Local businesses should track seven AI visibility metrics in 2026: mention rate, AI share of voice, citation rate, recommendation position, sentiment and accuracy, prompt coverage, and AI-referred conversions. Together they show whether ChatGPT, Google AI Overviews, Perplexity, Gemini, and Claude name your business when a customer asks for a plumber, dentist, or accountant.
Traditional rank tracking does not capture any of this. A page-one Google ranking says nothing about whether ChatGPT recommends you when someone types "who is the best emergency HVAC company near me?" AI answers are generated fresh, vary by phrasing, and often name only three to five businesses. If you are not one of them, you are invisible at the exact moment of decision. This guide defines each metric, explains how to measure it, and gives benchmarks and a practical tracking routine.
Why Local Businesses Need New Metrics for AI Search
Search behavior has split into two channels. Customers still use Google's blue links and map pack, but a growing share now start with an AI answer: Google AI Overviews at the top of results, ChatGPT conversations, Perplexity summaries, or Gemini responses on mobile. In each case the customer receives a synthesized recommendation instead of a list of ten links to evaluate.
That changes what "visibility" means. In classic search, you measure position on a results page. In AI search, you measure whether you are included in the answer at all, how prominently, how accurately, and with what source attribution. Rankings, impressions, and click-through rate cannot answer those questions, because many AI answers produce no click and no impression in your analytics.
Three structural differences make dedicated metrics necessary:
- Answers are non-deterministic. The same prompt can return different businesses on different runs, so a single check proves nothing. You need repeated sampling.
- Answers are shortlist-based. Most responses name a handful of options. Presence or absence matters far more than a rank of 7 versus 9.
- Answers vary by platform. A business can be recommended consistently in Perplexity and absent from ChatGPT, because each system draws on different sources and retrieval methods.
The seven metrics below are built around those realities. Each one answers a specific question you can act on.
The 7 Metrics at a Glance
| Metric | Question It Answers | How It Is Measured |
|---|---|---|
| Mention rate | How often does AI name us? | Percent of tracked prompt runs that include your business |
| AI share of voice | How do we compare with competitors? | Your mentions divided by all tracked-brand mentions |
| Citation rate | Does AI link to or source our content? | Percent of answers citing your domain |
| Recommendation position | Are we listed first or last? | Average order of appearance in answers |
| Sentiment and accuracy | Is what AI says true and positive? | Scored review of descriptions, hours, services, pricing |
| Prompt coverage | Across how many customer questions do we appear? | Percent of your prompt set with at least one mention |
| AI-referred conversions | Does AI visibility produce leads? | Calls, forms, and bookings tied to AI referral sources |
1. Mention Rate
Mention rate is the percentage of tested AI responses that name your business. It is the foundational metric: if it is zero, nothing else matters.
How to measure it: Build a list of 20 to 50 prompts a real customer would ask. Run each prompt on each platform multiple times, ideally five runs per prompt, because answers vary. Count the runs where your business name appears, then divide by total runs.
For example, a dental practice tests 30 prompts across ChatGPT, Gemini, Perplexity, and Google AI Overviews, with five runs each. That is 600 responses. If the practice appears in 96 of them, its mention rate is 16%.
Segment it. A blended number hides the useful detail. Break mention rate down by platform, by prompt type, and by geography. A roofer might show a 40% mention rate for branded prompts ("is Smith Roofing reputable?") but 4% for unbranded prompts ("best roofer in Tulsa"). The unbranded number is the one that reflects new-customer acquisition.
Benchmark guidance: Most local businesses that have never optimized for AI start below 10% on unbranded prompts. Reaching 25% to 30% across your core service-and-city prompts is a strong first-year target. Category leaders in small markets often exceed 50%.
2. AI Share of Voice
AI share of voice measures your portion of all brand mentions across a defined prompt set, compared with named competitors. It converts an isolated number into a competitive position.
Formula: Your mentions divided by the total mentions of you plus every tracked competitor, multiplied by 100.
Suppose you track yourself and four competitors across 500 responses. You are mentioned 80 times, and the competitors are mentioned 200, 150, 90, and 30 times. The total is 550, so your share of voice is 80 ÷ 550, or 14.5%. You rank third of five, and the data shows exactly who dominates the answers in your market.
Why it matters more than mention rate alone: Mention rate can rise while your competitive position falls, if rivals improve faster. Share of voice catches that. It also reveals which competitor to study. If one law firm holds 38% of share of voice, examine what sources AI cites for it: directories, press mentions, review volume, or well-structured service pages.
Choose competitors carefully. Track the businesses that actually appear in AI answers for your prompts, not only the ones you consider rivals. Often a directory or aggregator such as Yelp, Angi, or Healthgrades takes a large share. Record those too, since they tell you which third-party sources AI trusts in your category.
3. Citation Rate
Citation rate is the percentage of AI answers that link to or attribute information to your website or your profiles. Perplexity, Google AI Overviews, and ChatGPT with search all display sources. Being named is good. Being cited means the AI treated your content as evidence.
Track two layers:
- Owned citations: your website, blog, and service pages appear as a source.
- Third-party citations: your Google Business Profile, Yelp, BBB, industry directories, news coverage, or review sites are the sources that mention you.
The split is diagnostic. If you are mentioned but only through third-party sources, AI does not consider your own site authoritative. That points toward content work: clear service pages, FAQ sections that answer real customer questions, structured data, and specific local detail such as neighborhoods served, licensing, and pricing ranges.
Also record which pages get cited. If AI cites your "about" page but never your service pages, those service pages are probably thin or hard to parse. A citation log by URL becomes a prioritized content to-do list.
4. Recommendation Position
Recommendation position is where your business falls in the order of an AI answer. Models tend to list the strongest-fit option first, and users act on early entries far more often than later ones.
Scoring method: Record 1 for first mention, 2 for second, and so on. Average across all runs where you appear. Also track the share of answers where you are the top recommendation, which is the figure that predicts calls.
Distinguish three answer formats, because position means different things in each:
- Ranked lists: "Here are the top five." Position is literal.
- Single recommendations: "I'd recommend X." Only first position counts.
- Narrative summaries: businesses mentioned in paragraphs. Note whether you are featured as the primary option or as an afterthought.
A med spa that appears in 30% of answers but is first in 60% of those is in a stronger position than one that appears in 40% but is first in only 10%. Position combined with mention rate gives a more accurate picture of lead potential than either alone.
5. Sentiment and Accuracy
Sentiment and accuracy measure what AI says about you, not just whether it says your name. For local businesses this is a risk metric as much as a growth metric, because AI models regularly repeat outdated or wrong information.
Audit these facts every month:
- Business name, address, phone number, and service area
- Hours of operation, including holiday and emergency hours
- Services offered, and services wrongly attributed to you
- Pricing statements and insurance or payment acceptance
- Licenses, certifications, and years in business
- Review summaries and overall characterization
Score each response on accuracy (correct, partially correct, incorrect) and tone (positive, neutral, negative). An accuracy rate below 90% on core facts signals a source problem: conflicting data across directories, an outdated Google Business Profile, or stale pages on your own site. Fix the underlying source, then re-test. AI answers usually correct themselves only after the sources they draw from are corrected.
Sentiment also reflects your review profile. If AI summarizes you as "reliable but slow to schedule," that phrase almost certainly comes from recurring review language. It tells you what to address operationally and what to emphasize in new content.
6. Prompt Coverage
Prompt coverage is the percentage of your tracked prompts where you appear at least once. Where mention rate counts frequency, coverage counts breadth. It shows whether you own one narrow query or the full range of ways customers ask.
Organize your prompt set into categories:
- Service plus location: "best plumber in Fort Collins"
- Problem-based: "my water heater is leaking, who do I call?"
- Comparison: "Is a dentist or an orthodontist better for aligners?"
- Cost and process: "how much does a roof replacement cost in Ohio?"
- Trust and credentials: "licensed accountants who handle small business taxes"
- Urgency and availability: "24-hour HVAC repair near me"
Coverage gaps are your content roadmap. If you appear for service-plus-location prompts but never for cost or problem-based prompts, you need pages that answer those questions with specific figures and clear steps. Businesses that publish detailed answers to real customer questions gain coverage in exactly the categories where they were previously missing.
Refresh the prompt set quarterly. Customer language shifts, and new question types appear as AI use grows. Keep a stable core of 15 to 20 prompts so you can compare trends over time, and rotate the remainder.
7. AI-Referred Conversions
AI-referred conversions are the calls, form submissions, and bookings that originate from AI platforms. This is the metric that connects visibility to revenue, and it is the hardest to capture because many AI-influenced customers never click a link.
Three ways to measure it:
- Referral traffic in analytics. Look for sessions from chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and claude.ai. Create a saved segment so you can trend it monthly. Google AI Overviews traffic is blended into Google organic, so it requires separate inference.
- Intake questions. Add "How did you hear about us?" to your forms, booking flow, and phone script, with "ChatGPT or another AI tool" as an explicit option. Customers frequently report AI discovery here even when analytics shows direct traffic.
- Dedicated tracking. Use a distinct tracking phone number or a UTM-tagged link on profiles and pages that AI is likely to cite, so calls and clicks can be attributed.
Expect AI-referred volume to be small in absolute terms for most local businesses today, but it tends to convert at a higher rate than generic search traffic, because the customer arrives pre-qualified by a recommendation. Track conversion rate and average job value for this segment separately.
How to Build a Tracking Routine
The metrics only help if you collect them consistently. A workable routine has four parts.
Step 1: Define your prompt set
Write 30 to 50 prompts that mirror real customer questions. Pull language from call recordings, review text, Google Search Console queries, and front-desk notes. Include your city, your neighborhoods, and your highest-margin services.
Step 2: Sample repeatedly
Run each prompt on at least four platforms: ChatGPT, Google AI Overviews, Perplexity, and Gemini. Run each five times, since a single response is unreliable. Log the full response, the date, and the platform version or mode, such as search enabled.
Step 3: Choose manual or automated tracking
| Approach | Best For | Limitations |
|---|---|---|
| Manual spot checks | Initial baseline, one-location businesses | Time-intensive, small samples, hard to repeat consistently |
| Spreadsheet logging | Monthly trend tracking with a fixed prompt set | Needs discipline; about 3 to 5 hours per round |
| Automated AI visibility tools | Multi-platform, multi-location, ongoing monitoring | Subscription cost; verify methodology and sampling depth |
Step 4: Review on a fixed cadence
Run a full measurement monthly and a lighter check weekly on your top 10 prompts. Review results against your content and review activity. If you publish a new service page on the 5th and citation rate for that topic rises by the 30th, you have evidence of cause. Keep a change log so you can connect actions to movements.
Common Tracking Mistakes
- Testing only branded prompts. Asking AI about your own name flatters you. New customers ask unbranded questions.
- Relying on one run. One answer is an anecdote. Five runs per prompt is a minimum for a usable rate.
- Logged-in personalization. Your own chat history and location can bias results. Test from a clean session or a neutral account, and set the location deliberately.
- Ignoring competitors. A 20% mention rate is excellent in one market and weak in another. Share of voice supplies context.
- Treating AI as separate from the basics. Accurate listings, strong reviews, and clear website content feed every AI platform. Weak foundations cap every metric above.
Turning Metrics into Action
Each metric points to a specific fix:
- Low mention rate: strengthen entity clarity. Make sure your name, services, and service area are consistent across your website, Google Business Profile, and major directories.
- Low share of voice: study the competitors who dominate. Identify the sources that cite them and pursue the same placements.
- Low citation rate: publish detailed, well-structured pages with direct answers, specific numbers, and schema markup.
- Weak position: build review volume and recency, and add proof points such as credentials, awards, and years of experience.
- Accuracy problems: correct the source data, then retest.
- Narrow coverage: create content for the question categories where you are absent.
- Weak conversions: make contact easy on cited pages, with visible phone numbers, forms, and online booking.
Businesses with a site built for structured content and fast edits can act on these findings quickly. Platforms such as WorkspaceCMS include a form builder on every plan and booking and scheduling on Growth and Premium, which helps capture the visitors that AI recommendations send. A solid SEO foundation on the site supports the content work, while ongoing SEO and AI campaigns remain a separate add-on.
AI Visibility Tracking FAQ
What is AI share of voice?
AI share of voice is your percentage of all brand mentions across a set of tracked prompts, compared with competitors. Divide your mentions by total mentions of all tracked businesses. A plumber mentioned 80 times out of 550 total mentions has a 14.5% share of voice.
How often should a local business check its AI visibility?
Run a full measurement monthly across 30 to 50 prompts and four platforms, with five runs per prompt. Add a weekly check of your top 10 prompts to catch sudden changes, such as a competitor gaining position or an inaccurate fact appearing.
Which AI platforms should I track?
Track ChatGPT, Google AI Overviews, Perplexity, and Gemini at minimum. Each uses different sources and retrieval methods, so results differ by platform. Add Claude and Microsoft Copilot if your customers skew toward those tools or your tracking tool supports them.
Can I track AI visibility for free?
Yes. Write a prompt list, run each prompt several times per platform, and log results in a spreadsheet. Expect 3 to 5 hours per monthly round. Paid tools automate sampling and scale better for multiple locations, but manual tracking is enough to establish a baseline.
How do I know if AI is sending me customers?
Check analytics for referral sessions from chatgpt.com, perplexity.ai, gemini.google.com, and similar domains. Add an AI option to your "How did you hear about us?" question on forms and phone intake. Many AI-influenced customers appear as direct traffic, so self-reported attribution is essential.