How to Cold Email Local Businesses With Google Maps Data and AI: A Practical Setup and What It Costs

What a 100-lead test actually looks like

A builder documented on this site (builtwithagents.ai/strategy/openclaw-free-lead-generation-apify-google-maps) pulled 100 HVAC contractors in Phoenix from Google Maps in five to ten minutes, scored each one, and had AI draft a personalized email for each into Gmail, ready for review. About half of the businesses had an email address listed on their website. That second number is the one to remember: half. A list of 100 businesses is a list of roughly 50 people you can email.

This guide is the version of that pipeline that you can run this week, with a sending setup that will not burn your main domain and a legal minimum you should know before you hit send.

Why local businesses are a good target

Local service businesses have the problems an AI offer can solve and the owners read their own email. A plumber who misses calls, a dentist with an old website, a roofer who answers leads at night: these owners decide alone, quickly, and they are findable by category and city. You can describe the business in one sentence and the pain in another.

They also get pitched constantly, so most generic messages are ignored. Cold outreach here works only when the email shows you looked at their business and offers something specific.

What reply rates to expect

Be skeptical of the numbers floating around. One builder here (builtwithagents.ai/strategy/google-maps-claude-code-outbound-machine-local-leads) describes a Google Maps and Claude Code system and projects a 2 to 6 percent reply rate on hyper-local personalized email. He labels these projections, not verified results. Use them as a ceiling for a well-targeted campaign, not a forecast.

A fair planning assumption: out of 300 emails, you may get a handful of replies, a few conversations and maybe one client. If one client is worth a few hundred dollars a month to you, that is still a good return on a $50 setup. Plan for the low end.

What it costs

As of October 2026, Apify's Google Maps scraper lists prices from $1.50 per 1,000 scraped places on a pay-per-event basis, with extras such as contact enrichment (which pulls emails and social profiles from the business website) charged separately. A 500-place scrape with enrichment is a few dollars.

Instantly's pricing page lists the Growth plan at $47 a month ($37.60 a month billed annually) with 5,000 emails a month and 1,000 uploaded contacts. Other sending tools exist and are similar in price. You also need a sending domain and mailbox separate from your main business domain, which runs roughly $15 to $30 a month depending on how many mailboxes you buy, an estimate that varies by provider. AI writing on a small model costs cents for a few hundred leads. All in, expect $60 to $100 in the first month, much of it fixed.

The legal minimum

The US CAN-SPAM Act applies to business-to-business email. There is no B2B exemption. According to the FTC's compliance guide, you must use accurate From, To and Reply-To information, write subject lines that match the content, identify the message as an advertisement, include a valid physical postal address, give a clear way to opt out, honor opt-outs within 10 business days and monitor anyone you hire to send for you. Each violating email can carry a civil penalty of up to $53,088. Rules outside the US, including the UK and EU, are different and often stricter, so check them if you target those markets. This is not legal advice.

In practice: put your real name, business and mailbox address in the footer, include a one-line unsubscribe, and actually remove people.

Step by step

Step 1: Pick one niche and one city

Choose a trade you can help, such as HVAC, roofing or dental offices, and one metro area. A niche lets you write one offer and refer to real details. Write the offer in one sentence: for example, texting every missed call back within a minute so the lead is not lost.

Step 2: Scrape the businesses

In Apify, open the Google Maps scraper, enter the category and city (for example "HVAC contractor, Phoenix AZ"), set a limit of 300 to 500 places and turn on the contact enrichment option to pull emails from websites. Export the results as a spreadsheet. Keep the name, website, phone, rating, review count and email columns.

Step 3: Filter hard

Delete chains and franchises, businesses with no website or email, and anyone outside your target size. A business with 8 reviews and one with 800 are different prospects. Aim to end with 150 to 250 rows you would actually be glad to work with. A smaller, cleaner list performs better than a big one.

Step 4: Verify the emails

Run the list through an email verification service or the verification add-on offered by your scraper. Remove invalid and risky addresses. Sending to bad addresses hurts your sending reputation quickly, which is the main reason new cold email setups fail.

Step 5: Write one true personal line per lead

Give each row's website text or review summary to an AI model and ask for a single sentence noting something specific and verifiable: that they offer 24-hour emergency service, that their reviews mention fast arrival, that they just added a second location. Tell it to return nothing if it cannot find a real detail, and read at least 30 of the lines by hand. Fabricated compliments are the fastest way to look like spam.

Step 6: Write the email

Keep it under 90 words. Open with the personal line, state the problem you solve in one sentence, give one proof point you can honestly claim (if you have no client results yet, say you are looking for a first local case study and offering the setup free or discounted), and end with a low-effort question. No attachments, no images, no tracking links. Add your address and opt-out line.

Step 7: Set up the sending side

Buy a separate domain that is close to your brand, create one or two mailboxes, set up the authentication records your provider asks for (SPF, DKIM and DMARC), and warm the mailboxes in the tool for a couple of weeks before sending. Load your contacts in Instantly or a similar tool and start slowly, at a few dozen emails a day per mailbox, increasing only if bounces stay low.

Step 8: Reply by hand and track

Read every reply yourself and answer within hours. Track sent, bounced, replied and booked in a sheet. After the first 200 emails, change one thing at a time: the offer, the subject or the niche.

A worked example

Say you sell missed-call text-back to HVAC companies. You scrape 400 HVAC businesses in one metro for a couple of dollars. Filtering out chains and rows without a website leaves 250, and about half have a findable email, so roughly 125 people. Verification trims that to around 110. You write a personal line for each, read 30 of them, fix the weak ones and load the list. At a few dozen emails a day it takes about a week to send everything. If you land in the low single digits for reply rate, that is two to five replies. Two become calls, and with luck one becomes a paid pilot. That is a thin funnel, and it is the honest one: the work is repeating it across niches and cities, and improving the offer each round. These numbers are illustrative.

When not to do this

Do not do this with a purchased list of unknown quality, from your main domain, or without an unsubscribe. Do not do it if you have nothing to offer yet; a good offer beats a clever pipeline. And if your target customers are consumers rather than business owners, other channels work better and the rules are stricter.

Frequently Asked Questions