Neighborhood questions named different shops than citywide ones

On October 29, 2026, Uberall published a vendor study of which local businesses show up in AI answers. Uberall reported that the wording of the question changed which shops the models named.

Uberall sells software that tracks answers like these. Every figure below is company-reported. The brand figures are for coffee shops in San Francisco.

What Uberall counted

Over 14 days, Uberall gathered 11,705 answers from 12 models. The models included ChatGPT, Claude, Gemini, Grok, and Perplexity. The questions asked about local businesses in San Francisco.

For coffee, Uberall tracked 39 brands across those 12 models and three kinds of questions. Uberall reported that 51% of the 39 brands were mentioned by all 12 models, and that 92% appeared in at least four of the models.

Uberall added a limit. San Francisco coffee has clear market leaders. Uberall said fewer of the same brands would show up across models for dining, beauty salons, and dental practices. The 51% and 92% figures are this study’s coffee results.

Three ways of asking

Uberall reported three kinds of questions in the coffee study, and said those patterns showed up across the 11,705 answers.

Discovery questions asked for the best coffee shops, or for top-rated cafes. Uberall reported a mention rate of 79% to 86% for the top names. A mention rate is how often the answers named that brand. Uberall said these questions favored the brands with the strongest overall reputation.

A second kind of question named a use. Uberall’s examples were best to work from, and a business meeting spot. Uberall said the answers matched that use. One brand reached the top five in that group of answers, and only in that group, because it was known as a laptop-friendly place.

Neighborhood questions named a part of the city. Uberall’s examples were coffee near Union Square, and the best in the Mission District. Uberall said these answers put weight on where the shops are, and on the shop’s tie to that place, ahead of overall reputation.

The same brand, two mention rates

Starbucks is the worked example Uberall gives.

On discovery questions, Uberall reported a 0.2% mention rate for Starbucks. On neighborhood questions, the mention rate was 32.6%. The brand stayed the same. The question changed, and the rate changed with it.

A specialty coffee brand that led the discovery questions moved the other way. On neighborhood questions, Uberall reported a 1.3% mention rate for that brand. Uberall does not name it. Uberall’s read is that a physical address was likely not enough to connect the brand with the neighborhood. Uberall points to reviews, listings, and other public text that name the neighborhood.

Uberall reported that weekday and weekend recommendations were nearly identical across all 12 models over the 14 days.

Longer answers named more shops

Uberall reported that models like Claude Opus and ChatGPT o3 mentioned an average of 10 brands per answer. Uberall said those were the longest answers, and that they included established businesses and lesser-known ones. Models like Claude Haiku mentioned fewer than 5 brands per answer. Uberall said those names concentrated on well-known chains.

Write down every business the answer names. The shorter list can leave out a shop that the longer list includes.

Uberall describes ChatGPT, Claude, Gemini, and Grok as answering from what they learned in training, up to a fixed point in time. Uberall says that habit favors brands with a long, well-documented history online.

Uberall describes Perplexity as searching the live web on each question. Uberall wrote that a newer business with strong recent reviews can show up prominently in Perplexity and barely show up in ChatGPT. Keep those answers as two lists.

A test on one service you sell

Pick one service customers ask you for, and one neighborhood you serve. The coffee percentages stay with the San Francisco study. This test is a dated record of public answers. An edit to your page is not a promise that an AI tool will name your business.

Write two questions in words a customer would use.

Citywide: best [service] in [city].
Neighborhood: [service] near [neighborhood].

Ask each question in two tools. For one, use a tool Uberall groups with answers from earlier training: ChatGPT, Claude, Gemini, or Grok. For the other, ask Perplexity, which Uberall says searches the live web each time. Save four lists. On each, write the date and every business named.

Then apply this rule.

The citywide answer names you, and the neighborhood answer does not. Add the neighborhood and the service, in ordinary words, on your website and on your Google Business Profile. Name only a neighborhood you serve.
The neighborhood answer names you, and the citywide answer does not. The local words may already be public. Uberall said the discovery questions favored brands with the strongest overall reputation. Another page is not a promise that a citywide answer will include you.
No answer names you. Save the names and the date. The study does not state a cause.
One tool names you, and the other does not. Keep both lists.

What to do this week

Write one citywide question and one neighborhood question for one service you sell, in a neighborhood you serve.
Ask both in two AI tools. Save every business name and the date.
Read your service page and your Google Business Profile for that neighborhood and that service.
Where a neighborhood you serve is missing, add it in public text on the page and on the profile.
Keep the four lists separate. The first name in a short answer is one line in one list.

In New Reward’s Rival Map, each question your customers ask is marked you, your rival, or nobody, and every mark is dated.