/policies) let you store place IDs indefinitely, and its [service terms](https://cloud.google.com/maps-platform/terms/maps-service-terms) allow coordinates for up to 30 days. Content you show must credit Google Maps, and each review must credit its author and link to the original. Customers with a billing address in the European Economic Area have separate Maps terms.

OpenStreetMap makes the opposite trade. You may copy and adapt its data, but you must credit OpenStreetMap and its contributors, and anything you build on it and distribute must use the same licence. It also says it can’t provide a free map API for third parties, so large jobs work from a downloaded copy or another provider.

None of this is legal advice. The practical rule is to know which source every value came from, because each value carries its source’s conditions with it.

## How fresh is the data?

Businesses close and hours change, so every value needs a date.

- **Status.** Google marks each place as operational, temporarily closed or permanently closed, and operational doesn’t mean open right now.
- **Opening hours.** OpenStreetMap mappers can record when they last checked a place’s hours with a [check_date tag](https://wiki.openstreetmap.org/wiki/Key:check_date). A business’s own website may be newer than either.
- **Ratings and reviews.** Google [says](https://developers.google.com/maps/documentation/places/web-service/policies) it doesn’t verify reviews but checks for and removes fake content when it finds it. In the UK, [fake reviews have been banned since 6 April 2025](https://www.gov.uk/government/news/fake-reviews-and-sneaky-hidden-fees-banned-once-and-for-all), and websites must take steps to prevent and remove them. The same average from a dozen reviews and from several hundred is not the same finding, so keep the count.

## What should a good answer include?

Here is the kind of request you might give your AI. Illustrative:

> “Every independent gym in Manchester rated under 4.2, with their website and review count, in a sheet.”

For each business, a good answer gives:

- **Identity:** name, address and the source’s own ID, such as a Google place ID, so the row can be refreshed later.
- **The rating in context:** the rating, the number of reviews behind it, the source and the date.
- **Website and hours:** with where each came from and when it was last confirmed.
- **Status:** open, temporarily closed or permanently closed.
- **How “independent” was judged:** no source holds that field, so the agent should say what it used, such as finding no other branches.
- **Credit:** the attribution each source requires.
- **Gaps, stated:** “no website listed” or “hours not confirmed” as facts, not blank cells.

It should also respect the source’s terms: an answer you read once is one thing, a spreadsheet you keep is another. For checking the result, see [how to check where an AI answer came from](https://ikag.dev/blog/how-to-check-ai-answers), and for the ways an agent reaches outside data, [how AI agents get live data](https://ikag.dev/blog/how-ai-agents-get-live-data).

## Where IKAG fits

Local business data is one of the next kinds of tools IKAG is building. Your AI, connected once, would ask for the list, and IKAG is designed to call a places source within its terms and return where each value came from, where the source supports it. You would pay per call instead of a subscription to a listings tool, and no prices have been set. IKAG is opening to early users and data partners first.

## Questions

### Can an AI agent read Google Maps reviews?

Through Google’s Places API it can get a place’s rating, its number of ratings and up to five reviews, billed per request. Google’s terms require you to credit each review’s author and forbid copying and saving reviews outside its services.

### Is scraping Google Maps allowed?

Not on Google’s terms. Its Maps Platform terms forbid scraping, exporting or caching Maps content outside its services, including copying and saving business names, addresses and user reviews, and its general terms forbid automated access that ignores its robots.txt instructions. Use the official API within its terms, or open data; this is not legal advice.

### Can I use OpenStreetMap data commercially?

Yes, under the Open Database License: credit OpenStreetMap and its contributors, and share any adapted database you distribute under the same licence. OpenStreetMap says it cannot offer a free map API for third parties, so large jobs work from a copy of the data or another provider.

### How fresh is local business data?

It depends on the source and the field. Opening hours change most often: OpenStreetMap can record when a mapper last checked them, Google marks places as temporarily or permanently closed, and a business’s own website is often the latest word.

## Sources

- [Google for Developers: Place data fields, Places API (New) (last updated 15 October 2026)](https://developers.google.com/maps/documentation/places/web-service/data-fields)
- [Google for Developers: REST resource places, Places API (New) (last updated 6 August 2026)](https://developers.google.com/maps/documentation/places/web-service/reference/rest/v1/places)
- [Google for Developers: Places API usage and billing (last updated 15 October 2026)](https://developers.google.com/maps/documentation/places/web-service/usage-and-billing)
- [Google for Developers: Policies and attributions for Places API (last updated 15 October 2026)](https://developers.google.com/maps/documentation/places/web-service/policies)
- [Google: Google Maps Platform Terms of Service (last modified 26 August 2026)](https://cloud.google.com/maps-platform/terms)
- [Google: Google Maps Platform Service Specific Terms (last modified 15 October 2026)](https://cloud.google.com/maps-platform/terms/maps-service-terms)
- [Google: Terms of Service, UK version (effective 30 July 2026)](https://policies.google.com/terms)
- [OpenStreetMap: Copyright and licence (checked 17 October 2026)](https://www.openstreetmap.org/copyright)
- [OpenStreetMap Wiki: Verifiability (last edited 28 November 2024)](https://wiki.openstreetmap.org/wiki/Verifiability)
- [OpenStreetMap Wiki: Key opening_hours (last edited 5 September 2026)](https://wiki.openstreetmap.org/wiki/Key:opening_hours)
- [OpenStreetMap Wiki: Key website (checked 17 October 2026)](https://wiki.openstreetmap.org/wiki/Key:website)
- [OpenStreetMap Wiki: Key check_date (last edited 18 May 2026)](https://wiki.openstreetmap.org/wiki/Key:check_date)
- [Overture Maps Foundation: Places guide (September 2026 release)](https://docs.overturemaps.org/guides/places/)
- [Food Standards Agency: UK food hygiene rating data API (checked 17 October 2026)](https://ratings.food.gov.uk/open-data)
- [Schema.org: LocalBusiness (version 30.1, 16 September 2026)](https://schema.org/LocalBusiness)
- [RFC 9309: Robots Exclusion Protocol (September 2022)](https://www.rfc-editor.org/rfc/rfc9309.html)
- [GOV.UK: Fake reviews and sneaky hidden fees banned once and for all (6 April 2025)](https://www.gov.uk/government/news/fake-reviews-and-sneaky-hidden-fees-banned-once-and-for-all)