Share of Model: How to Track Your AI Search Visibility
You rewrote your service pages. You added specific numbers and FAQ sections. You cleaned up your structured data. Now what? Most home service businesses doing GEO work have no idea whether any of it is landing. They are optimizing blind, with no way to tell if ChatGPT is mentioning them, how often Perplexity cites their site, or whether a homeowner asking Google’s AI Overview for an HVAC company in their city ever sees their name.
This is the measurement problem that GEO has had since the beginning. Unlike SEO, where Google Search Console shows you impressions, clicks, and ranking positions, there is no official dashboard for AI visibility. The search engines do not publish data on how often they recommend your business in AI-generated answers. You have to go find it yourself. Here’s how.
The Metric That Matters: Share of Model
The primary GEO measurement concept is Share of Model (SoM): how often your business appears in AI-generated responses for relevant queries, compared to your competitors. It is the AI equivalent of market share in search results.
A simple way to think about it: if you run 20 relevant prompts across ChatGPT and Perplexity and your business name appears in 8 of those responses, your SoM for that prompt set is 40%. If your top competitor appears in 14 of those same 20 prompts, their SoM is 70%. The gap tells you how much ground you need to close and where.
SoM is not a perfect metric. AI responses are non-deterministic, meaning the same prompt asked twice can produce different answers. That is why you need a consistent method and a meaningful sample size to get a signal worth acting on.
Building Your Prompt Set
The first step is defining a set of 15 to 25 prompts that reflect how homeowners in your market actually look for your services. These fall into four categories:
- Direct recommendation queries: “Who is the best [trade] company in [city]?” or “What HVAC company should I call in Nashville?” These are the queries where appearing means you are being actively recommended.
- Problem-first queries: “My AC stopped working in Dallas, who should I call?” or “Emergency plumber in Phoenix at night.” These reflect the way most homeowners in a crisis actually phrase their problem to an AI assistant.
- Comparison queries: “What are the top-rated roofing companies in [city]?” or “Compare HVAC companies near me in Atlanta.” These prompt the AI to generate lists, which is where strong brands get cited by name.
- Specific service queries: “Who does water heater replacement in [city]?” or “Best drain cleaning company in [city].” These test whether the AI associates your business with specific services, not just your trade in general.
Run every prompt on ChatGPT (GPT-4o), Perplexity, and Google AI Overviews. Record whether your business name appears, what position in the list if it does, which competitors are mentioned alongside you, and whether any of your URLs are cited as sources. That raw data is your baseline.
The Manual Tracking Method (Free)
A spreadsheet and 30 minutes per month is enough to get started without any paid tools. Set up columns for: prompt text, platform (ChatGPT / Perplexity / Google AIO), date, business mentioned (yes/no), position in list if mentioned (1st, 2nd, 3rd, etc.), competitors mentioned, and source URLs cited.
Run your full prompt set once at the start of each month. Record results for your business and your top two to three competitors. After three months, you have a trend line: are you appearing more or less often, moving up or down in list position, and gaining or losing ground against competitors?
The three-month minimum matters because AI models update their underlying knowledge at irregular intervals, and a single month of data can reflect a snapshot rather than a trend. Month-over-month changes of less than 10 percentage points are usually noise. A 15 to 20 point shift in either direction is a real signal.
Paid Tools Worth Knowing
Several platforms now automate the manual tracking process. They run your prompt sets across multiple AI engines on a schedule, track changes over time, and compare your visibility against competitors.
| Tool | Best For | Starting Price (2026) |
|---|---|---|
| Otterly.AI | Automated prompt monitoring across ChatGPT, Perplexity, Gemini | ~$99/month |
| LLMrefs | Citation tracking and competitor benchmarking | ~$79/month |
| Profound | Citation source analysis and content gap identification | ~$149/month |
| HubSpot AEO | Combined SEO + AI visibility dashboard | Included in Marketing Hub |
For a home service business doing its own GEO work, the manual method is sufficient for the first six months. The paid tools become worth it when you need to track more than 30 prompts, monitor multiple service areas, or report results to stakeholders who need a dashboard rather than a spreadsheet.
What Good Looks Like
Current benchmarks for home service businesses with active GEO programs:
| Metric | Starting Out (0–3 months) | Established (6+ months) |
|---|---|---|
| Mention rate across prompt set | 5–15% | 30–50% |
| List position when mentioned | 3rd–5th | 1st–3rd |
| Citation rate (URL appears as source) | 0–5% | 10–25% |
| Competitor gap (SoM vs top competitor) | 20–40 points behind | Within 15 points |
These ranges reflect the broader market: a 2026 analysis found that 44% of AI prompts return zero brand mentions. Most businesses are invisible. A mention rate above 25% puts you in the top tier of local home service visibility in AI-generated answers.
Reading the Results: What Different Patterns Mean
Your prompt data will show patterns that tell you specifically what to fix, not just that something is wrong.
You appear in comparison queries but not recommendation queries. The AI knows you exist but does not have enough authority signals to actively recommend you. Focus on review volume on Google and Yelp, which AI engines use as credibility proxies, and on external citations: mentions in local news, home service directories, and neighborhood-level community sites.
You appear in ChatGPT but not Perplexity. Perplexity leans heavily on Reddit, forums, and real-time web sources. If you have no mentions in Reddit threads or local community forums, Perplexity has no reason to include you. Search Reddit for your city plus your trade, find relevant threads, and make sure your business information is accurate and consistent in the responses others have given.
You appear in early list positions but are not cited with a URL. The AI knows your name but is not pulling content from your website. Your site’s content may not be structured for extraction: add or improve your FAQ sections with self-contained answers, add specific pricing ranges, response times, and license numbers to your service pages.
You appear inconsistently: present in some runs, absent in others. This is normal below a certain authority threshold. The AI is on the fence about including you. This is the stage where improving the specificity of your content and adding more external mentions has the fastest impact.
The One Number to Watch
If you do nothing else, track your mention rate for the single prompt that matters most to your business. For an HVAC company in Dallas: “What HVAC company should I call in Dallas?” Run that one prompt on all three platforms, once a month, and track the yes/no result. That is one data point, three responses, six minutes of work.
After six months you will have a clear direction of travel: up, flat, or down. Up means your content and authority work is landing. Flat means something you are doing is not reaching the AI’s knowledge base. Down means a competitor is pulling ahead. All three outcomes tell you something actionable, and none of them require a paid tool or an agency to interpret. That is the starting point for measuring GEO: not a dashboard, just a question you ask every month and a record of what the AI says back.