The observation
A common worry among small business owners goes like this. "Big companies have the budget, the brand and the ad spend. AI will just recommend them." It is a reasonable fear. In a small field test run in October 2026, it was not what we saw.
What we saw, in two towns on two days, was that local businesses filled most of the recommendations and the national chains mostly sat in the paid ads. We have a short line for it: The big chains are in the ads. The local businesses are in the answer.
That does not mean chains are shut out of everything. It means that on the specific questions we asked, the assistants were not choosing by size. The reasons are worth understanding, because they give a small business something concrete to work on.
This page looks at the local side of the story. For the general rule about how assistants choose, see how AI assistants pick businesses. For the paid side, see ads in AI overviews. If you want a team to build the evidence for you, our page on how a team builds the proof AI looks for describes what that work is.
What we saw, category by category
In early October 2026, Matt Goren asked AI assistants for local recommendations in Richmond, Virginia. The assistants were Google AI Mode, ChatGPT logged out, Claude, and Gemini for one category. He covered coffee, pizza, AC repair, plumbers and oil change.
Coffee. Across 5 answers, the assistants gave 27 recommendations. No national coffee chain was among them.
Pizza and AC repair. No national chains in any answer.
Plumbers. Local companies came first in every answer. National franchise brands appeared only in Google's Sponsored ads shown under the answer.
Oil change. The assistants did name national chains first. But they picked locations one by one, using reviews, and always placed them next to independent shops. One assistant told the user to skip one specific chain location because of one bad review.
That last result matters for the whole picture. It shows that even when a chain wins, it wins as a specific location with a specific review record, not as a brand. A chain with strong local reviews can do well. A chain location with a recent bad review can be told to skip.
We are not naming any chain or any local business. Names would turn a pattern into a ranking, and it is a pattern we want you to test, not a list to memorize.
The small-town re-run
The same day, the test was repeated in DeLand, Florida, using Google AI Mode and ChatGPT. It held, and it looked stronger.
A plumber with 19 reviews was in the answer. Meanwhile, companies with 4,600 and 10,400 reviews, including national franchise brands, sat in the Sponsored ads.
Think about what that says. The company with 19 reviews was not winning on volume. By any simple count it was tiny next to the others. It was named anyway, which suggests the assistants were reading something other than total review count. Plausible candidates are what the reviews said, whether other sources such as local guides or forums mentioned the plumber, and how closely the business matched the exact question.
We cannot prove which of those it was from this test alone. What we can say is that a small number of reviews did not keep a local plumber out of the answer, and a very large number did not get a franchise into it.
One more difference showed up. In the small town the assistants agreed on the top coffee shop. In Richmond, each assistant picked a different one. A smaller market has fewer businesses and fewer sources, so agreement is easier. In a bigger market, the field is wider and the assistants disagree.
Why this might be happening
We should be careful here. We saw a pattern, and these are explanations that fit it, not facts we have proven.
First, the assistants built their answers from what other people said, never from the businesses themselves: review sites, local forums and blogs, community polls and the local paper. One assistant even called a company's description of itself a lead to check rather than an endorsement. If assistants treat a company's own words as a claim and outside voices as evidence, then the business with more outside voices about the specific job has the edge.
Second, local businesses are often talked about in local places. A town has a newspaper, a subreddit, a community page, a few bloggers who love a particular dish. A national brand's footprint is spread thin across every town. In any one town, it may have fewer voices speaking about that specific location.
Third, recommendation questions are specific. "Best pizza near me" is about one place and one kind of food. A national brand is built to be the same everywhere, which is its strength and also means little about any specific place.
Taken together, this suggests a rule: the local business with the most outside voices backing it for the exact job asked about gets named. A big ad budget can put a chain in the sponsored slot, but not in the answer itself. We explain the rule in more detail in our flagship piece on how assistants choose.
The limits, stated plainly
Before anyone builds a plan on this, the limits need to be clear.
- The whole test is two places and one day each, across five categories.
- Part of it ran on a signed-in account, which may have colored some answers.
- It shows a pattern worth retesting, and proves nothing on a national scale.
A few more cautions apply. Assistants change often, so what we saw in October 2026 may not hold next spring. Different categories behave differently, as oil change showed. And an assistant asked the same question twice may name different businesses.
There is also a reading we want to head off. "AI prefers small businesses" is not what we found. We found that, in these examples, assistants did not choose by size. A small business with no independent evidence is not helped by being small. A local business that is vouched for in several places is helped by that evidence, and the evidence is what you can build.
If you want to check it for yourself, ask your own five questions in two or three assistants and look at who is in the answer and who is in the ads. It takes about an hour.
What a small business should do about it
The good news is that the work is the kind a local business can actually do, without an ad budget.
Get specific reviews. Ask customers to name the job. "Fixed our burst pipe at 2 a.m." gives an assistant something to match. Our page on reviews and AI recommendations shows how to ask without pressure.
Be known for one job at a time. A page, a set of reviews and a mention for each exact job you want to be recommended for.
Show up in local places. Local press, community pages, niche guides, and the forum where your neighbors ask for recommendations. We cover the do and do not of that in Reddit and AI search.
Complete your maps profile. It is a main source for local questions. The full list of local sources is in AI search optimization for local business.
Write your own pages plainly. Say what you do, where and for whom in the first sentence. That is the claim that outside voices will confirm.
Keep ads if they pay. The paid slot is a different channel from the answer. Just do not expect it to put you inside the answer.
Give this work a quarter before you judge it. Reviews arrive one at a time, a local article takes weeks to land and a forum recommendation cannot be rushed. Measure monthly with the same five questions, write down who is named, and look for movement in how you are described, not only whether you appear. Small shifts in wording, such as being called the place for a specific job, are often the first sign that independent evidence is starting to add up.
A fair word for franchise and multi-location owners
If you own a franchise location or manage several, none of this is bad news. It is a clearer map of where your effort counts.
The oil change result is the useful one. Chains were named, but location by location, on reviews. That means each location is judged as its own business. A location with detailed, recent, specific reviews and a complete maps profile can be named. A location with thin or stale evidence can be skipped, even under a famous name.
So the work is local even if the brand is national. Each location needs its own accurate profile, its own steady flow of reviews that mention the actual service, and its own presence in local places such as community pages and local press. A national marketing team cannot do that from one office. The owner on the ground can.
One caution. Brand rules sometimes limit what a location can publish or claim. Check them before changing profiles or pages, and work inside them.
An illustrative example, not a real case
Here is a made-up example, labeled as such, to show how the pieces fit. We are not describing any real business.
Imagine two pizza places in a mid-sized town. One belongs to a national chain. The other is a family shop. A resident asks an assistant for the best pizza for a birthday party of fifteen kids.
The chain has a polished site and a large ad budget. Its location page says "great pizza, fast delivery." The family shop has a modest site, but a local parent blog wrote about its party room, a community group thread has three comments saying the owners were patient with a big group, and a dozen reviews mention "party" and "kids" by name.
The question is about a specific job: a party for fifteen children. Which business do independent sources connect with that job? The family shop. The chain may still buy the ad slot under the answer. The family shop is more likely to be the sentence inside it.
The lesson is not that being a family shop wins. It is that the shop had several independent voices matched to the exact job, and the chain had its own claim and nothing else.
What assistants cannot see about you
It helps to remember that an assistant knows only what it can read. If something good about your business is not written down anywhere public, it does not exist for the assistant.
That covers a lot of small business strengths. Your twenty years of experience, the way you stay late to finish a job, the fact that half your customers are repeat customers. If none of that appears in reviews, local articles, forum comments or your own pages, the assistant cannot weigh it.
The fix is not exaggeration. It is writing the truth down in the places that count. Put the facts on your own pages in plain sentences. Make it easy for customers to mention the specifics. Give a local writer something real to report. Each one turns a private strength into a public one that an assistant can find.
For the part of this that lives on your own site, our guide to writing for LLMs shows how to phrase things so they can be understood and quoted.
How RunOctopus handles this
Outside evidence takes time and reviews are yours to earn. The clear, specific pages that evidence points back to are the part RunOctopus writes and publishes for you.
RunOctopus is a done-for-you service. Its AI, Ollie, researches a business, then writes and publishes pages directly on the business's own website. A team plans, writes, reviews and publishes alongside Ollie. It works with Shopify, WordPress, WooCommerce and Wix. There is nothing to install, nothing to approve, nothing to upload and nothing to schedule, and pages go live on their own.
The first week brings 8 long-form articles, 6 pages and a custom interactive tool built for the business's niche. After that it is 175+ pieces of content a month, written, published and linked. RunOctopus also reports on where ChatGPT, Perplexity, Claude and Gemini name the business. The 90-day promise is more visibility on Google and in AI answers within 90 days, measured against the 90 days before joining, or every dollar back.
For scale, All Angles Creatures, the reptile supply store where RunOctopus started, went from 0 to 4,500+ Google clicks a month in five months (Google Search Console, 28 days to Sep 29, 2026).
You can see the plan before deciding anything. Paste your address into the free site read and Ollie reads the site and shows the plan. No account, no call.