AI Won't Recommend a Commodity Dealership. Here's the Six-Point Test We Run.

What commodity means to an AI engine

Google said the quiet part in May 2025. In its guidance on succeeding in AI search, Search Central wrote that publishers should focus on making unique, non-commodity content that visitors will find helpful.

Its AI features guide goes further, explaining that the AI systems look at a variety of sources, so a unique viewpoint that stands out is what gets used.

Here is what that means mechanically. When a buyer asks an engine which Honda dealer to trust near Naperville, the engine assembles an answer from sources it can quote. A page that says what every page says adds nothing. The engine already knows dealerships have factory-trained technicians. It needs a page that tells it something it does not know.

A commodity dealership, to an engine, is any store whose pages could belong to any store. Template copy, OEM boilerplate, stock photos, no prices, no names, no numbers. The engine reads a thousand of those and treats them as one, then reaches for the source that is not interchangeable.

That is why the store with a real service price list gets named and the store with a beautiful template does not. Differentiation is not a branding exercise here. It is the mechanical precondition for being quoted at all.

The evidence from 661 dealerships

Our 2026 Dealership AI Search Study read 113,094 AI prompts across 661 US franchise dealerships and four engines. The findings map directly onto the commodity problem.

Read those three together and the conclusion is uncomfortable for our industry, ours included. You cannot out-code, out-review or out-platform the store down the road. You can only out-specify it.

What 15 real service pages scored

We wanted to see the commodity problem at page level, so this week we pulled the service pages of franchise rooftops the engines had cited in six real markets during our audits, and scored the 15 we could fetch on the same six-dimension test we run on our own content. Another 29 blocked our crawler or had no service page we could reach, which is its own finding.

The results, from pages that are live right now:

• ●6 of 15 scored zero. Not one dimension passed.
• ●9 of 15 scored one or less. Only 4 cleared the three-dimension floor we hold our own posts to.
• ●0 of 15 named a technician or service advisor. Not one human being.
• ●10 of 15 showed no price. The median density of named numbers was zero per 1,000 words.
• ●0 of 15 passed the outcome test. One page used a single outcome verb; none described what happened to a customer.
• ●Median length: 552 words, most of it copy that could sit on any rooftop's site.

The 4 pages that passed had one thing in common. Each published a real service price list, 27 to 41 prices on the page. Pricing alone lifted them out of the commodity pile, because a price is a fact only that store can state.

A small sample, and we say so in the limits below. But 15 pages the engines were already citing in their markets, and 6 of them had nothing an engine could quote. Imagine the pages the engines skipped.

The six-dimension test, published

Every GALAXY content draft A3 Brands ships runs through a deterministic commodity test before a human reviews it. The framework comes from Google's own commodity-versus-non-commodity guidance, adapted for dealership work, and the principle fits in one sentence. Commodity content tells readers what to know. Non-commodity content tells them what we did, what happened, and why it mattered.

The six dimensions, exactly as our reviewers apply them:

1. 1.Situational specificity. The page describes an identifiable case, not a category.
2. 2.Named numbers. At least three specific, verifiable numbers tied to outcomes.
3. 3.First-person practitioner voice. The dealership, or the agency, is the actor doing the work.
4. 4.Proof of work. Evidence of real operational practice, not boilerplate.
5. 5.Stakes or outcome. The page names the consequence, the decision, or the result.
6. 6.Unrepeatable angle. At least one element that depends on local data, named people, or a process nobody else runs.

The verdict rules are strict. A post passes with no failed dimensions. It comes back to the writer at one or two misses. It fails outright at three misses, or at any score with zero unrepeatable angle, because a page any competitor could have written is a commodity by definition.

We publish the test because it costs us nothing to share and it costs a competitor everything to fake. You cannot pass it with a better template.

What each dimension looks like at a store

The six dimensions sound like writing advice. At a dealership they are operational facts, and each one has a commodity version and a non-commodity version.

Notice what the non-commodity column has in common. Every line required someone at the store to supply a fact. That is the point, and it is also the problem, which we get to below. The service-page fight against quick-lube chains we covered in why quick-lube chains outrank service departments is won or lost entirely in that right-hand column.

How we do it: research before writing

Nobody can write a non-commodity page about a store they do not know. So the process starts with research, not copy, through the GALAXY system, the layer that studies a dealership before a page exists.

Brand DNA captures what the store can say that is true only of it. Founding year, named people, signature programs, the awards with dates, the things the GM says in the hallway that never made it onto the website.

Audience DNA captures the questions buyers in that market ask, in their words. The service question, the lease-end question, the trim question. Our study found the questions are the map, and most stores never published the answers.

Competitor DNA captures who the engines cite instead, which we measure directly rather than guess. Our guide to AI visibility competitor research walks through the method, and the finding repeats in every market: the store with the specific page owns the answer.

The research produces a fact base. The writer works from that base and from nothing else, which is how the six dimensions get satisfied by design rather than by decoration.

How we do it: write inside the test, gate before publish

The draft is written inside a closed world. The writer may use the store's facts, the market research, and verified third-party sources. The writer may not invent a number, a name, or an outcome. If a dimension cannot be satisfied from the fact base, the page waits for the fact rather than shipping a placeholder.

Then the gates run, in order.

The commodity test, the deterministic six-dimension check, runs first. A draft that fails comes back with the failing dimensions named.

The prose gate runs second. It catches the patterns that make a page read machine-written, the triple-parallel cadences, the colon reveals, the rhetorical questions, the empty adverbs. We wrote about the whole discipline in what 23,745 AI citations reveal, and the short version is that engines quote pages that read like a person who did the work.

Source verification runs third. Every third-party statistic links a primary source that we fetched and confirmed contains the number. A 200 status code is not verification. The page containing the figure is.

Human review runs last. The gatekeepers are people, not scripts. The scripts exist so the people spend their judgment on judgment calls instead of catching boilerplate.

Every post we have shipped since these gates went in passed all four before it went live, this one included.

How we do it: measure, then measure again

A non-commodity page is a hypothesis until the engines cite it. So the last step is measurement, and it is the step most content programs skip.

Each client store gets an AI visibility score at the start, the same audit anyone can run free through Saggy. The audit asks the engines the four questions buyers in that market ask and counts, per engine, whether the store is cited, mentioned, or absent, and who got named instead.

Then the pages ship, and the score gets re-run monthly. Citations move first, referral traffic from the engines second, leads third. A band change or a question flipping from absent to cited is signal. A three-point wiggle is noise.

The outcomes are the only proof that matters, and we publish them with the analytics attached. A Houston CDJR store picked up 93% more leads in 60 days. A Florida Toyota store went from 104 leads a month to 284. Both programs were built on the loop above: research the store, write inside the test, measure the engines, repeat.

The bottleneck is the store, not the writer

Vendors leave this part out of the pitch.

Every non-commodity element on a page is a fact only the store can supply. The price list. The technician's name and years. The Saturday queue. The 31 battery terminals. A writer cannot invent those, and a writer who does has produced commodity content with extra steps, plus a liability.

Most dealerships do not supply them. Not from bad faith, but because nobody at the store has ever been asked, and the service director has a drive to run. The single biggest reason dealership content stays commodity is that the specific facts never leave the building.

So the real work of a program like ours is intake. Structured questions, on a schedule, to the people who hold the facts. The 20 questions the floor answers every week. The prices as of this month. The one thing about this store that no other store can claim. It takes an hour a month from someone who knows the operation, and it is the hour that decides whether the engines ever have anything to quote.

If a vendor promises non-commodity content without asking your team for anything, they are describing a template with your logo on it.

What you can do this week

None of this requires an agency to start.

Limits and disclosure

The examples in the dimension table are illustrations, written to show the shape of a non-commodity line, not quotations from a specific client page.