AI, ads, and roofing:
what I’d use now.

Give ad tools better direction.

Ad platforms are taking on more of the work of creating and selecting ads. Google expanded its text-guidelines beta in February, giving advertisers ways to tell its AI which terms and concepts to avoid. Meta’s October research describes an ad recommendation model that learns across its advertising surfaces.^[1][2]

For a roofer, that makes a clear brief useful. Tell the system what you sell, where you work, and what you can promise.

For example, a roof-repair campaign could use:

• The actual towns your crew serves.
• The roof types and repair work you handle.
• A clear explanation of what the inspection includes.
• Real repair photos and a short explanation from your estimator.
• A page about roof repair with a clear inspection request.

Review the generated ads and the pages they send people to. Check claims about warranties, prices, insurance, and timing against what the business can deliver. Google’s text guidelines are a beta feature; check availability in the account before planning around them.

I would test a few useful angles: a common leak problem, an inspection walkthrough, and a completed repair. Track which one brings suitable customers.

Make your real work easy to find.

Google published a dedicated AI-search optimization guide in May. Its guidance keeps established SEO at the foundation and points businesses toward original experience, useful information, and relevant photos and video.^[3][4]

A roofing company already has good material for this. Your team can explain what caused a leak, why one roof could be repaired, or what led a homeowner to choose a replacement.

With the customer’s permission, turn a completed job into a short project story. Explain the problem, what the team found, the options, and the work completed. Use photos that show those details. Add the general area and link to the relevant service page.

That gives a homeowner something useful to judge and gives search systems clearer information about the business. It does not guarantee an AI mention.

Google also made its dedicated generative-AI performance reports available worldwide by August 31. Those reports can show which pages receive impressions in supported AI features. Use that to watch discovery, then check calls and inspection bookings separately.^[5]

For a more detailed starting plan, read my AI search guide for contractors.

Use AI to help with the first conversation.

GoHighLevel’s current Voice AI documentation describes structured call flows and appointment booking against configured calendars. Booking support depends on the builder and available actions; some setups need a configured integration.^[6][7]

Here is a roofing example I would test: an after-hours caller wants a roof inspection. The assistant collects their name, number, town or ZIP code, and reason for calling. If the request fits your services, it offers a time from the right calendar. A team member reviews the record.

Give the assistant a clear route to a person when the caller needs one. HighLevel distinguishes transfer to another AI agent from transfer to a human phone number. Check the actual call-transfer action and destination.^[8]

Start with a small job for the assistant: collect accurate details and help arrange the next conversation. Leave roof diagnosis, pricing decisions, and insurance questions with the people qualified to answer them.

Before putting it on a live number, test a normal booking, an unavailable time, an address outside your service area, and a caller asking for a person. Confirm the calendar entry, notification, and handoff. Check the account’s available features and costs too.

Follow the lead all the way to the job.

On October 28, Microsoft announced general availability of its Conversions API. It gives advertisers a way to send conversion information from sources such as a website or CRM directly to Microsoft Advertising.^[9]

The useful roofing application is to connect advertising with what happens after the form or phone call.

1. 01InquiryA call or form arrives
2. 02InspectionA suitable visit is booked
3. 03Job wonThe customer accepts
4. 04RevenuePayment is collected

Keep those stages separate in your records. Track the source, service requested, inspection outcome, job value, and collected payment. Review the margin on the work you win.

Once that record is dependable, ask your provider which verified outcomes can be sent back through the ad platform’s supported integration. Microsoft’s announcement does not establish that a particular GoHighLevel connection is ready.

I would judge a marketing change by the suitable work it produces and what that work contributes to the business. A lower cost per form is one part of that picture.

Keep AI’s business facts current.

An October Nielsen Norman Group study of regular Claude users found problems with stale and hard-to-find context. It was a study of AI users, rather than roofing companies, but the operating lesson is useful here: AI needs current information to work from.^[10]

I would keep one approved roofing brief with your services, coverage area, current offer, booking rules, warranty wording, and real project examples. Name the person who updates it.

Use that brief when asking AI to draft an ad, organize a project story, or prepare answers to common questions. Have someone who knows the business check the work before it reaches a customer.

What I would do this week.

Choose one service and one area you can serve well. Then make these four improvements:

1. 01 · Show the work

   Publish one useful job story.

   Use real photos, explain the problem and the work, and link to the right service page.

2. 02 · Improve the ad

   Give it a clear brief.

   Check the offer, area, claims, and destination. Test a specific homeowner concern.

3. 03 · Check the response

   Walk through an inquiry.

   Call and submit the form. Confirm who responds and what happens when the office is closed.

4. 04 · Close the loop

   Record what becomes work.

   Keep inquiries, inspections, jobs won, and payments separate. Use them to guide the next change.