Quick Summary
An AI visibility score is a 0-100 measure of how visible your dealership is inside AI answers. It combines two halves: readiness, whether AI engines can read, understand and quote your website, and presence, whether ChatGPT, Gemini, Perplexity and Google AI Overviews actually name your store when buyers ask real questions.
Presence carries most of the weight, because citations are the outcome that pays. With 25% of new-vehicle buyers already using AI tools during shopping, the score has become the fixed point dealers use to know whether that shift is working for them or against them.
What You Should Know
For GMs
- One in four new-vehicle buyers now uses AI tools while shopping. The stores AI names first are collecting those buyers, and the score tells you whether that is you.
- Read the band, not the decimals: Invisible, Named only, Sometimes cited, Usually cited, or Default answer. Band changes are real; three-point wiggles are noise.
- One of the four measured questions is a service question. A low presence score is leaking repair orders to the chains, not just leaking sales.
For Marketing Directors
- The readiness-versus-presence gap assigns the work: low readiness is website work you control, low presence with high readiness is authority and third-party work.
- Your rank tracker cannot see this channel. Pair the score with GA4 referral segments for chatgpt.com, perplexity.ai and gemini.google.com to watch the traffic side.
- Hold vendors to the same standard the score holds itself to: stated engines, stated sample size, stated confidence interval. No method, no budget.
For Dealer Principals
- The Default answer band, 82 and up, is unclaimed in most markets right now. First store to earn it becomes the recommendation every AI gives by default.
- A high score built on third-party citations instead of your own site is rented, not owned. The report tells you which one you have.
- This is a monthly instrument, not a one-time audit. The engines, your competitors and your review footprint all move; the score is how you notice.
“The first thing a GM asks us is what's a good score. Wrong first question. The right one is which half is dragging, because a readable site nobody cites and a cited store with an unreadable site need opposite work, and the total hides that.”
Ryan Boyle
Director, A3 Brands
Ask ChatGPT "best Subaru dealer near Prescott" and it gives an answer. It names stores. It leaves stores out. It does the same thing thousands of times a day, in every market, for every marque, and most dealers have no idea which side of that answer they are on.
An AI visibility score exists to end the guessing. It is one number, 0 to 100, that tells you whether the AI engines buyers now use can read your website and whether they actually recommend your store when it matters.
We built our own scoring engine at A3 Brands, the Saggy AI analyzer, and it runs about 200 live engine checks per dealership. This guide explains what the score measures, how the math works, what the bands mean, and how dealers use the number to make decisions. If you want the manual version first, our 15-minute AI search check covers what you can see with your own eyes.
What an AI visibility score actually measures
An AI visibility score answers two questions that sound similar and are not.
Can AI engines read your site? Crawlers for ChatGPT, Gemini and Perplexity have to fetch your pages, parse them, and pull clean answers out of them. A site that blocks AI crawlers, buries answers in scripts, or publishes thin template pages fails this test before any buyer asks anything. We call this half readiness.
Do AI engines actually recommend you? When a buyer asks "which Kia dealership near Topeka has the best reviews," an engine either cites your store, mentions it in passing, or leaves it out. We call this half presence, and it is measured by asking the engines real questions and reading the answers, not by inspecting your website.
Both halves matter because each one fails differently. A perfect website nobody cites is invisible. A store that gets cited today on a site engines struggle to read is one redesign away from losing the position.
The score exists to put both facts in one place, weighted toward the one that pays.
Why dealers need one now
The buyer behavior already moved. Cox Automotive's 2025 Car Buyer Journey Study, released in January 2026, found 19% of all vehicle buyers and 25% of new-vehicle buyers used AI tools or AI-generated overviews while shopping. That was the study's first year even tracking the question.
The search results moved with it. Pew Research monitored 900 US adults across 68,879 Google searches in March 2025: 60% of question-style queries produced an AI summary, and when one appears, clicks on the traditional results underneath drop from 15% to 8%.
Half the click, roughly, for the stores that only optimized for the old blue links. The buyers did not disappear. They are reading the AI's answer, and the AI's answer names somebody. We wrote about what that does to lead volume in rankings look fine but leads are down.
Here is the operational problem: you cannot manage what you cannot see. Your GA4 dashboard does not show ChatGPT conversations. Your rank tracker does not know what Perplexity said about you yesterday. An AI visibility score is the instrument that makes this channel measurable, the way a ranking report made classic SEO measurable twenty years ago.
The Channel the Score Measures
25%
New-Vehicle Buyers Using AI
Used AI tools or AI overviews while shopping (Cox Automotive Car Buyer Journey, Jan 2026)
60%
Question Queries Get AI Answers
Share of question-word Google searches that produced an AI summary (Pew Research, March 2025)
15% → 8%
Click Rate When AI Answers
Clicks on traditional results nearly halve when an AI Overview appears (Pew Research)
Readiness: the half you control
Readiness is a crawl of your own website, graded the way an AI engine experiences it. Our version scores it across the three service lines we actually work in:
SEO: findable and indexable.
Classic crawl health. Can engines fetch the pages, are they indexed, is the structure clean, do AI crawlers get through or does the firewall block them.
AEO: answers a buyer question.
Answerability. Does the page state the answer plainly enough for an engine to extract it, or does the buyer question get answered nowhere on the site? Answer-first structure, real FAQ content, and pages organized around questions score here.
GEO: worth quoting as a source.
Citability. Depth, authority signals, schema markup, named numbers, and content specific enough that quoting you is safer than paraphrasing a directory.
Readiness is the half you control completely. Every point of it is a change you can ship on your own website. That also means it is the half your SEO partner can fix directly, which is exactly why a score that measured only readiness would flatter the vendor and tell the dealer nothing about outcomes.
Presence: the half that pays
Presence is measured from the outside. The audit asks the engines the four questions a real buyer in your market asks, built from your own store's marque and city:
- ●best {marque} dealer in {your city}
- ●where should I buy a {marque} near {your city}
- ●best place to service a {marque} in {your city}
- ●which {marque} dealership near {your city} has the best reviews
Each answer gets read and classified. Cited means the engine named your store as a recommendation or source. Mentioned means you appeared but were not the answer. Absent means the buyer just got sent to somebody else, and one of those four questions is a service question, so this is your fixed ops pipeline too, not just sales.
Presence carries 55% of the total score, deliberately more than all three readiness lines combined. Readiness is a means. Citations are the result the dealer is buying. A score weighted the other way would reward tidy websites that never win a recommendation, and we have audited plenty of those.
55%
Of the Score Is Presence
Actual citations across ChatGPT, Gemini, Perplexity and Google AI Overviews outweigh all three website-readiness lines combined, because citations are the outcome a dealer is buying.
How the score is computed
The methodology matters, because a visibility number built on three casual ChatGPT prompts is a vibe, not a measurement. Here is how ours runs.
The engines.
ChatGPT, Gemini and Perplexity each get asked every question three times, because engines answer with real variance. Google AI Overviews gets asked once per question, because Google serves a cached overview and re-asking returns the same answer. In total the audit reads roughly 200 live engine responses.
The sample size.
We size the prompt set by confidence interval, not by taste. Eight prompts produced a 95% confidence interval of plus or minus 18 points, which is too sloppy to put a band label on. Twenty prompts brings it to about plus or minus 10, which is tight enough to act on.
What is excluded.
Claude is deliberately left out: it has no live web-search surface a car buyer would actually use, so including it would understate every dealer in the country.
The weighting.
SEO readiness 15%, AEO readiness 15%, GEO readiness 15%, presence 55%. One number comes out, 0 to 100, and it maps to a band.
How One Audit Runs
Crawl the Site
Fetch and grade the dealer's website for SEO, AEO and GEO readiness.
Derive the Market
Pull the store's marque and city from its own pages to build real buyer questions.
Ask the Engines
Four buyer questions across ChatGPT, Gemini, Perplexity and Google AI Overviews, about 200 live responses.
Score and Band
Cited, mentioned or absent per answer, weighted into one 0-100 score with a named band.
The five bands, from Invisible to Default answer
A number alone invites arguing. The bands say what the number means for the store.
Invisible (0-24).
AI engines cannot read the site and do not surface the store. Buyers asking AI in your market never hear your name.
Named only (25-44).
Engines know the store exists but send buyers elsewhere. You show up in passing while the recommendation goes to a competitor.
Sometimes cited (45-64).
Cited for some buyer questions, absent for others. Most commonly a store that wins the sales question and loses the service question, or wins in one engine and not the rest.
Usually cited (65-81).
A reliable source for most buyer questions. The work here is closing the specific gaps, question by question.
Default answer (82-100).
The store AI recommends first. This is the position every other band is fighting for, and in most markets right now it is sitting unclaimed.
The bands are deliberately blunt. A GM does not need decimals. A GM needs to know whether the store is the answer, an option, or absent.
The Five AI Visibility Bands
| Band | Score | What It Means for the Store |
|---|---|---|
| Invisible | 0-24 | AI engines cannot read the site and never surface the store |
| Named only | 25-44 | Engines know the store exists but send buyers elsewhere |
| Sometimes cited | 45-64 | Cited for some buyer questions, absent for others |
| Usually cited | 65-81 | A reliable source for most buyer questions |
| Default answer | 82-100 | The store AI recommends first |
What our own audit data shows
Numbers from the field beat theory. In August 2026 we ran the full Saggy audit against six franchise rooftops in six states, one week, same method, and the results explain why this measurement exists.
The spread is enormous.
Same week, same audit: scores ranged from 27 to 100. An Audi store in Maryland scored 100 and is the default answer in its market. A CDJR store in Texas scored 27, named by the engines in fewer than half of answers while buyers got sent elsewhere.
The market is more crowded than any dealer assumes.
Answering just four buyer questions, the engines cited between 121 and 214 distinct domains per market, roughly 1,300 citations read per audit.
Even the winner owns a minority.
Stores that ranked first for AI citations in their market captured only 15% to 24% of them. The rest went to Cars.com, AutoTrader, DealerRater, CarGurus and local press, third parties collecting the visibility dealers assume is theirs.
The engines disagree about the same store.
One Chevrolet store was cited in 45% of ChatGPT answers and 95% of Perplexity answers. A single-engine spot check would call that store either failing or winning, and both verdicts would be wrong, which is exactly why the score samples every engine repeatedly instead of trusting one conversation.
Six Rooftops, One Week of Audits
27-100
Score Range
Same week, same method, six franchise stores in six states
121-214
Domains Cited Per Market
Distinct domains the engines cited answering four buyer questions
≤24%
Citation Share of the Winner
Even rank-1 stores captured under a quarter; directories took the rest
Reading your score: the gap is the diagnosis
The single most useful thing in the report is not the total. It is the gap between the two halves, because the gap tells you what kind of problem you have.
Readiness high, presence low.
Your site is readable and the engines still do not cite it. The content exists; nothing yet makes it the obvious source to quote. This is usually an authority and specificity problem: the engines are quoting review platforms and directories instead of you.
The fix lives in why AI cites five signals, not in another technical audit.
Presence high, readiness low.
You are being cited despite your site, not because of it. The position rests on third-party pages, reviews, directories, press, and none of that is under your control. One platform policy change or one aggressive competitor and the citations move. This score pattern looks like winning and is actually fragile.
Both low.
The engines are not reaching for you, and the site gives them no reason to start. The order of operations matters here: fix readiness first, because presence work with an unreadable site behind it washes out.
Both high.
The work is holding the position, and the score's job becomes monitoring, because every competitor's agency is now pitching them the same channel.
What actually moves the score
The score is an instrument, not a lever. These are the levers, roughly in the order we pull them for clients.
Open the doors.
Let AI crawlers through the firewall, fix the crawl blocks, and get the basics indexed. Cheap, fast, and it caps everything else if skipped.
Build pages that answer the four questions.
A buyer asking "best place to service a Mazda in Fort Walton Beach" needs a page that actually answers that, with pricing, proof and specifics. Answer-first structure is the difference between being source material and being wallpaper. Our teardown of what 23,745 AI citations reveal shows exactly what format wins.
Get quotable.
Named numbers, real inventory facts, staff and store specifics, and schema that labels all of it. Engines quote the page that says "oil changes from $89, 45 minutes, Saturday hours" over the one that says "quality service you can trust."
Feed the third-party record.
Engines lean on reviews and local coverage when deciding who to recommend. Your review volume, ratings and Google Business Profile accuracy show up in presence even though they live off your site.
None of this is exotic. It is the same work as getting cited by ChatGPT, pointed at the questions your specific market asks, and measured instead of assumed.
What the score is not
A measurement you trust has to be honest about its edges.
It is not a lead count.
The score measures whether AI recommends you, not how many buyers acted on the recommendation. It is a leading indicator. Leads confirm it later, in your CRM.
It is not a rank tracker.
There is no position two in a ChatGPT answer. You are the recommendation, a mention, or absent, which is why the score reads in bands instead of positions.
It is not perfectly stable.
Engines answer with real variance, which is why the methodology asks each one repeatedly and reports a number with a known confidence interval of about plus or minus 10. A three-point wiggle between runs is noise. A band change is signal.
It is not permanent.
Engines update, competitors publish, review counts shift. A score is a photograph, which is exactly why we track it monthly for clients through AI visibility monitoring instead of framing one good number.
Anyone selling you a one-decimal AI score with no stated method, no sample size and no error bar is selling a screenshot.
How to check your score free
We built the scoring engine into a free tool on this site: Saggy, our AI visibility analyzer. You give it your dealership's website. It fetches the site, works out your marque and market, builds the four buyer questions, runs them across ChatGPT, Gemini, Perplexity and Google AI Overviews.
It hands back your score, your band, and the per-question breakdown of who the engines named instead of you.
It is the same roughly-200-call audit we run at the start of client engagements, and it spends real API money per run, so it is limited to one run per store per day.
What you do with the number is the actual point. If readiness is the laggard, the fix list is concrete and it is yours to ship or hand to your SEO partner. If presence is the laggard, you now know which questions and which engines you are losing, and who is winning them. Either way you have replaced "how are we doing in AI?" with a number, a band, and a named gap.
That is the whole function of an AI visibility score. Not a vanity metric. The instrument panel for a channel that one in four new-vehicle buyers is already using.
The Score Is the Start, Not the Service
A3 Brands uses the AI visibility score as the shared instrument between dealer and agency: the audit names the gaps, the SEO, AEO and GEO work closes them, and the monthly re-score proves whether it worked. Every client engagement starts and ends with the same number.
Key Takeaways
- ✓An AI visibility score is a 0-100 measure combining readiness, whether AI engines can read and quote your site, and presence, whether they actually cite your store in answers.
- ✓Presence carries 55% of the score because citations are the outcome that pays; a perfect website nobody cites is still invisible.
- ✓The score is measured by asking ChatGPT, Gemini, Perplexity and Google AI Overviews the four questions real buyers ask about your marque and city, across roughly 200 live responses.
- ✓25% of new-vehicle buyers already use AI tools while shopping per Cox Automotive, and clicks on traditional results drop from 15% to 8% when an AI Overview appears per Pew Research.
- ✓Five bands translate the number for a GM: Invisible, Named only, Sometimes cited, Usually cited, and Default answer.
- ✓The gap between readiness and presence is the diagnosis: readable-but-not-cited is an authority problem, cited-despite-the-site is a fragile position resting on third parties.
- ✓Treat any AI score with no stated methodology, sample size or confidence interval as marketing, not measurement.
- ✓In A3's own August 2026 audits of six franchise rooftops, scores ranged 27 to 100 and even rank-1 stores captured only 15-24% of their market's AI citations.
Founder & President, A3 Brands
Tim spent a decade distributing products to 3,000+ dealerships, ran the Internet Sales department at Baker Automotive Group, and served as Acura's Field Program Manager and Digital Strategist at Shift Digital before founding A3 Brands — one of very few SEO agencies built exclusively for car dealerships.
Frequently Asked Questions
What is a good AI visibility score for a dealership?
Which AI engines actually matter for car buyers?
Is an AI visibility score the same as my Google ranking?
Why does my AI visibility score change between runs?
Can I improve my AI visibility score myself?
How often should a dealership check its AI visibility score?
Sources & References
- Cox Automotive 2025 Car Buyer Journey Study (released January 2026) — 19% of all buyers and 25% of new-vehicle buyers used AI tools or AI-generated overviews while shopping; 2,300 consumers surveyed fall 2025
- Pew Research Center (March 2025) — 60% of question-word queries produced an AI summary; clicks on traditional results drop from 15% to 8% when one appears; 900 US adults, 68,879 searches, March 2025
- A3 Brands AI visibility scoring methodology — Four buyer questions across ChatGPT, Gemini, Perplexity and Google AI Overviews; ~200 live responses per audit; 95% confidence interval of ±10 points; presence weighted 55%
- A3 Brands Saggy audit data (August 2026) — Six franchise rooftops in six states; ~1,300 citations read per market across 121-214 distinct domains; rank-1 stores captured 15-24% of citations; per-engine cited rates for the same store diverged as far as 45% vs 95%
What's Your Store's AI Visibility Score?
Run the same audit we run for clients: your marque, your market, four real buyer questions across ChatGPT, Gemini, Perplexity and Google AI Overviews. Free, about 200 live checks, one run per store per day.