What AI Says About Your Business: We Ran the Test

12 min read

Editor’s note. We set the method and scoring rules before seeing the results. The answers below are reproduced as generated, including the ones that complicated our theory. Peer-reviewed research and commercial evidence are identified separately.

TL;DR Tap to expand the short version

What AI says about your business is now something customers read before they buy. More and more of them ask ChatGPT, Gemini or Google’s AI to describe a company before they walk in or check out.
A popular claim in marketing circles says these tools sometimes sand off what makes a business distinctive. They instead show a generic version. We wanted to see what happens when AI describes a real business whose facts already exist online.
We built the Flattening Test with four real businesses and differentiators we verified from public sources first, to see if the distinctive fact survived in the AI responses.
We found that the tasty, visible, quotable stuff (a signature dish, a secret menu) survived about 83% of the time. The facts about how the business is run (family owned, never franchised, only two locations) survived zero times. One answer even made up details that were wrong.
The takeaway is that the danger is not that AI forgets you. It is that AI remembers your gimmick and forgets your backbone. There is a simple, free way to check your own business, below.

Echo-chamber claim
Supporting evidence
Who loses most
The four-company test
- Our scoring logic
- Results
A one-hour audit
Test limits
FAQ
Related reading

The flattening claim

A customer once searched for your website and reviews before deciding whether to visit. Now they may ask ChatGPT or Gemini to explain your business in a few lines.

That answer can become the first impression, even though the business did not write it and may never see it.

Sometimes the description arrives with its edges sanded off. A forty-year refusal to franchise becomes “a beloved local institution.” A kitchen built around one clear idea becomes “quality seasonal ingredients.” The detail a customer might repeat to a friend disappears into language that could describe a competitor.

Language models learn from huge amounts of existing text. When they describe a business, do they preserve what makes it different or return the familiar language of its category?

A growing group of companies sells audits of how brands appear in AI answers. Their pitch assumes that models do more than make occasional mistakes. It assumes they systematically blur the differences between businesses.

Published evidence for that claim remains limited, so we ran a small test.

Large studies of AI brand visibility usually measure whether a brand appears and whether the mention is positive. They rarely test whether a model preserves the facts that give customers a reason to choose one business over another.

Evidence behind the claim

Related research offers a reason to take the question seriously. In a Science Advances study, 293 participants wrote short stories with or without AI help. Judges preferred the assisted stories, but those stories were 10.7% more similar to one another.

An ICLR 2024 experiment found the same convergence in argumentative essays. A later study of 2,200 admissions essays found that each human-written essay introduced new ideas at two to eight times the rate of an AI-written one.

A 2026 paper attributes part of the effect to typicality bias. Human reviewers tend to reward answers that feel familiar and expected, so training can steer models toward established patterns rather than unusual ones.

Those studies examine AI helping people create essays, stories and arguments. Describing an existing business is different because the facts may already be available online.

A model may therefore describe a well-known business accurately even if AI-assisted writing tends to converge.

The earlier research supports the hypothesis that AI may flatten business descriptions. It does not prove it.

Businesses with more to lose

Any flattening would matter most to businesses that have invested in being different.

A business that resembles its competitors loses little when AI describes it in average terms. The description is broadly accurate.

A company that spent decades refusing to franchise can lose far more. Its most important distinction may vanish behind the same category language.

That creates a practical risk for businesses whose advantage rests on ownership, operating choices or a story that is less visible than the product itself.

The underlying problem is old. Good work and public reputation do not always match. AI adds another layer between a business and the customer, which makes the gap easier to miss.

The four-company test

We fixed the scoring rules, businesses, facts and prompts before reading any answer.

Each business came from an everyday category, and every fact in the test was checked against the company’s materials and at least one external source.

Business	Category	What makes it different (all verified)	Where we confirmed it
In-N-Out Burger	Fast food burgers	Never franchised and still family owned; no freezers, microwaves or heat lamps in the kitchen; a famous secret menu; deliberately stays close to its own supply depots, which caps how far it can spread	Company history, decades of press coverage
Zingerman’s Delicatessen	Delicatessen	Flatly refuses to franchise; grows by starting new businesses that all stay in Ann Arbor; famously teaches other companies its open management style	Own FAQ (“it is in our plans to NOT franchise”), Michigan archive, NPR
St. John, London	Restaurant	Built its whole reputation on “nose to tail” cooking under chef Fergus Henderson; its signature roast bone marrow dish; a deliberately bare, white dining room	Own site, thirty years of food press
Ted Drewes, St. Louis	Frozen custard stand	Invented the “concrete,” a shake so thick it is handed to you upside down; sells Christmas trees grown on its own farm in Nova Scotia; just two locations after nearly a century	Own history page, Route 66 marker, Wikipedia

We asked the same three questions about each business, using language a customer or marketer might use.

The question we asked	Exact wording	Why it is in the test
The customer question	“Describe [business] in about 100 words for someone deciding whether to go there.”	What a would-be customer clearly types.
The marketing task	“Write a short description of [business] for a local business directory listing. About 60 words.”	The kind of copy AI tools crank out for businesses at scale.
The direct challenge	“What specifically makes [business] different from a typical [category]?”	The AI’s best shot. If the difference does not show up when you ask for it head on, it will not show up anywhere.

Our scoring logic

Each verified fact received a yes-or-no score. The model had to identify the detail, not merely gesture towards it.

A vague gesture did not count. “Committed to quality and community” scored zero against “refuses to franchise,” because the first could describe almost any deli in the country, while the second could only describe this one.

For this first round, we tested two current AI models from Anthropic’s Claude family, using cold prompts with no live web search enabled. The goal was to see what these tools surfaced without assistance, because that is often the version of a business a customer encounters when they ask an AI about it.

Results

Business	Question type	Facts kept	What went missing
In-N-Out	Marketing task	1 of 4	Never franchised, no freezers or heat lamps, the supply cap
In-N-Out	Direct challenge	2 of 4	Never franchised (softened, see below), the supply cap
St. John	Customer question	4 of 4	Nothing
St. John	Marketing task	3 of 4	The bare white room
St. John	Marketing task	3 of 4	The roast bone marrow
Ted Drewes	Direct challenge	2 of 4	The Nova Scotia tree farm, the two-locations-only fact

Every distinctive fact about how a business operates disappeared from the answers we tested, including the prompt that asked directly what made each business different.

The In-N-Out answer shows how a true description can still blur the main distinction.

Asked directly what sets the chain apart, the AI wrote: “Family-owned and privately held, In-N-Out has resisted aggressive expansion, keeping quality control tight across a relatively small regional footprint.”

The wording is accurate, but it could describe many regional chains.

The checkable fact, zero franchises since 1948, became a general statement about controlled growth. A reader could accept the answer without noticing what was missing.

Ted Drewes kept its famous upside-down concrete but lost the Christmas tree farm and two-location detail. The answer also introduced errors, saying the business was “open since 1931” and “operates only seasonally.” The first St. Louis stand opened in 1930, the business dates to 1929, and its main stand closes only in January.

The test therefore found two separate problems. Models omitted structural facts and sometimes filled the gap with confident errors that sentiment checks would miss.

St. John complicates the pattern. It retained ten of twelve facts across three answers, including nose-to-tail cooking, bone marrow, the white dining room and the chef.

Its distinctive details also dominate the restaurant’s public story.

Decades of food writing have made “nose to tail” inseparable from St. John. In-N-Out’s burgers dominate its coverage, while its ownership model appears more often in business reporting. The models kept what was most visible.

Six answers cannot establish a general rule. They suggest that models preserve famous, concrete details more reliably than facts about ownership, growth and operating choices.

A larger test may show whether that pattern holds across sectors and less famous businesses.

A one-hour audit

A basic check needs no subscription or technical team. A notepad and an hour are enough.

Start by writing down what makes your business different, then split the list in two. One group is what customers can see or experience: a signature product, a service nobody else offers, a detail that immediately separates you from competitors. The other is how the business is built: family ownership, never franchised, only two locations, employee ownership, thirty years in the same spot. Keep the facts a competitor could not honestly claim, and only include details you can verify with a source. Our pilot suggests the second group is the one most at risk, so start there.
Ask an AI tool the same three questions we used. Open ChatGPT, Gemini, or whichever tool your customers are likely to use, enter your business name, and ask without guidance or corrections. Do not feed it the answer you hope to get. You want to see what a stranger would see.
Then score the response fact by fact. Do not judge whether the description sounds positive. Judge whether the details that make you different are clearly there.
A polished paragraph that leaves out every meaningful distinction is still a failure. Watch for two traps: the soft blur (“resisted aggressive expansion” instead of “never franchised”) and invented details, such as an incorrect founding year. The first hides what makes you different. The second creates a false version of your business.
The fix starts with making those facts easier to find. AI systems can only surface details that are clearly available somewhere. If your most important facts exist only in a brochure, a PDF, or your own head, they are unlikely to become part of the public description.
State them plainly on your website: “We have never franchised. Every location is company owned.” Then make sure credible outside sources mention those same facts, using the same two-source standard applied in our test. The way machines read your website is separate analysis.
Put a reminder in the calendar and do it again next quarter. These tools change without telling anyone, so a good answer today is not a promise for June. Run the same three questions every few months and keep the dated results in one document. It is the cheapest way to see, over time, whether the AI is learning your real story or losing it.

Treat it like checking how a map describes your address. The aim is to see whether a machine introduces the business with the facts that matter.

Limits of the test

This was a small pilot using four famous, well-documented businesses and two models from one provider. That design gave the models favourable conditions but limits what the results can support.

The findings do not show how other models, live web search or smaller businesses would perform. They identify a pattern that deserves a larger and more varied test.

FAQ

What does it mean when people say AI flattens a business?

Brand flattening is the loss of specific differentiators when AI describes a business in generic terms. In this pilot, visible or sensory facts survived about 83% of the time, while structural facts about how the businesses operated did not appear across six answers.

Is there real research behind this, or is it marketing hype?

Research shows that AI-assisted writing can become more similar even when readers prefer it. Whether that finding predicts how models describe businesses remains open.

Why would AI drop the facts about how my business is run?

Highly repeated sensory facts, such as signature dishes, dominate online coverage. Structural details such as ownership or operating model appear less often, so models are more likely to omit them. A larger test is needed to confirm that explanation.

How can I check what AI says about my own business?

List the facts that genuinely distinguish the business, separating customer-visible traits from operational ones. Ask major AI tools to describe the company, write a directory listing and explain what makes it different. Record missing, blurred or invented facts and repeat quarterly.

Will paying a GEO agency fix this for me?

Most visibility tools track mentions and sentiment rather than specific differentiators. Businesses can begin by stating those facts clearly on their own sites and seeking independent coverage that confirms them.

AI recommendations, on the reputation signals that decide whether an AI recommends you at all.
Agentic checkout, on where AI referred customers are clearly showing up.
false order confirmations, on what happens when the AI gets your customer’s reality wrong in the other direction.

Spotted an error, have a correction, or want to pitch a story? Contact [email protected].

How AI Models Describe Four Real Businesses

Table of Contents

The flattening claim

Evidence behind the claim

Businesses with more to lose

The four-company test

Our scoring logic

Results

A one-hour audit

Limits of the test

FAQ

What does it mean when people say AI flattens a business?

Is there real research behind this, or is it marketing hype?

Why would AI drop the facts about how my business is run?

How can I check what AI says about my own business?

Will paying a GEO agency fix this for me?

Like this:

How AI Models Describe Four Real Businesses

Table of Contents

The flattening claim

Evidence behind the claim

Businesses with more to lose

The four-company test

Our scoring logic

Results

A one-hour audit

Limits of the test

FAQ

What does it mean when people say AI flattens a business?

Is there real research behind this, or is it marketing hype?

Why would AI drop the facts about how my business is run?

How can I check what AI says about my own business?

Will paying a GEO agency fix this for me?

Related reading

Share this:

Like this:

Discover more from Industry Contents