My leads agent finds businesses through Google but keeps only Google's place ID

In plain English

My leads agent now has a first step: finding local businesses to check. It asks Google for businesses of a given kind in each postcode district (the first half of a postcode, such as SP1) near Salisbury. It drops chains and any business already on file, then shows Simon the list. Google’s terms say a business’s name and address from Google Maps must not be copied and saved. Simon chose to keep only Google’s code for each place. Every other detail of a business comes from somewhere else, such as Companies House. For now the step saves nothing at all. It does not read the trade directories, because their terms rule out automated or business use.

Context

The finding step was built last on purpose (decision 0015), so no money went on searches before the checks that follow them worked. Before one of my agents wrote any code, it read the terms of each place the step could look, on 14 November 2026.

Google. Google Places is Google’s service for searching its business listings from a program rather than a web page. To cache is to keep a copy to use again. The Google Maps Platform Terms of Service (last changed 26 August 2026) say the customer will not “copy and save business names, addresses, or user reviews” (section 3.2.3(a)(iii)). They also say it will not “cache Google Maps Content except as expressly permitted under the Maps Service Specific Terms” (section 3.2.3(b)).

The Service Specific Terms (last changed 28 October 2026) allow two exceptions:

• A place ID, Google’s own code for a place, may be kept: “Customer may cache (a) place_id from Places API” (General Service Terms, section 3, “Google ID Caching”).
• A map position may be kept for up to 30 days (section 14.3, under “Places API (Legacy and New)”).

The Terms of Service also rule out using “latitude/longitude values from the Places API as an input for point-in-polygon analysis”: working out from a map position whether a place sits inside an area (section 3.2.3(c)(iv)). Google’s attribution rules (section 3.2.2(b) of the Terms of Service) point to the Places API policies (last updated 9 November 2026). Under “Google Maps attribution requirements”, those say “It must always be clear to end users which content is provided by Google Maps”. So results shown without a map must be labelled as coming from Google Maps.

These are the terms for a customer whose billing address is outside the European Economic Area, which the UK is no longer part of. Both documents say so in their opening paragraphs, before section 1.

The trade directories. Decision 0015 says my agents look in trade directories for contact details. The terms of all three that Simon named stand in the way of an agent reading them:

• Checkatrade’s Acceptable Use Policy (last updated August 2025) says not to “use bots, scrapers, or other automated tools without consent” (section 4, “Prohibited Uses”, under “Damage the Platform”).
• Which? Trusted Traders’ terms (undated, read 14 November 2026) say: “You must not use or encourage others to use our website (or its contents) for their or your commercial use or in a way that could damage the reputation of Which?” (section 3, “Other important things you need to know”).
• TrustATrader’s terms (last updated 28 June 2021) say visitors agree “not to use our Website for any commercial or business purposes” (under “Our responsibility for loss or damage suffered by you”, in the part for visitors).

Options

1. Save the full results. A list of names and addresses to work through later. Google’s terms forbid it.
2. Save only the place IDs. Allowed. A later run can skip a place already shown. Each record can say which place it came from. A search costs the same whether or not its results are new, so it saves no money.
3. Save nothing. Show the list, then forget it. A business is skipped next time only once it is in the records store (the small database from decision 0016), through the business check.

Decision

Simon chose option 2 on 14 November 2026, with one rule added: keep Google’s place ID only. A business’s name, address and company number go into the records store only from another source, Companies House or the business’s own website, never copied from Google.

This step doesn’t save anything yet. Keeping place IDs and confirming each business from another source before it goes in the store is issue #74. Until then, one of my agents built the finding step like this:

• Three fields only. The search asks Google for the place ID, the name and the address. It never asks for the map position, so the area is judged by the postcode in the address.
• The list says where it came from. It starts “Results from Google Maps” and says it is not saved.
• A spending cap. Each request is billed. A run makes no more than 20 unless Simon sets another number. It says how many it made.
• The area. The step searches 15 districts: SP1 to SP11, BA12, SO20, SO51 and BH24. For now that is the 20-mile area around Salisbury from decision 0013, plus four borderline districts Simon added: SP8 (Gillingham), SO20 (Stockbridge), SO51 (Romsey) and BH24 (Ringwood). BA10 (Bruton) is left out, as it lies beyond 20 miles. Simon may refine the list.
• No directories. The agent never reads Which? Trusted Traders, Checkatrade or TrustATrader.

Google’s Terms of Service ask that an app’s own terms of service say it uses Google Maps and point to Google’s terms and privacy policy (section 3.2.2(a)(i)). Simon chose not to publish any for this, because the finding step is a tool only Simon uses.

Simon also chose how the directories are used by hand. Contact details are looked up by hand on the business’s own website and on Checkatrade, whose terms ban only automated reading. Which? Trusted Traders and TrustATrader are not used at all, even by hand, because their terms allow personal use only. My agents still never read any directory. Decision 0015 now carries a note saying the same.

Update, 14 November 2026: Simon raised the spending cap from 20 requests to 40. Issue #74 built the rest of this decision. The finding step now keeps each place ID it shows. A business goes into the records store only once Companies House confirms it.

Update, 14 November 2026: decision 0019 lets Simon confirm a sole trader or partnership by hand from its own website. It also explains why the GOsC and HCPC registers can’t confirm a business and how letters now use the name a business gives on its own website.

Consequences

• The finding step never saves a list of businesses from Google, on Simon’s computer or in my repository.
• Until issue #74 is built, a business shown but not taken forward appears again in the next search that finds it. It costs no extra searches.
• Under Simon’s rule, a record’s details come from Companies House or the business’s own website, so the record does not depend on what Google sent. Google’s Places API policies say, in the section on data storage and caching: “Consult your legal counsel to determine whether your implementation complies with the terms”. No lawyer has looked at this.
• If the finding step is ever used by anyone other than Simon, the terms of service question comes back.
• Checkatrade is now the only directory used for contact details. It is used only by hand. For a business that isn’t on Checkatrade and has no working website, there are fewer places to find its contact details.