By Dealer AI Guy Editorial Staff.
TL;DR
- Inventory depth can help when a shopper asks for a specific vehicle that must be in stock. That does not make inventory volume a general AI visibility strategy.
- Zach Billings' podcast example tested inventory-oriented prompts. It did not establish that a large lot makes a dealership more trustworthy, more locally relevant, or more likely to be recommended across every kind of AI search.
- Reviews, corroborating third-party information, useful local content, crawl access, and clear dealership facts all help machines decide whether they can confidently mention a store.
- A dealer with 80 vehicles can beat a dealer with 800 for recommendation, reputation, service, and expertise prompts if the smaller store gives AI systems better evidence.
- Dealers should measure visibility by prompt category, not chase a single made-up "AI ranking."
Zach Billings made a claim on the Car Dealership Guy podcast that deserves a closer look. While discussing dealership visibility in AI search, he said "more inventory is a huge lever" and argued that dealer groups can sometimes "outweigh a single point dealership" because they have so much inventory.
That sounds plausible. It is also much broader than the evidence he described.
Inventory matters when the question is about inventory. If somebody asks an AI assistant to find a new Ford F-150 under $50,000 within 25 miles, a dealership with more matching trucks has more chances to qualify. Nobody should find that surprising.
But that does not mean a bigger lot is more visible to AI in general. It means product availability matters for a product-availability query. Turning that observation into a general rule about AI visibility confuses eligibility with recommendation.
One gets you into the candidate set. The other determines whether the system has enough confidence to put your name in the answer.
What Billings' test actually measured
In the podcast episode, Billings described an ongoing Wikimotive study using prompts such as finding F-150s for sale under a certain price. His team looked at factors including website platform, inventory on the page, page content, schema, and llms.txt.
He said the strongest predictor was quality content on the page. Inventory was second. The technical items, including schema and llms.txt, were near the bottom.
That is an interesting result. It is not evidence that inventory volume controls all AI visibility.
The prompt itself requires available inventory. A store with no matching F-150 cannot be a useful answer, no matter how good its content is. Once matching inventory exists, a larger selection may improve the odds that the store satisfies the shopper's exact constraints.
Change the prompt and the relevant evidence changes with it:
- "Who has a red 2026 F-150 under $50,000 near me?" depends heavily on live inventory.
- "Which Ford dealer near me has the best reputation?" depends much more on reviews, local prominence, and corroboration.
- "Where should I service my diesel truck?" depends on service capabilities, expertise, location, hours, and customer experience.
- "Which dealer is transparent about fees?" depends on policies, reviews, third-party mentions, and evidence of actual pricing behavior.
There is no honest way to collapse those four questions into one inventory-based ranking factor.
Inventory can make you eligible. It cannot make you credible.
A large inventory creates more vehicle pages and more possible matches. It does not tell an AI system whether the dealership is honest, responsive, nearby, knowledgeable, or good to do business with.
AI assistants are answer systems. When they recommend a business, they need enough evidence to explain why that business belongs in the answer. Lot size is weak evidence for most of the questions that come before a purchase.
Consider two stores.
Dealer A has 800 vehicles, thin vehicle pages, inconsistent hours across the web, generic city pages, vague service information, and reviews that repeatedly mention surprise fees.
Dealer B has 80 vehicles, accurate business information, detailed service pages, model-specific buying guidance, clear policies, strong recent reviews, and consistent third-party mentions.
Dealer A may have an advantage when the buyer asks for a rare trim that Dealer B does not stock. Dealer B can still be the better answer to "Which dealership should I trust?", "Where should I service this vehicle?", or "Who is best for a first-time buyer?"
That distinction is the whole issue. Inventory breadth answers "Do they have one?" It does not answer "Should I choose them?"
Reviews are evidence, but they are not magic either
Billings was on firmer ground when he discussed reviews and consensus. A recommendation engine needs corroboration. A dealership cannot establish its own reputation by publishing a page that says it is the best dealer in town.
Google says its local results are based mainly on relevance, distance, and prominence. Its own guidance also says that more reviews and positive ratings can help local ranking. That is a much more defensible explanation for why one dealership appears in a local recommendation than raw vehicle count. Google explains those local ranking factors here.
Reviews still need context. Volume, recency, rating, platform diversity, and the language customers use can all affect the picture. Five hundred reviews about oil changes do not automatically prove that a store is the best place to buy a used EV. A useful audit looks at what the reviews actually establish, not just the number beside the stars.
Content matters because AI needs something it can use
The most revealing part of Billings' own summary was that quality content ranked ahead of inventory in the inventory study.
That makes sense. A live inventory feed tells a system what is present. Good content explains the details around the decision: pricing policies, model differences, warranty coverage, service capability, trade-in process, financing limits, dealership geography, and who the store is best equipped to help.
Most dealership content fails here because it was written to occupy a URL, not answer a question. Swap the city and model name, and the same paragraph could sit on 300 dealer websites. That gives an answer engine very little worth quoting.
Useful content is specific enough to verify. It names the actual policy, process, capability, inventory constraint, or local consideration. It answers the follow-up question a shopper will ask next. It also stays consistent with the dealership's Business Profile, inventory, review footprint, and third-party listings.
Google's current guidance for appearing in its AI search features is refreshingly unromantic: follow the same SEO fundamentals, make pages crawlable, provide helpful and reliable content, and ensure structured data matches what visitors can see. Google says useful, non-commodity content will likely matter more over time than the usual bag of AI-search tricks. Read Google's guidance for generative AI search.
That does not mean the technical layer is irrelevant. It means the technical layer supports the evidence. It does not replace it.
Schema and llms.txt are not a substitute for substance
Dealer marketing loves a checkbox because a checkbox is easy to sell.
Add schema. Upload llms.txt. Install another script. Announce that the site is now "AI ready."
None of those steps can rescue weak facts, stale inventory, contradictory business information, or generic copy.
Structured data can help a machine identify entities and attributes on a page. It should be accurate and match the visible content. Crawler access also matters. OpenAI, for example, documents separate controls for OAI-SearchBot, GPTBot, and ChatGPT-User, so a careless robots.txt rule can affect how OpenAI products discover or use a site. OpenAI documents those crawlers here.
Those are table stakes. They are not proof that a dealership deserves a recommendation.
The same caution applies to llms.txt. Google now says plainly that it ignores llms.txt and that the file neither helps nor hurts visibility in Google Search. Other systems may choose to use it, but it is not a universal ranking factor, and no dealer should be told that adding one will suddenly make ChatGPT recommend the store. Even Billings' own account placed it near the bottom of the factors his team studied.
There is no single AI ranking
Dealers are being sold "AI rank tracking" as if ChatGPT had one stable results page with ten blue links.
It does not work that way.
Answers vary by platform, model, location, prompt wording, available web results, and whether the system searches the live web. Google AI Mode, ChatGPT, Perplexity, Gemini, and Copilot do not all retrieve, weigh, or cite information in the same way.
Academic work on generative engine optimization supports the narrower point that source presentation can affect visibility in generated answers. The original GEO research paper found that changes such as adding citations, quotations, and statistics could improve visibility in the systems the researchers tested. It did not establish "inventory count" as a universal lever, and research results from a controlled benchmark should not be presented as a permanent recipe for commercial AI products.
A serious dealership audit therefore tests prompt groups, not one vanity score:
- Inventory prompts: specific year, make, model, trim, price, condition, and distance.
- Recommendation prompts: best dealer, most trusted dealer, first-time buyer, luxury experience, transparent pricing.
- Fixed-ops prompts: service, tires, collision, recalls, EV repair, diesel expertise, pickup and delivery.
- Reputation prompts: reviews, fees, responsiveness, warranty complaints, trade-in experience.
- Comparison prompts: dealer versus dealer, local group versus independent store, franchise versus used-car specialist.
Then record whether the dealership appears, how it is described, which sources are cited, and whether the facts are correct. That produces something a GM can act on.
What dealers should do instead of buying the inventory myth
Start by separating availability from authority.
For availability, keep inventory pages crawlable, current, fast, internally linked, and explicit about price and vehicle attributes. Make sure important details exist in the rendered page, not only behind scripts or forms.
For authority, build evidence around the questions that inventory cannot answer:
- Keep name, address, phone, hours, departments, and service areas consistent.
- Earn reviews on the platforms buyers and search systems actually encounter.
- Publish useful pages about real policies, services, models, ownership questions, and local conditions.
- Explain fees, warranties, delivery options, return or exchange policies, and service capabilities plainly.
- Use structured data that matches visible facts.
- Check robots.txt and other crawler controls so search systems can reach the pages you expect them to use.
- Track visibility across several prompt categories and platforms, then fix the evidence gaps behind the misses.
Big groups should do this too. Scale can create more pages, more reviews, and more mentions, but it also creates more inconsistency. The advantage belongs to the dealer whose information is easiest to find, verify, and use.
The bottom line
Zach Billings did not prove that inventory size determines AI visibility. He described an ongoing test in which inventory was useful for inventory-specific prompts, while quality content was an even stronger predictor.
That is a much narrower and more reasonable conclusion.
A large lot may help a dealership answer "Who has the vehicle?" It does not automatically answer "Who should I trust?", "Who is best?", or "Where should I go?"
The dealerships that show up consistently will be the ones that give AI systems enough accurate, accessible, corroborated information to make a recommendation they can defend. Sometimes inventory will be part of that answer. Often it will not.
Frequently asked questions
Does having more inventory help a dealership appear in AI search?
It can help for prompts that require a specific vehicle, trim, price, or condition. More matching inventory creates more chances to satisfy the request. It is much less relevant to prompts about reputation, service quality, transparency, or trust.
What influences whether ChatGPT recommends a car dealership?
There is no published universal ranking formula. In practice, useful signals can include crawlable dealership information, matching inventory, local relevance, reviews, consistent business facts, useful content, and corroborating third-party sources. The mix changes with the question and the platform.
Will schema or llms.txt make ChatGPT recommend my dealership?
No single file or markup type guarantees a recommendation. Accurate schema can clarify page information, and crawler controls can affect access. llms.txt remains an emerging convention, not a universal ranking switch.
How should a dealer measure AI visibility?
Test a repeatable set of inventory, recommendation, fixed-ops, reputation, and comparison prompts across the major answer platforms. Record mentions, descriptions, citations, factual errors, and changes over time.
Check what AI actually says about your dealership
Run the free Dealer AEO Audit to test whether AI assistants can find, understand, and recommend your store. It is a better starting point than guessing how many vehicles you need to park on the lot.