Someone in your town opened ChatGPT last week and typed something close to “best [your trade] near me.” They got back a short list. Three or four businesses, a sentence about each, maybe a link. You were either on that list or you were not.
According to BrightLocal’s 2026 research on AI and local business recommendations, published in March 2026, 45% of consumers have used AI tools to find local businesses, up from 6% the year before. That study surveyed a representative panel of 1,002 US adults. AI is now the third most-used source for finding a local business, behind only Google and Facebook.
The useful part of this is not the headline. It is that we know, roughly, where those AI answers come from, and you can go check your own.
The number that moved fastest
Six percent to 45% in twelve months is a step change, not a trend line.
The same BrightLocal research breaks it down by tool: ChatGPT at 31%, Google’s AI Mode at 23%. It also breaks it down by age, and this part tends to surprise owners. Adults 30 to 44 lead at 64%. The over-60 group sits at 24%. If your customer base skews toward homeowners in their thirties and forties, which describes most contractors, dentists, insurance agents, and family attorneys, you are looking at the group that picked this up fastest.
Pew Research Center’s June 2026 report on Americans and AI, based on 5,119 US adults surveyed in February 2026 through its American Trends Panel, sets the wider frame. Half of US adults, 49%, have used an AI chatbot at all, up from 33% in 2024. Around a quarter use one daily. And 60% say they read the AI summaries that appear at the top of search results.
That last figure matters more than the chatbot numbers do. Most people have not switched away from Google. They are reading an AI answer inside Google.
What people ask AI, and what they do next
The Pew research found that searching for information is the single most common use of chatbots, at 42% of users. Not writing emails, not making pictures. Looking things up.
And people do not take the answer at face value. Among AI users in BrightLocal’s study, a subgroup of 455 people inside that 1,002-person panel, 88% check the reviews an AI cites. Just over half, 51%, are checking whether the review is legitimate. Another 37% want to see where it came from. Nearly all of them, 97%, at least sometimes verify an AI recommendation against real reviews, and 42% always go to the native review platform to look for themselves.
So the AI answer works like a shortlist. Your customer then goes and checks you out the ordinary way, on Google, on Facebook, on your own site. Two things have to be true for that to end well. You have to make the shortlist, and you have to survive the check.
Where AI gets its local answers
BrightLocal ran a hands-on test of this, published in July 2025 and updated that October, written up in AI Search Makes Local Listings More Important Than Ever. Worth saying plainly before the findings: this was a small sample. Twenty searches across 10 industry niches, repeated on Google AI Mode, Gemini, Perplexity and ChatGPT Search, run by hand. Directional, not definitive.
The direction is consistent, though. Business websites were the dominant source across every model and every industry tested. In an earlier BrightLocal test, ChatGPT used business websites as a source 58% of the time. Yelp showed up as a source in 33% of the tested searches. The models also leaned heavily on industry-specific directories, the kind that only matter inside your particular trade. Instagram, Facebook and YouTube appeared as cited sources too.
Your website, directories, and social and review profiles. That is the raw material an AI has to work with when your name comes up.
Why ranking in ordinary Google still predicts getting cited
This next finding should take some pressure off.
Semrush analyzed 5,000 randomly selected keywords and more than 150,000 unique citations in a July 2025 study comparing Google AI Mode against traditional search and other LLMs. It found that 92% of AI Mode responses displayed a link sidebar, averaging around seven unique domains, with 51% domain overlap and 32% URL overlap against Google’s own top 10 results. When links appeared below the response instead, which happened on 7% of queries, the overlap rose to 89% by domain and 80% by URL. Semrush also found that 13.49% of AI Mode queries included a local pack.
The broader conclusion from that same study is the one to hold onto. Domains with more Google top-10 rankings were consistently more likely to be cited, across every model tested. The work that gets you ranking in regular Google search is the work that gets you cited in AI answers. There is no separate track to run on.
Google’s own answer to “how do I optimize for AI”
Google publishes documentation on this. The AI features and your website page in Google Search Central, last updated in December 2025, states: “There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary.”
Pages need to be indexed and eligible to show with a snippet. That is the requirement. Google explicitly says no new machine-readable files, no AI-specific text files, and no special schema are needed. Among its listed best practices, Google names keeping your Business Profile information up to date.
Which means anyone selling you an AI optimization package built on proprietary files and secret schema is selling you something Google says does not exist.
The three-source audit
Set aside an hour. You need a phone and a laptop. Work through these in order.
1. Your website
Open your homepage on a phone. Can you find your phone number, address, and hours without scrolling past the fold or opening a menu? If not, fix that first. Nothing further down this list earns anything while the two details every visitor came for sit three taps deep.
Then write down the ten questions customers ask you on the phone every week. Do you take insurance. Do you charge for estimates. How far do you travel. Search your own site for each answer. Every question without a plain answer somewhere on a page is a question an AI cannot answer about you.
2. Your listings
Open your Google Business Profile. Check hours, including holiday hours, then categories, service area, and phone number. Search your business name plus your city and write down every directory that comes back in the first three pages. Check each one. Then find the two or three directories specific to your trade, the ones your customers know by name, and check those. Same business name, same address, same phone, everywhere. It is dull work. Do it anyway.
3. Your reviews and social profiles
Look at Google, Facebook, and the main review platform for your industry. Are the review counts thin? Are recent reviews sitting there with no response? Given that 42% of AI users in BrightLocal’s 2026 study always go check native review platforms, a thin or stale review profile can undo a citation you just earned. Confirm that your Facebook and Instagram profiles carry the same address, hours, and phone number as everything else.
Then run the test itself. Ask ChatGPT and Google’s AI Mode for the best business in your category in your city, and see who gets named and what gets cited. Save that answer somewhere so you can compare in ninety days.
What to do this month
Done properly, none of this looks impressive from the outside. A Business Profile that is still accurate on a Tuesday in February, not just the week it was built. The same phone number in every listing. A website that answers the ten questions your customers ask, in the words they use, on pages that load fast and get indexed. Reviews that get a reply. Then the same pass again next quarter.
That is maintenance, and maintenance is what gets dropped when you are busy running the business. Most of our digital marketing work is exactly this: doing the ordinary thing on schedule, on your behalf, so it stops living on your list unfinished. There is no shortcut version of it, and Google has already said so in its own documentation.
Run the audit above this week, and if the picture is worse than you expected, resist the urge to fix all three sources at once. Pick the one that failed hardest and give it one evening. Then put a note in your calendar for ninety days out to ask ChatGPT and Google’s AI Mode the same question again, and compare the answer to the one you saved. That second answer is the only scoreboard here that belongs to you.