Structured Data is Critical for Dealers To Get Into AI Search Results

[douglaskarr]

"44% of car shoppers have already used AI tools, and 97% say it will influence their decisions." according to Cars.com

Early AI search engines were a bit of a mess. They tried to guess context and often hallucinated or got the details wrong. But the tech has evolved. Today’s AI Overviews (AIO) and Generative Engine Optimization (GEO) rely on structured data and its validation. And the results are increasingly helpful and accurate. Schema markup has always been advantageous for SEO, but now it's a must-have as more customers are turning to AI. And they are!

When your homepage, inventory search, and VDPs are properly tagged, you aren't just indexing; you’re giving the AI a roadmap. Platforms that have leaned into this are seeing their inventory show up directly in AI-generated answers. And these aren't just random clicks; they are high-intent leads who are being served your specific vehicle because the AI actually understands your business, location, as well as your inventory.

There’s been a lot of buzz in the dealer industry lately about the Savvy Dealer AI Compatibility Tester. As with most "audits and analyzers", we were skeptical at first, so we took a deep dive. We reviewed the tool's results and saw how our CMS stacked up. If you haven't checked your site's "AI readiness" yet, it’s worth the two minutes to see where you stand. Check your score!

Full disclosure: We’re pretty proud that the Overfuel platform consistently hits a 100/100 on this analysis. It’s a testament to why our clients are seeing such strong performance right now—when you combine a fast UX with structured data, the search engines (and AI bots) love you.

 

[SavvyNick]

Appreciate the love on the AI Compatibility Tester! Savvy Dealer built it for exactly this reason, dealers deserve to know where they stand before AI search passes them by. Great to see Overfuel scoring well! The rising tide lifts all boats that are actually built for it.

 

[Greg_Gifford]

actually, it has been definitively proven that AI models do not rely on schema markup

 

[Matt Kelley]

I just ran a dealer site through and it got an F. I gave it the VDP url directly and it said it didn't have schema.

So I gave the .md report to Gemini and my URL source code to see what I am missing and part of its response was:

1. The "VDP lacks Vehicle/Product schema" claim is False​

• The Report Claims: 0/9 points for has_vehicle_schema, 0/4 for price_in_schema, and 0/3 for vin_in_schema.
• The Technical Reality: Your source code contains a highly detailed "@type": "Car" schema.
• The Proof: Your code explicitly defines the Price as $5994 (Line 70) and the VIN as [######] (Line 57). The grader is likely hard-coded to only look for the word "Vehicle" or "Product" and is ignoring the more accurate "Car" type.

 

[SavvyNick]

I just ran a dealer site through and it got an F. I gave it the VDP url directly and it said it didn't have schema.

So I gave the .md report to Gemini and my URL source code to see what I am missing and part of its response was:

1. The "VDP lacks Vehicle/Product schema" claim is False​

• The Report Claims: 0/9 points for has_vehicle_schema, 0/4 for price_in_schema, and 0/3 for vin_in_schema.
• The Technical Reality: Your source code contains a highly detailed "@type": "Car" schema.
• The Proof: Your code explicitly defines the Price as $5994 (Line 70) and the VIN as [######] (Line 57). The grader is likely hard-coded to only look for the word "Vehicle" or "Product" and is ignoring the more accurate "Car" type.

Hey Matt, I'm reaching out to our team for clarification. I'll reach out with their response.

 

[craigh]

actually, it has been definitively proven that AI models do not rely on schema markup

I would be interested on how this was definitively proven, although the operative word here may be rely.
Many LLMs retrieve data from an index and that index is created with a crawler and those crawlers use schema markup to categorize data.

In a Crawl -> Retrieve -> LLM structure, it stands to reason that the Retrieval layer is being impacted by the Crawl layer which uses schema markup. This is different than what ChatGPT may serve you in a browser, but it impacts Perplexity, Google's "AI Mode", etc. If you ask Perplexity if it uses schema markup when retrieving results from the web it confirms that it does.

I think you could argue they don't "rely" on schema markup, but I don't think it would be fair to say they don't use it.

 

[Matt Kelley]

I think in context to this conversation @Greg_Gifford is more accurate than the original post claims.

The tool is showing it is a Fact that it has to have it (even when it does have it still says it doesn't) -- This website can easily be seen and sited by Perplexity and ChatGPT just fine. And I tested 5 different non-Overfuel sites. It is true Overfuel was the only one to get 100/100 though interestingly enough, everything else failed haha

If this forum is to help dealers and provide valuable resources and accurate information, Greg is providing better information.
=====

Your Website Needs AI Optimization​

Your site is invisible to AI search engines. Let us help you fix that and get discovered by customers using ChatGPT, Perplexity, and other AI tools.
Get a free consultation on how to optimize your dealership website for AI-powered search.

 

[craigh]

Since we have access to the tools, I asked for some Gemini deep research on the subject of schema and it's relation to SEO. The net result is that there are 2 conflicting studies - 1 found no correlation between schema and AI results and the other found strong correlation between schema and AI results, specifically in newly published content.

Ultimately we will hopefully move towards a new standard and it looks like the leading contender there today is llms.txt

A proposal to standardise on using an /llms.txt file to provide information to help LLMs use a website at inference time.

 

[DjSec]

Today’s AI Overviews (AIO) and Generative Engine Optimization (GEO) rely on structured data and its validation.

Generative Engine Optimization (GEO) is what SEO's were supposed to be doing all along, they are just calling a part of SEO GEO so they can charge you again for the same services they already charged you for.

When your homepage, inventory search, and VDPs are properly tagged, you aren't just indexing;

Schema markup does not cause indexing, nor is it a direct ranking factor. Adding JSON-LD to a site will not magically make a dealership appear in AI Overviews or generative answers. If a site is not already competitive in organic search, adding schema alone will not change that.

you’re giving the AI a roadmap.

A roadmap to where?

Platforms that have leaned into this are seeing their inventory show up directly in AI-generated answers.

The claim that platforms “show up directly in AI-generated answers” because of tagging oversimplifies how modern AI retrieval works. AI Overviews pull from indexed content and authoritative sources. If you are not ranking and showing up in those sources you can add all the code you want, it will not help!

And these aren't just random clicks; they are high-intent leads

Studies show AI-generated answer click-through rates are around 1.76% to 0.61%, so if you can show up and again the schema has nothing to do with you showing up you are lucky to get a click.

There’s been a lot of buzz in the dealer industry lately about the Savvy Dealer AI Compatibility Tester.

Analyzer tools are checklist-based validators. They measure markup presence not authority, not ranking, not trust signals, and not real-world performance.

If you can achieve 100/100 by pasting a small JSON-LD block, then the score itself is not measuring competitive advantage, it’s measuring compliance with a template.

Schema is useful. Fast UX is useful. Clean structure is useful.

But positioning it as a breakthrough AI advantage is marketing ... not a new discovery.

I like the tool, it is useful the marketing is not!

Just to be clear AI Overviews don’t scrape random sites.

They are grounded in:

• Top-ranking organic results
• Google’s Knowledge Graph
• Trusted publisher sources
• Authoritative local signals

If you are not already competitive in:

• Top 10 (realistically top 5)
• Or strongly authoritative in a topic

You’re unlikely to be cited.


Schema does not override this.

 

[Rick Buffkin]

I just tried the tool with one of our websites and got the blocked message. Looks like CloudFlare is blocking the session.