How to Measure Your AI Search Visibility in 2026

TL;DR: Ask the four big AI engines the same 10 customer questions every month, score each answer 2 points if your business is recommended, 1 if mentioned, 0 if absent, and track the total out of 80. Pair that with AI referral tracking in your analytics and a monthly check of the data sources AI engines read. In 2026, 45% of consumers already use AI to find local businesses, but ChatGPT recommends only about 1.2% of them. The bar is low. Measuring is how you find your gap and claim your spot.

The Short Answer

AI search visibility is simple to define: out of all the times a potential customer asks an AI engine a question your business should win, how often does the AI actually name you?

You measure it in three layers:

1. Prompt testing. Run a fixed set of real customer questions through ChatGPT, Perplexity, Gemini, and Google AI Overviews every month and score the answers. 2. Referral tracking. Count the visitors and leads that arrive at your website from AI engines. 3. Input auditing. Check the sources AI engines learn from, your Google Business Profile, reviews, schema, and citations, because those inputs predict next quarter's answers.

None of this requires enterprise software. A spreadsheet, an hour a month, and discipline will beat what most agencies call "AI monitoring." Here is the exact process we run for Mi Assist Studio clients.

Why This Number Matters in 2026

The shift happened faster than almost anyone planned for. A 2026 survey of US consumers found that 45% now use AI tools like ChatGPT, Gemini, and Perplexity to find local businesses. One year earlier that number was 6%. That is a seven-fold jump in twelve months, and among 30 to 44 year olds, the group booking med spa visits, hiring HVAC techs, and choosing family lawyers, adoption hits 64%.

Two more numbers frame the opportunity:

* ChatGPT recommends only about 1.2% of local businesses. Nearly every business in your market is invisible to the engine 45% of consumers now consult. A 2026 analysis found AI search left local businesses out of 46% of relevant answers entirely. * Google's grip on discovery is loosening. Google's share as the top review discovery platform fell from 83% to 71% in a single year, while AI Overviews now appear on roughly 48% of searches. A July 2026 analysis put organic click-through around 8% when an AI summary appears, versus about 15% without one. The clicks that remain increasingly go to businesses cited inside the AI answer, which earn about 35% more organic clicks than those left out.

The takeaway is not that Google is dead. It is that a growing slice of your future customers never see a list of ten blue links at all. They see one synthesized answer, and either your business is in it or a competitor is.

You cannot manage what you do not measure. So measure it.

Layer 1: The 10-Prompt Visibility Test

This is the core metric. It takes about an hour a month.

Step 1 - Write 10 questions your real customers ask. Not keywords. Full, conversational questions with buying intent, the way people actually talk to ChatGPT. Pull them from your call logs, review mentions, and intake forms. Examples for Florida verticals:

Vertical	Example test prompt
HVAC (Melbourne)	"My AC is blowing warm air. Who is a reliable HVAC repair company in Melbourne, Florida that can come out same day?"
Med spa (Miami)	"Which med spa in Brickell has the best reputation for natural-looking Botox results?"
Restaurant (Orlando)	"Where should I book an anniversary dinner in Winter Park with good reviews and outdoor seating?"
Law firm (Tampa)	"Who is a well-reviewed personal injury attorney in Tampa that offers free consultations?"

Step 2 - Run each prompt in four engines. ChatGPT, Perplexity, Gemini, and a normal Google search to catch the AI Overview. Use a fresh session or incognito window each time so your history does not contaminate the answers. Keep the wording identical month to month.

Step 3 - Score every answer.

* 2 points: your business is named and recommended, with a detail like your phone number, a service, or a review theme. * 1 point: your business is mentioned but not recommended, or appears in a list without differentiation. * 0 points: you are absent.

Ten prompts across four engines gives a maximum score of 80. Log the score, the date, and which competitors were named instead of you. That competitor column becomes your roadmap, because whoever the AI cites is whoever the AI trusts, and their citation sources can be reverse engineered.

Step 4 - Repeat monthly, same prompts, same engines. One test is a snapshot. Six months of tests is a trend line you can manage against, and the trend is the metric that matters. AI answers are probabilistic, so expect some month-to-month wobble. Three consecutive months of movement is signal. One month is noise.

Layer 2: Track AI Referrals in Your Analytics

Prompt testing tells you what the engines say. Referral tracking tells you what that is worth in traffic and leads.

Set up a channel group or segment in your analytics for AI referral sources. The referrer domains to capture in 2026: chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, and claude.ai. In GA4, build a custom channel with a regex match on those referrers. In a CRM with first-touch and last-touch attribution, tag them as their own source so AI-referred leads are visible in your pipeline, not lumped into a generic referral bucket.

Two practical warnings from running this for our own clients:

* A large share of AI clicks arrive with no referrer at all. Clicks out of AI mobile apps and some desktop apps show up in analytics as Direct traffic. If Direct visits to deep pages on your site start climbing while nothing else changed, AI engines are likely sending people your way without credit. Watch for Direct sessions landing on specific service or blog pages rather than your homepage. * AI engines sometimes send visitors to pages that do not exist. They occasionally construct a plausible URL on a trusted domain instead of citing a real one. Check your 404 logs monthly. A 404 URL that keeps receiving traffic is an AI engine telling you exactly what page it wants you to have. Build that page, and you convert a dead end into a landing page with pre-qualified demand.

Then connect referrals to revenue. AI-referred visitors tend to arrive pre-sold because the engine already made the recommendation. Track their lead conversion rate separately. Across client accounts we consistently see AI-referred visitors convert at a higher rate than generic organic traffic, which means even modest AI traffic can outproduce a larger volume of unqualified clicks.

Layer 3: Audit the Inputs AI Engines Read

Your visibility score is a lagging indicator. The sources AI engines learn from are the leading indicator. Once a month, check the five inputs that drive local AI answers:

Input	What to verify monthly
Google Business Profile	Categories, hours, services, and photos current. Q&A answered.
Reviews	New reviews flowing on Google plus one secondary platform. Response rate near 100%.
Website structure	Service pages with clear pricing signals, FAQ content, and LocalBusiness schema markup that validates.
NAP consistency	Name, address, phone identical across your site, GBP, Yelp, and industry directories.
Mentions	New citations, local press, chamber listings, supplier pages, or vertical directories added since last month.

Inconsistency is the quiet killer here. If your website says 6 PM and your Google Business Profile says 5 PM, an AI engine loses confidence in both and may omit you rather than risk a wrong answer. Every inconsistency you fix raises the probability the engine names you next month.

What a Good Score Looks Like

Benchmarks from the visibility tests we run:

* Untouched local business: 0 to 6 out of 80. This is where almost everyone starts, which matches the data showing ChatGPT recommends just 1.2% of local businesses. * After 90 days of focused AI SEO work: 15 to 30 out of 80, usually led by Perplexity and AI Overviews, which refresh their sources fastest. * Market leader in a niche: 40+, with at least one engine naming the business unprompted with a differentiating detail.

The honest timeline: first movement typically shows in 60 to 120 days, because AI engines need to re-crawl and re-weigh their sources. That lag is exactly why measuring monthly matters. Without a score, AI SEO feels like faith. With one, it is a trend line you manage like any other business number.

FAQ

How often should I test my AI search visibility?

Monthly. Same 10 prompts, same four engines, logged in the same spreadsheet. Weekly testing adds noise without insight, and quarterly testing lets a competitor hold the recommendation slot for three months before you notice.

Do I need a paid AI visibility tool?

Not to start. The 10-prompt test and a GA4 channel group cost nothing but an hour a month. Paid trackers become worth it when you manage multiple locations or need daily competitive monitoring across dozens of prompts.

Why does my business rank on Google but never show up in ChatGPT?

Because the inputs differ. Google rankings lean on backlinks and on-page optimization. AI engines lean on consistent structured data, review text, and mentions across sources they trust. A page-one Google business with thin reviews and inconsistent NAP data is invisible to AI engines, and the 2026 data shows that describes most local businesses.

How long does it take to improve the score?

Expect first citations in 60 to 120 days after fixing your data foundation, with compounding gains after that. Perplexity and Google AI Overviews typically move first, ChatGPT moves slower but sticks longer once you are in.

---

If you want the score without running the test yourself, we will do it for you. Mi Assist Studio runs this exact visibility audit for Florida businesses, ten real customer prompts across four AI engines, scored and benchmarked against your top three competitors, with the gap analysis that shows what to fix first. Get your free AI visibility check or read what AI SEO is and why it matters if you are starting from zero.