Measuring Your GEO Results: Track Which AI Tools Are Citing Your Business

Most home service contractors doing GEO work are flying blind. They spend time building citations, updating their Google Business Profile, and adding schema markup, but they have no way to see if any of it is generating calls. Standard Google Analytics does not capture most AI search traffic. Perplexity clicks show up as referral or direct traffic with no label. ChatGPT has no referral data at all because it does not send users to websites the same way search engines do. Without a specific setup to catch this traffic, you are making changes and hoping they work.

In 2026, AI search handles 22 percent of all searches, up from 15 percent a year ago. The traffic it sends converts at 14.2 percent, compared to 2.8 percent for standard Google organic results. That gap means a single AI-referred lead is worth roughly five standard organic clicks. If you are doing GEO work and not tracking it, you are both missing the wins that should reinforce what you are doing and missing the gaps that should tell you where to focus next.

Why Standard Analytics Miss Most AI Search Traffic

AI tools handle referral attribution in completely different ways, and most of them are opaque by default:

• Perplexity sends traffic with a referrer header of perplexity.ai, so it shows up in GA4 if you filter for it. Most contractors have never set up that filter.
• ChatGPT often does not send users to websites at all. It summarizes information and either mentions your business name or it does not. There is no click to track.
• Google AI Overviews can send traffic that is counted inside your existing organic totals, with no separate label, making it impossible to separate from standard SEO results.
• Microsoft Copilot sends referral traffic tagged as bing.com, which most contractors already track but do not separate from regular Bing search.

The result is that a meaningful share of your leads may already be coming from AI referrals and showing up as direct traffic or buried inside organic totals. You need a deliberate setup to surface them.

Method 1: Manual Prompt Testing (30 Minutes Per Week)

The fastest and cheapest way to track your GEO performance is to run a set of test prompts manually in ChatGPT, Perplexity, Google AI Mode, and Copilot each week. This takes about 30 minutes and catches major changes in whether you are being cited.

Set up a simple spreadsheet with these columns: Date, AI Tool, Prompt, Result (cited or not cited), Source URL cited if any. Run the same five prompts every week and log the results. Use prompts that match how homeowners actually ask for help:

• "Who does emergency [trade] repair in [your city]?"
• "Best [trade] contractor near [city or neighborhood]?"
• "How much does [your main service] cost in [city]?"
• "Is [your business name] a good [trade] company?"
• "Which [trade] companies are licensed and insured in [your area]?"

The last prompt is particularly useful because it tests whether AI tools are pulling your license and certification data. They do this when that data is properly structured in your schema markup and GBP profile. If that prompt does not return your business after you have added license information, the issue is either that AI tools have not re-indexed the source yet or that the data is in the wrong format.

Track your results by week. If you were cited in 2 out of 5 prompts in week 1 and 4 out of 5 in week 6, something changed in your favor. Cross-reference that with any GEO changes you made during that period. That is your evidence that a specific action worked.

Method 2: GA4 Referral Filters for AI Traffic

Set up a custom channel group in GA4 to catch AI referral traffic that you can currently identify. In GA4, go to Admin, then Data Display, then Channel Groups. Create a new channel and add referral rules for these domains:

• perplexity.ai
• chat.openai.com
• copilot.microsoft.com
• you.com
• phind.com

Label this channel "AI Referral" and monitor it monthly. The volume will be low at first, especially if your GEO work is in early stages. But even one or two sessions per month from Perplexity is a signal that AI tools are surfacing your website as a source worth citing. That gives you a baseline to measure against.

For call-based businesses, GA4 referral data only tells you about web sessions, not phone calls. Most contractor leads come from calls, not form fills. This is where GA4 attribution misses the most revenue, which is why you need the next method alongside it.

Method 3: Call Attribution Questions

The simplest attribution method for phone-first businesses costs nothing. Add two questions to every intake call, either asked by your call handler or added to a booking form:

1. "How did you hear about us?"
2. "Did you use any AI tools, like ChatGPT or Perplexity, before you called?"

Log the answers in whatever CRM or spreadsheet you use to track leads. After 90 days, count how many callers mentioned AI tools. Even without perfect attribution, this gives you a directional read on whether homeowners in your market are using AI search before they call and whether your business is the one they find.

In markets where GEO has been prioritized for three to six months, it is common to see 8 to 15 percent of callers mention AI tools unprompted when asked. If that number is near zero after 90 days of GEO work, one of three things is true: the optimization has not reached threshold yet, the market is not yet using AI search heavily for your trade, or the signals you have built are not strong enough for AI tools to pull reliably.

Method 4: Paid AI Visibility Tools

If you want daily automated tracking rather than manual weekly checks, several tools now monitor your AI citation frequency across multiple platforms simultaneously:

For a single-location contractor, Mentionable or LLMrefs at the lower price point covers the main platforms and tracks your query set automatically. You get a weekly report showing which prompts returned your business name, which AI tool cited you, and what source URL was used. That last detail is actionable: if Perplexity is consistently pulling from your Yelp profile rather than your website, you know your website content needs strengthening for that query type.

The Metric That Matters: Share of Model

The GEO metric equivalent to Google's "search impression share" is called Share of Model (SoM). It measures how often your business appears in AI responses compared to your top competitors for the same set of prompts. If you run five prompts across four AI tools, that is 20 total queries. If your business appears in 8 of them, your SoM is 40 percent for that prompt set.

Track SoM monthly. A SoM that grows from 15 percent to 35 percent over three months tells you your GEO work is generating real visibility gains. Flat SoM across three months while you are actively building citations tells you the changes you are making are not moving the needle and you need to try something different. Every paid tool listed above reports this metric automatically. If you are doing manual tracking, calculate it yourself from your weekly spreadsheet.

Three Actions This Week

1. Run the five prompts above right now in ChatGPT and Perplexity. Log your results in a spreadsheet: cited or not cited, and if cited, what source URL the AI used. Do this same check every Monday for the next six weeks. After six weeks you will have enough data to see directional change from any GEO work you do between now and then. This takes 15 minutes the first time and 10 minutes every week after.
2. Set up the GA4 referral channel for AI domains. Add perplexity.ai, chat.openai.com, and copilot.microsoft.com as referral sources in a custom "AI Referral" channel group in GA4. Check it in 30 days and record the session count. Even one session per month from an AI source is the beginning of a measurable trend, and a trend you can point to when deciding whether to invest more time in GEO.
3. Add the two attribution questions to your next 20 intake calls. Ask every caller this week how they found you and whether they used an AI tool before calling. Log the answers. By week four you will have your first real data on whether homeowners in your market are using AI search before they call, and whether your business is the name they find there when they do.

GEO work without measurement is the same mistake as running Google Ads without conversion tracking: you spend time and effort and have no way to know if it worked. These three methods cost nothing except 30 minutes per week, and they give you the baseline data that makes every future optimization decision faster and more defensible. Start with the manual prompt test today. Everything else builds on top of it.