How ChatGPT Recommends Local Businesses

ChatGPT recommends local businesses that its search crawler can reach and that other websites already vouch for, and it names far fewer of them than Google Maps does. SOCi's Local Visibility Index, published January 2026, found ChatGPT recommended 1.2% of brand locations, against a 35.9% average appearance rate for the same brands in Google's local 3-pack. Perplexity and Google AI Mode lean harder on pages that already rank in ordinary Google results.

For a Las Vegas plumber, dentist or personal injury firm, most of that selection happens before a customer types the question. Two things changed in the past year. In November 2025, Whitespark's Local Search Ranking Factors report scored each factor for AI search visibility for the first time, and in March 2026 BrightLocal reported that 45% of consumers now use AI to find local business recommendations, up from 6% in 2025. We read Google's, OpenAI's and Perplexity's own crawler documentation on September 25, 2026 for this post.

Think of three gates. The first is reach: each engine sends its own crawler, and if your robots.txt or hosting setup turns it away, nothing else here matters. The second is proof: whether lists, articles and review platforms describe you the way your own site does. The third is fit at the moment of the question, which is where profile details like hours and services come in.

The cut is steep. SOCi's 2026 Local Visibility Index announcement covers more than 350,000 locations across 2,751 brands with 50 or more locations each, scored on 120 metrics across five industry categories and 42 subcategories. Divide 35.9% by 1.2% and you get about 30, which is why SOCi calls ChatGPT nearly 30 times more selective than the 3-pack. The locations ChatGPT did name averaged 4.3 stars. Ranking well in Maps did not carry over automatically either: in retail, only 45% of the brands leading traditional local search were also among the most recommended in AI results.

The audience is real now. BrightLocal's AI trust analysis of its 2026 Local Consumer Review Survey, published March 10, 2026 from 1,002 US adults surveyed through SurveyMonkey (455 of them AI users), found 31% had used ChatGPT for local recommendations and 23% Google AI Mode. Among AI users, 63% trust the recommendations, but 97% at least sometimes check them against real reviews. The AI answer is a shortlist, not a verdict.

The 13 signals, which engine reads them, and which ones you control

The table pulls every named input from the seven sources into one place. The last column only holds a number where a source published one.

Signal	Engine	Source and date	Owner controls it?	Weight or number
Page indexed and eligible for a Google snippet	Google AI	Google Search Central: AI features, read Sep 2026	Yes	The only technical requirement
OAI-SearchBot allowed in robots.txt	ChatGPT	OpenAI crawler overview, read Sep 2026	Yes	About 24 hours to take effect
PerplexityBot allowed in robots.txt	Perplexity	Perplexity crawler guide, read Sep 2026	Yes	Up to 24 hours to take effect
Presence on expert-curated best-of lists	All	Whitespark 2026 Local Search Ranking Factors, Nov 2025	Partly	AI score 179, rank 1
A dedicated page per service	All	Whitespark, Nov 2025	Yes	AI score 170
Prominence on industry-relevant domains	All	Whitespark, Nov 2025	Partly	AI score 167
Quality of unstructured citations	All	Whitespark, Nov 2025	Partly	AI score 160
Authority of third-party review sites	All	Whitespark, Nov 2025	Partly	AI score 156
Business open at time of search	Google AI	Whitespark, Nov 2025	Yes, via accurate hours	Local pack score 189
Address shown on the profile	Google AI	Whitespark, Nov 2025	Only with a real storefront	Local pack score 176
GBP predefined services filled	Google AI	Whitespark, Nov 2025	Yes	Local pack score 134
Complete, accurate data on non-Google platforms	All	Whitespark citation factors, Nov 2025	Yes	Citations about 8% of local pack weight
Rating at or above the 4.3 average	ChatGPT	SOCi 2026 Local Visibility Index, Jan 2026	Partly	Average of picked locations, not a cutoff

Seven of the 13 rows are fully in the owner's hands, and most of them cost an afternoon. Read the scores carefully: rows 4 to 8 are AI visibility scores, while rows 9 to 11 were scored for Google's local pack. They belong here because Google's AI features page lists keeping Business Profile information current among its general practices, and AI Mode sits on the same search stack.

Perplexity tracks Google rankings closely; ChatGPT mostly does not

Semrush's AI Mode comparison study, published July 21, 2025, ran 5,000 randomly selected keywords on desktop through Google's top 10 organic results, AI Overviews, AI Mode, ChatGPT and Perplexity, and analyzed more than 150,000 unique citations. Perplexity showed over 91% domain overlap and 82% URL overlap with Google's top 10. AI Mode showed about 54% domain overlap and 35% URL overlap. ChatGPT showed the weakest alignment of all.

In plain terms, if you already rank on page one for a service, you are probably in the pile Perplexity draws from. AI Mode showed a sidebar of about 7 domains in 92% of responses, and that sidebar had 51% domain overlap with the top 10, so page one gets you roughly half way. ChatGPT is where a strong Google ranking helps least, which is why the off-site signals in the next section carry more weight there. Semrush also found commercial queries produced answers about 2 times longer than informational ones, and who is the best emergency plumber near me is a commercial query.

Reach comes first. OpenAI's crawler page says OAI-SearchBot surfaces websites in ChatGPT search and recommends allowing it in robots.txt. GPTBot is a separate agent used for model training, and blocking it does not control search appearance. Perplexity says PerplexityBot surfaces and links sites in its results and is not used to train foundation models. If a security plugin or CDN sits in front of your site, check that it is not quietly turning these agents away; that kind of setup is part of our AI integration work.

A skeptical owner will point out that Google says there is nothing special to do, and for Google that is correct. Its documentation says there are no additional technical requirements, no AI text files, and no special schema.org markup needed for AI Overviews or AI Mode. Anyone selling an llms.txt file as the fix for Google AI is selling something Google says you do not need.

Three of the top five AI factors live on someone else's website

Whitespark's 2026 report asked 47 local search experts to score 187 factors. The top five for AI search visibility were presence on expert-curated best-of lists (179), a dedicated page for each service (170), prominence on industry-relevant domains (167), quality and authority of unstructured citations (160), and authority of third-party review sites (156). The report itself notes that three of those five are citation factors.

Unstructured citations are mentions of your business in articles, roundups and local news, as opposed to directory listings with fixed fields. For a restaurant that might be a neighborhood dining guide; for a dentist, a community sponsor page or a health directory. You cannot write these yourself, but you can make them more likely: answer reporters, sponsor things that get written up, and give the people who compile best-of lists something specific to say about you.

The one top-five factor fully in your hands is the service page. A single Services page listing ten things gives an engine nothing to match a narrow question against. A page per service gives it a direct answer to quote and a URL to cite, and it is the cheapest item on the list. That is the center of how we approach AI SEO for Las Vegas businesses.

Keep the old weights in view. The same report puts approximate local pack weights at about 25% for Business Profile signals, 18 to 20% for reviews, 15% for on-page, 14% for links, 12% for behavioral signals and 8% for citations. Citations are a small slice of Maps ranking but most of the top of the AI list, and that gap is the shift worth acting on.

A one-office dentist and a three-truck plumber, worked through

Take a single-location dental practice and assume 40 general practices compete for the same part of the valley. Apply SOCi's rates as a rough ratio: 35.9% of 40 is about 14 practices that would turn up in a 3-pack at some point, while 1.2% of 40 is 0.48, less than one practice. SOCi measured chains, so treat that as a picture of the cliff, not a forecast.

Now the rating. Assume the practice sits at 4.1 stars on 120 reviews and wants to reach the 4.3 average SOCi measured among ChatGPT's picks. With n new five-star reviews, solve (4.1 × 120 + 5n) / (120 + n) = 4.3. That gives 492 + 5n = 516 + 4.3n, so 0.7n = 24 and n is about 34. If the front desk collects three a week, that is roughly 12 weeks of steady asking, with no new low ratings along the way.

The controllable rows take far less time. In one week the practice can confirm its pages are indexed, allow both search crawlers, split one services page into dedicated pages for implants, emergency visits, clear aligners and children's cleanings, and fill in Google's predefined services to match.

Now a plumbing company with three trucks, run from the owner's home on a service-area listing with no public address. The address factor scored 176 for the local pack, number 7 among Whitespark's new factors, and this business cannot claim it honestly. Renting a mailbox to fake a storefront puts the listing at risk. The leverage is elsewhere.

Open at time of search scored 189, number 5 among the new factors for local pack importance. A plumber who takes calls around the clock should list hours that say so, because a 10 pm slab leak search that finds the business closed passes it by. Those hours are also a promise: if the profile says open, someone or something has to answer at 2 am. Assume six core services, from water heaters to slab leak repair, and that is six pages to write, plus outreach to the trade's best-of lists, the factor that scored 179.

Where these numbers do not hold

SOCi's index covers only brands with 50 or more locations, so no single-location Las Vegas business is in the sample. The 1.2% rate and the 4.3-star average describe chains, and 4.3 is the average of who got picked, not a published cutoff. A 4.2-star independent can still be named.

Whitespark's scores are the judgment of 47 practitioners, not measured clicks or calls. They show what professionals believe matters; they do not prove any single factor causes a recommendation.

Semrush's overlap data dates from July 2025, ran on desktop, and used random keywords rather than local ones. It is still the latest cross-engine overlap study Semrush has published, since its June 2026 AI Visibility Index did not repeat the figures, but engines change retrieval quietly. BrightLocal's AI figures rest on 455 AI users nationwide, not a Nevada sample.

Robots.txt is not an off switch. Perplexity says Perplexity-User, which fetches a page when someone asks a question, generally ignores robots.txt, and OpenAI's ChatGPT-User visits pages on a user's behalf. Blocking the search crawlers removes you from the index those engines draw on; it does not keep every AI agent away.

Finally, the effort may not pay. If nearly all your work comes from referrals, repeat clients or a few commercial accounts, AI recommendations matter less than service. Law firms also carry advertising and testimonial rules a plumber does not, so check those before any review push. If you are unsure the channel deserves a budget line, that is a consulting question before it is an SEO project.

Questions owners ask

Do I need an llms.txt file to show up in AI search?

Not for Google. Its documentation says AI Overviews and AI Mode need no AI text files, no new machine readable files and no special schema markup. A page that is indexed and eligible for a snippet qualifies. OpenAI and Perplexity ask only that their search crawlers be allowed in robots.txt. Spend that money on service pages and third-party mentions instead.

Does blocking GPTBot keep my business out of ChatGPT?

No. GPTBot is OpenAI's training crawler, and OpenAI says blocking it does not control search appearance. The agent that matters for being named in ChatGPT search is OAI-SearchBot. You can opt out of training and still allow search, which is what most owners actually want. Check both lines in robots.txt, because a blanket rule blocking all bots catches both.

ChatGPT tracks Google organic results less closely than Perplexity or AI Mode, and it names far fewer locations than the map pack shows. A competitor that appears on best-of lists, gets mentioned on industry sites and holds strong reviews on authoritative platforms can win even with a weaker Maps position. In SOCi's retail data, Maps leaders and AI favorites often diverged.

What to do this week

Check reach. Open your robots.txt and confirm neither OAI-SearchBot nor PerplexityBot is disallowed and no blanket rule blocks all bots, then confirm in Google Search Console that your service pages are indexed.
Close the cheap gaps. Compare the services you sell to the pages you have, write the first missing page, fill in every Google predefined service that applies, and correct your hours, holidays included.
Map who gets named. Ask ChatGPT, Perplexity and Google AI Mode for the best business of your type in your part of town, and write down every business named and every site cited. Those cited lists and directories are next month's outreach list.

If you want a second pair of eyes on those three checks, the free 15-minute audit on our AI SEO page starts with exactly this list.

Drafted with AI assistance, researched, edited, and fact-checked by Elias Musleh on October 19, 2026.