Measurement & Reporting · 9 min read · July 15, 2026

How Many Roofing Enquiries Come From AI Recommendations? What the Numbers Actually Show

More homeowners than ever are asking ChatGPT, Gemini, or Perplexity to recommend a roofer before they ever type a query into Google. Yet almost no roofing business tracks how many of its actual renovation enquiries trace back to that shift. This guide shows you how to make AI-driven recommendations measurable in the roofing trade, what numbers are realistic today, and where the common measurement mistakes happen.

Why the AI recommendation suddenly matters in the roofing trade

Not long ago, nearly every roof renovation followed the same path: the homeowner typed "roofer near me" into Google, clicked one of the top results, and collected two or three quotes. Now a new step comes first. More and more people start by asking ChatGPT or Perplexity questions like "What does a new roof cost for a single-family home?" or "How do I find a reputable roofer?" The AI answers, and it sometimes names specific businesses, criteria, or regions.

For you as a roofer, that means part of your future customer base is building its shortlist inside a chat window you never see. If the AI recommends your competitors there, or simply doesn't know your business exists, you lose the enquiry before it's ever made. That's why the question matters so much: how many renovation enquiries are already coming through this AI recommendation , and how do you actually measure it?

Nobody in the roofing trade can hand you an exact percentage today. But you can build a solid picture of your own numbers. It just takes the right tracking in place instead of a guess.

What GEO actually means for a roofing business

GEO stands for Generative Engine Optimization. The name sounds technical, but the idea is simple: you make sure generative AI systems know your business, classify it correctly, and recommend it at the right moment. It's the successor to classic SEO, except the stage is no longer a Google results page but the answer inside a chat. For a roofer, good GEO means that when someone asks "Roof renovation on an older house: tiles or metal?", your name or your region shows up in a sensible answer.

The difference from Google is significant. Google gives you ten blue links and lets the user choose. An AI assistant usually gives one summarised answer with one to three named options. If you're not one of them, you effectively don't exist to that person. Visibility shifts from a ranking to a yes-or-no. That makes a single mention worth far more than a sixth-place spot on Google.

For roofers, there's an added layer: the work is strongly local and strongly trust-based. Nobody buys a roof off a shelf. Here, AI tends to recommend criteria and types of business rather than anonymous products. That favours well-run specialist firms, provided they lay their facts out clearly on the web.

The questions homeowners are really asking the AI

Before you can measure anything, you need to know what people are actually asking. Conversations with roofing businesses and common search patterns point to a few clear themes. Cost questions dominate: "What does a roof renovation with insulation cost for 120 square metres?" or "Is over-rafter insulation worth it compared with between-rafter insulation?" Anyone asking questions like these is close to making a real investment, which makes them exactly your target customer.

Alongside those are trust and selection questions: "How do I recognise a reputable roofer?", "What should I ask before hiring one?", "How long does a full re-roof take?" And then the technical decision questions: tile versus concrete roof tile, adding solar panels at the same time, repair versus full replacement after storm damage.

The point is that for nearly all of these questions, the AI can bring a regional specialist into the conversation, or name criteria that describe you exactly. Once you know which questions your customers are asking, you can test those same questions yourself and see whether, and how, your business shows up in the answers.

How to measure the AI recommendation: four practical methods

The first method is asking your customers directly. Add one line to your enquiry form and your phone script: "How did you hear about us?" with an extra option for "Through ChatGPT, Gemini or another AI." It sounds simple, but it's the most honest source you have. Many businesses discover this way, for the first time, that AI recommendations are already sending them enquiries.

The second method is testing it yourself. Ask the AI systems the questions typical of your region, for example "Good roofer for a roof renovation near Rosenheim." Note whether your business comes up, which competitors get named, and what the AI says about you. Repeat this monthly and you'll see a trend instead of a single snapshot.

The third method is your web server logs. Analytics tools now surface AI crawlers and referral sources such as chatgpt.com or perplexity.ai. A look at your logs or your analytics dashboard will tell you whether visitors coming from AI answers are reaching your site at all. The fourth method is a dedicated monitoring tool that automatically tracks how often, and in what context, your business is named in AI answers.

What numbers are realistic in the roofing trade right now

Now for the honest part, because overstating this doesn't help you. At most regional roofing businesses, the share of enquiries that arrive directly and consciously through an AI recommendation is still small today, a low single-digit share of first contacts for most firms. That sounds modest. But it was close to zero just two years ago.

What matters more than the current number is the direction it's moving. Use of generative AI is growing fast: ChatGPT alone reached 900 million weekly users by early 2026, Google's Gemini app passed a billion monthly users soon after, and Google's own AI Overviews now reach more than two billion people a month. For a decision as large as a roof renovation, more and more homeowners are doing this kind of research before they ever pick up the phone. On top of that there's a hidden layer: many customers tell you on the phone "I found you online", even though an AI answer was the actual trigger. So the real influence is larger than what gets recorded directly.

Run the numbers for your own business. If you get 300 enquiries a year and even a modest share of them are AI-driven, that can easily be a dozen or more enquiries you'd otherwise never trace. With an average renovation in the five-figure range and a decent close rate, that's an order volume worth paying attention to.

The common measurement mistakes, and how to avoid them

The most common mistake is all-or-nothing attribution. A customer says "Google" was the source, so you log it as a Google enquiry. But they may well have asked ChatGPT first, remembered your name, and then searched for you directly. The AI was the first nudge, and it never shows up in your numbers at all. That's why the open-ended question "What first brought you to us?" tells you more than a fixed dropdown ever will.

The second mistake is the one-off test. You ask ChatGPT once, your business doesn't come up, and you conclude AI brings you nothing. But AI answers shift depending on the exact wording, the location context, and the model version. A single query tells you almost nothing. Only repeated tests over several weeks, with varied phrasing, give you a reliable picture.

The third mistake is looking only at clicks. Unlike Google, an AI answer often needs no click at all: the answer in the chat is enough on its own. Pew Research found that when an AI summary appears in search results, people click through to a traditional result in only about 8% of visits, compared with 15% when no summary appears. The customer remembers your name and calls you directly. If you judge AI's impact only by website clicks, you'll badly underestimate it. Measure the mention itself, not just the traffic it sends.

What you can do to get recommended by AI more often

Generative AI builds its answers from what's already said about you on the web. The first lever is clean, consistent facts: the same business name, the same address, the same phone number, and the same list of services on your website, in trade directories, on your Google Business Profile, and on review sites. Contradictions make the AI less confident, and an uncertain system would rather recommend a competitor with a clean, consistent data trail.

The second lever is real content that answers the exact questions your customers have. A guide on your site titled "Between-rafter or over-rafter insulation: which suits my roof?" or "Storm damage to your roof: what to do first" gives the AI something concrete to draw on. Write the way your customers actually ask questions, with clear answers, concrete numbers, and regional detail. That's exactly the kind of passage AI systems tend to quote.

The third lever is reviews and third-party mentions. When customers recommend you in forums, on review platforms, or on local directories, that reinforces your name in the material AI systems train on and search. An Ahrefs analysis of roughly 75,000 brands found that how often a brand gets mentioned across the web correlates with AI citation rate far more strongly than backlinks do. Ask satisfied customers for an honest review. It helps with Google, and increasingly with AI recommendations too.

A realistic three-month roadmap

Start small and concrete. Month one: add the "how did you hear about us" question, including the AI option, to your enquiry form and phone script, and put together a simple list of ten questions your customers typically ask about your region. Test those questions once with ChatGPT and Perplexity and note whether, and how, you show up. That's your baseline measurement, your starting point.

Month two: clean up your facts. Check your name, address, and services across every platform and remove any contradictions. Write two guide articles on your most common customer questions, for example the cost and process of a roof renovation. At the same time, ask five satisfied customers for a review. This work pays off for Google and for AI recommendations at once.

Month three: repeat your tests and compare them against your baseline. Are you showing up more often now? Are customers naming AI as a source more often on the phone? This is how a hunch turns into a real, repeatable measurement, and how you start making decisions about your visibility where your customers will actually look first, instead of only after the enquiry has already gone to a competitor.

Common questions

How can I tell whether a renovation enquiry came from an AI recommendation?

The most reliable way is asking directly at first contact: "What first brought you to us?" with an AI option included. Your web server logs can also show referral sources such as chatgpt.com or perplexity.ai. But expect a hidden gap, since many customers still say "Google" or "the internet" as the source even after an AI answer was the real trigger.

Is GEO even worth it yet for a small regional roofing business?

Yes, precisely because the competition is still thin. Most roofing businesses haven't bothered to lay their facts out cleanly yet. The effort involved is manageable: consistent business data, a handful of real guide articles answering customer questions, and active reviews. This pays off for Google and for AI recommendations at the same time, so there's no downside.

Why does ChatGPT name my business today but not tomorrow?

AI answers shift depending on wording, location context, and model version, so a single query is never a reliable signal. An Ahrefs study found that only around 6-8% of the URLs ChatGPT cites even overlap with Google's top 10 results for the same query, which shows how differently these systems select sources. Test the same questions repeatedly over several weeks, with slightly different phrasing. That pattern tells you whether your business is mentioned consistently or just came up by chance. Consistent facts across the web raise your odds noticeably.

Read on

Content & Answer Pages