How AI Assistants Recommend Local Businesses

Consumers moved faster than anyone expected

In one year, asking an AI assistant for a local recommendation went from a curiosity to a mainstream habit. 45% of consumers said they had used AI tools for local business recommendations, up from 6% the previous year (BrightLocal), which makes AI the third most popular source of local business recommendations (BrightLocal 2026).

The adoption is not evenly spread. 64% of adults aged 30 to 44 have asked an AI tool for a business recommendation, against 24% of people over 60. Among people who use AI for this, 31% used ChatGPT and 23% used Google's AI Mode (BrightLocal).

The wider base is consistent with that. 49% of US adults use AI chatbots, up from 33% in 2024, 42% use them to search for information, and 60% say they read AI search summaries (n = 5,119 US adults, February 2026) (Pew Research Center).

Trust is following use rather than leading it. 63% of active AI users trust AI recommendations and only 10% distrust them, while 53% of non-users say they do not trust them (BrightLocal).

What this actually changes about your traffic

The behaviour change worth planning around is not that people ask AI. It is that fewer of them click afterwards.

Pew instrumented 68,879 Google searches from 900 US adults in March 2025. Users who saw an AI summary clicked a traditional result in 8% of visits, against 15% for those who did not, and clicked a link inside the summary itself in just 1% of visits (Pew Research Center). 26% of users ended their session entirely on a page with a summary, against 16% without.

Longer questions are far more likely to trigger a summary: 53% of searches of ten or more words produced one, against 8% of one or two word searches. Local questions tend to be long.

The referral side is small but growing. Similarweb estimated AI platforms sent 1.13 billion referral visits to the top 1,000 sites in June 2025 against 191 billion from Google Search, roughly 0.6%, though AI referrals grew 357% year over year and more than 80% came from ChatGPT (Similarweb). Around 30% of marketers report search traffic declines as consumers shift to AI tools (HubSpot).

What is being sold to you that nobody can prove

"AI ranking factors" with percentage weights. No AI platform publishes its local recommendation signals. Any table assigning percentages to them is invented.

"Guaranteed ChatGPT visibility." There is no submission process, no verified ranking dashboard and no purchase that places a business in a model's answer.

"Schema markup gets you into AI answers." Structured data helps machines parse a page. It is sensible practice, and it is not a documented recommendation signal for any assistant.

"Publish 50 AI-optimised pages." Volume tactics that worked in early search are the same tactics answer engines are built to discount, and thin pages carry their own risks with Google.

"AI has replaced reviews." The opposite is visible in the data: 88% of consumers fact-check the reviews an AI tool cites (BrightLocal).

If a vendor cannot tell you which observable input changed and how they measured the result, they are selling a story.

Aim to be citable, not ranked

The useful reframe is that an answer engine is not ranking you. It is assembling a short answer and citing a few sources, so your job is to be one of the easiest sources to cite correctly.

That means being consistent everywhere a machine can read you, being unambiguous about what you do and where, and having recent, specific, human evidence that supports the claim. 88% of AI summaries cite three or more sources (Pew Research Center), so the practical question is whether your business is easy to include among them.

There is a second reason to care about the wider corpus. Web search was enabled on only 34.5% of ChatGPT queries in February 2026, down from around 46% in late 2024 (Semrush), which means most answers are drawn from the model rather than a live fetch. You cannot influence a model's weights. You can influence whether the public record about your business is accurate, plentiful and consistent.

The honest summary: this is reputation and data hygiene work with an uncertain payoff timeline, not a channel you can dial up.

What you can actually influence

1. Fix your listings first. Name, address, phone, hours, categories and service area, identical everywhere. Expert consensus in Whitespark's 2026 study puts three of the top five AI search visibility factors in the citation and structured listing category (Whitespark). Expert consensus is not platform disclosure, but it is the best available evidence.

2. Complete your Google Business Profile properly. Google reports complete profiles get seven times more clicks, regularly updated profiles get five times more views, and consumers are 29% more likely to consider a business with a detailed profile (Google Business Profile Playbook 2026).

3. Keep earning real reviews, unrewarded. Review corpus is both an input to answers and the thing consumers check afterwards, and 88% of them do check (BrightLocal).

4. Answer the long questions plainly on your own site. Long, question-shaped queries are the ones that trigger AI summaries. Write pages that state the answer in the first two sentences.

5. Say the specifics out loud. Parking, accessibility, dietary options, appointment lead times, price band, neighbourhood. Assistants recommend on constraints, and unstated facts cannot be cited.

6. Publish first-party evidence. Customer video and quoted testimonials on your own pages give a crawler something specific to attribute to you rather than to a directory.

7. Measure what you can. Check referral traffic from AI domains in your analytics, and test a handful of prompts monthly.

8. Do not buy signals. Purchased followers, likes or views are directly prohibited by the FTC rule.

Toutedly contributes to exactly one part of this: it produces first-party customer footage and captions you can publish on your own pages and profiles. It does not influence AI rankings, and no tool can honestly claim to.

Signals, ranked by how much you can influence them

The table below separates what is known from what is guessed. Every row marked as inference is exactly that.

Two things follow from this table. First, almost everything you can influence is work you would do anyway for conventional local search. Second, anyone offering to move the bottom three rows is describing something that cannot currently be verified.

The traffic that does arrive behaves well. Adobe reported AI-referred visitors converting 31% more than other sources, bouncing 33% less and viewing 13% more pages (Adobe Analytics). That is a reason to make the landing experience good, not evidence that AI visibility can be bought.

Wording and prompts you can copy

Business description, written to be parsed:

[Business Name] is a [category] in [neighbourhood], [city]. We [specific service] for [customer type]. Open [days and hours]. [Parking or transit detail]. [Booking method].

A page answer, written for a long question:

Do you take walk-ins on Saturday? Yes. We hold [number] chairs for walk-ins every Saturday between [time] and [time]. The wait is usually [range] minutes. Booked appointments are taken first.

Specifics worth stating in plain sentences:

Wheelchair accessible entrance from [street]. Gluten free options on every menu section. Same day appointments usually available before 11am. Typical price range $[x] to $[y].

Prompts to test your own visibility, run monthly:

"Best [category] in [neighbourhood], [city]" "Where can I get [specific service] near [landmark] on a Sunday?" "[Category] in [city] that is good for [specific constraint]"

Unrewarded review ask, to keep the corpus growing:

If you have a minute, an honest review on Google helps other people find us. No discount attached, we are not allowed to offer one.

Caption for first-party video published on your own pages:

Dana came in for [service] and described what happened, in her own words. Dana received a discount for sharing her honest feedback.

What not to write, for comparison:

"Voted the best [category] in [city] by AI" or any claim of AI endorsement you cannot evidence.

Toutedly generates the caption and hashtags for each submission so first-party evidence is easy to publish on your own pages, which is the only lever on this page it touches.

Your first 30 days

Treat month one as an accuracy audit, not a growth campaign.

Repeat the prompt test monthly. Track what changes and, more importantly, whether anything you fixed showed up in the answer.

What you can and cannot do legally

This is general information for US businesses, not legal advice, and rules change.

You cannot buy engagement signals. Section 465.8 of the FTC rule addresses the sale and purchase of fake indicators of social media influence, including followers and views, where the buyer knew or should have known they were fake (full rule text). Buying an audience to look authoritative is now explicitly in scope.

You cannot pay for the reviews that feed these answers. The rule prohibits compensation in exchange for, or conditioned expressly or by implication on, a review expressing a particular sentiment (16 CFR 465.4), and Google separately prohibits incentives for any review and removes incentivized content from Maps (Google policy).

You cannot filter who you ask. The FTC tells marketers not to seek reviews only from customers expected to be positive (FTC guide for marketers).

You must disclose incentivized testimonials you publish, including on pages you hope an assistant will read. A reward creates a material connection that must be disclosed clearly and conspicuously (16 CFR 255.5).

Do not claim an AI endorsement. "Recommended by ChatGPT" is an advertising claim like any other. If it is not substantiated and current, it is deceptive.

Be careful with AI-generated content about your business. Fabricated testimonials or reviews generated by a model are covered by the rule's prohibitions on fake and misrepresented reviews (Federal Register).

Mistakes to avoid while this is still unsettled

Paying for AI optimisation you cannot audit. Ask what input changed and how it was measured. If there is no answer, there is no service.

Rewriting your whole site for machines. Pages that read like specification sheets lose the humans who still click.

Ignoring listings because AI feels new. The most credible evidence points at citations and structured listings (Whitespark), which is unglamorous work.

Abandoning reviews. Reliance on Google alone for reviews dropped 12% while AI, TikTok and Apple Maps grew (BrightLocal), but consumers still verify what AI tells them.

Panicking about traffic without measuring it. Check your own analytics before acting on a headline. AI referral share is still small in absolute terms (Similarweb).

Assuming today's behaviour holds. Adoption moved from 6% to 45% in twelve months. Any tactic on this page should be revisited every quarter.

Where Toutedly fits

Toutedly does not optimise anything for AI, and it would be dishonest to imply otherwise. What it does is produce first-party evidence that you can publish where machines and people both read: your own pages, your Google Business Profile, your social accounts.

The mechanism is simple. A campaign holds one prompt, one reward and one QR code. A customer scans the code in your venue, records 15 to 30 seconds of honest feedback on their own phone, and receives the reward for the recording rather than the sentiment. You get the video, a drafted caption with the disclosure line, hashtags, an @mention, and the customer's contact details with consent.

That output helps with the one input on this page you fully control: specific, recent, human evidence about what your business actually does, published under your own name. It does not put you in an AI answer, it does not verify that a customer posted a public review, and it never rewards a Google or Yelp review.

Plans are Free at $0, Pro at $49 a month and Business at $99 a month. If you are working through the 30 day audit above, the useful order is listings first, then reviews, then your own published evidence.

Questions people ask