How ChatGPT Decides Which Businesses to Recommend

When someone asks ChatGPT for a recommendation, it does not pull a name from a ranked list the way Google returns ten blue links. It assembles an answer from patterns in its training data and, for current questions, from live web results. Then it favors businesses that are mentioned consistently across trustworthy sources, described clearly, and backed by third-party signals like reviews and directory listings. Understanding how ChatGPT recommends businesses comes down to one idea: the model recommends what it can confidently recognize and verify. In our October 2026 test of 120 AI answers across eight cities, ChatGPT named 28.8% of the businesses in Google's top 20 map results that had 500 or more reviews and 2.6% of those with fewer than 50, and it was the only one of five AI engines to cite the Better Business Bureau (bbb.org appeared in 25% of its answers). The rest of this guide explains where those answers come from and how to influence them honestly.

No one outside OpenAI knows the exact internal weighting, and anyone who claims to is guessing. So everything below is framed as how AI engines generally behave, based on observable patterns and published vendor documentation, rather than a peek inside the model.

How ChatGPT sources its answers

ChatGPT draws on a few different sources depending on the question, and each one rewards different things.

Training data. A model is trained on a large snapshot of text from the web and other sources up to a cutoff date. If your business was written about in articles, forums, reviews, and directories before that cutoff, the model may already "know" you exist and what you do. This is slow to change. You cannot edit training data after the fact, which is why long-term brand presence matters more here than any quick fix.

Live web browsing and search. For timely or specific questions ("moving companies that can do a cross-country move next month"), ChatGPT can retrieve live results and summarize them, listing the pages it used. OpenAI has stated this retrieval runs through a search layer, and at various points that has reportedly included Bing-sourced results. The practical takeaway is stable regardless: pages that rank well and read clearly are easier to retrieve and quote.

Retrieval and grounding. When the model browses, it grounds its answer in the specific passages it pulls back. It tends to lift sentences that already read like a direct answer, then attribute them. Content buried under long introductions is harder to ground against, so it gets quoted less often.

The important nuance: a recommendation can come from training data, live retrieval, or both at once. That is why two people can ask the same question and get slightly different names, and why a business with no live ranking can still surface if it built a strong reputation before the cutoff.

What ChatGPT actually read in our test

In October 2026 we asked ChatGPT (gpt-5.5, with web search on) "Who is the best {moving company | wedding photographer | solar installer} in {city}? Give me your top picks." in eight cities, and put the same question to Perplexity, Gemini, Claude and Google AI Mode. The full results are in our AI recommendation study. All 24 of ChatGPT's answers searched the web and cited their sources, with 4.9 source links per answer on average, the fewest of the five engines.

The sites ChatGPT read were mostly third parties that judge businesses: review platforms such as ConsumerAffairs, moveBuddha, EnergySage and The Knot, plus Reddit threads and Better Business Bureau listings. It cited bbb.org in 6 of its 24 answers, and none of the other four AIs cited it once. If your BBB profile, your Reddit mentions and your listing on your industry's main review platform are thin or out of date, you are missing from the pages ChatGPT reads before it picks.

You can see this in how ChatGPT explains its choices. For movers in Charlotte it opened with "Here are my top picks for moving companies in Charlotte, NC based on current 2026 roundups, BBB listings, customer-review signals, and local reputation." For a solar installer in the same city it wrote: "That combination of local presence, longevity, and review volume makes it my safest “first call” recommendation."

Here is how ChatGPT compared with the other four engines on the same 24 questions.

What makes a business "citable"

AI engines lean toward sources they can recognize and verify. A handful of factors tend to make a business easier to recommend.

• Consistent business information. Your name, address, phone number, and service area should match everywhere they appear: your site, Google Business Profile, directories, and review platforms. Conflicting details make a business harder to identify with confidence.
• Brand mentions across many sources. A name that appears in independent articles, forum threads, review sites, and directories reads as more established than one that only appears on its own website. Breadth of mention is a recognition signal.
• Structured, scannable content. Short paragraphs, clear headings, lists, and direct answers are easier to extract and quote than dense prose. Structured data (Organization, LocalBusiness, Service, FAQPage schema) helps machines understand what a page is and who it is about.
• Third-party authority. Reviews, ratings, mentions in reputable publications, and listings in respected directories act as outside validation. A business the wider web vouches for is safer to recommend than one that only vouches for itself.
• Topical depth. Covering your services thoroughly, with a connected set of genuinely useful pages rather than one thin page, signals expertise the model can draw on.
• Clarity about who you serve and where. Stating your service areas and specialties plainly helps the model match you to the right question instead of guessing. A wedding photographer who says "documentary-style weddings and elopements, will travel" is easier to match than one who says "capturing your special day".

None of these are tricks. They are the same signals a careful human would use to decide whether a business is real, relevant, and worth mentioning.

What our data says about reviews

Third-party authority showed up clearly when we matched ChatGPT's picks against the Google Maps top 20 for the same city and industry.

Review volume tracked ChatGPT's choices more closely than map position did. Of the 73 map businesses with 500 or more Google reviews, ChatGPT named 28.8%. Of the 196 with fewer than 50, it named 2.6%. The map businesses it named had a median of 435 Google reviews, against 64 for the ones it skipped, while their average star ratings were almost identical (4.78 and 4.81). Map position mattered less: ChatGPT named 13.9% of the businesses in the top three map spots and 7.9% of those ranked 11 to 20.

For an owner, the lesson is to make review volume a weekly habit. Ask every finished customer for a review on Google and on your industry's main platform, because ChatGPT repeats review counts in its answers as evidence ("citing 248 reviews" in Denver, "4.5 stars with 1,141 reviews" in Phoenix). Pushing a 4.8 average to 4.9 on a small base is unlikely to change much, since named and skipped businesses had the same ratings.

Why it sometimes recommends competitors or out-of-area providers

Business owners are often surprised when ChatGPT names a competitor, a national chain, or a provider in the wrong city. A few honest explanations:

• More and clearer signals. If a competitor is mentioned in more places, has more reviews, or has clearer service pages, the model has more reason to recognize them. Citation often follows visibility, not quality of work.
• Training data lag. The model may be working from an older snapshot. A newer or rebranded business can be invisible simply because it was not yet well documented at the cutoff.
• Weak location signals. If your service area is not stated clearly and consistently, the model may default to bigger, better-documented national brands or providers in a nearby major city.
• Ambiguous or generic queries. Broad questions ("good moving company") pull broad, well-known answers. The more specific the query, the more a well-optimized smaller business can surface.
• Retrieval gaps. If your pages are hard to crawl, slow, or thin, live retrieval may skip them in favor of competitors that are easier to read and quote.

The pattern across all of these is the same: the model recommends what it can recognize and verify, and competitors sometimes simply give it more to work with.

Our test adds a number to the "wrong business" problem: 72.8% of ChatGPT's picks were not in the Google Maps top 20 for the city at all. ChatGPT builds its shortlist from review platforms and ranking pages, so a business that ranks well on the map but is missing from those pages can be passed over for one that ranks lower on the map and appears on every list.

How often ChatGPT agrees with the other AIs

Of the 151 businesses ChatGPT recommended, 61.6% were also named by at least one other AI for the same city and industry, the highest share of any engine we tested. Claude overlapped most (46.4% of ChatGPT's picks), partly because Claude gives the longest lists at 10.4 businesses per answer. Google AI Mode overlapped least, at 18.5%.

Agreement depends heavily on the industry. In moving and solar, where a few review platforms dominate, ChatGPT's picks were widely shared. In wedding photography, where style and taste vary, most of its picks were its own.

If you are a wedding photographer, being named by ChatGPT and being named by Gemini or Perplexity are close to separate contests, so check each engine on its own. If you are a mover or a solar installer, the engines lean on the same handful of review platforms, so a stronger profile on those platforms is likely to show up in several engines at once.

→ Quick check: when a customer asks ChatGPT who to hire, is your business the one AI names? Find out free. Get my free GEO audit.

Look up every business ChatGPT recommended

The table below lists all 151 recommendations from ChatGPT's 24 answers, with the position ChatGPT gave each business, how many of the other four AIs named it too, and its Google Maps rank and review count where it appeared in the map top 20. Filter by industry or search for your city to see who ChatGPT named in your market and what those businesses have in common.

A blank Maps rank means the business was outside Google's top 20 map results for a search from the city center, which was true for 110 of the 151 picks. That gap is the clearest sign that ChatGPT's shortlist comes from the web pages it reads rather than from the map.

What business owners can do

You cannot edit a model's training data, but you can shape almost every signal it relies on. This is the core of generative engine optimization (GEO), also called AI SEO or AEO, and it overlaps heavily with good SEO.

• Fix consistency first. Audit your name, address, phone, and service area everywhere they appear and make them identical. This is the cheapest, highest-leverage step.
• Earn genuine mentions. Get listed in reputable directories, encourage honest customer reviews, and pursue legitimate coverage and guest contributions in places your industry actually reads. Start with the sites ChatGPT cited most in our test: moveBuddha, ConsumerAffairs, BBB and Reddit for movers, The Knot and WeddingWire for wedding photographers, and EnergySage for solar installers.
• Restructure your key pages. Lead with a direct answer, use short paragraphs and headings, and add specific, true details. Our guide on how to write content AI will cite walks through this in depth.
• Add structured data. Implement Organization, LocalBusiness, Service, and FAQPage schema so machines can parse who you are and what you offer. Our free schema generator handles the common types.
• State your service area clearly. Make it unmistakable on the page, not just implied.
• Keep content current. Update important pages so they reflect this year, not three years ago. Freshness helps with live retrieval.
• Make sure AI crawlers can reach you. Check your robots.txt does not block AI user agents, and that your pages load quickly and render their content. The AI crawler checker tests this for free.
• Strengthen your Google Business Profile. It is one of the most widely read sources of local business data. Our guide on Google Business Profile and AI recommendations covers what to fix.

For the full step-by-step version aimed at ChatGPT specifically, see our guide on how to optimize for ChatGPT search. Most of this work compounds. The same signals that help ChatGPT recommend you also help you rank in Google and other AI engines.

Free GEO audit

Be the business AI recommends

Buyers now ask AI for a shortlist before they visit a single website. Find out whether your business is on it, and how to get there first.

Tested across ChatGPT · Perplexity · Google AI Overviews · Gemini

Get my free GEO audit

The short version: signals that drive recommendations

If you only remember one list, make it this one. AI engines tend to recommend businesses with:

• Consistent core information across the web (name, address, phone, service area).
• Frequent, independent brand mentions in articles, forums, directories, and reviews.
• Strong third-party validation through reviews and reputable listings.
• Clear, structured, scannable pages that answer questions directly.
• Schema markup that explains what each page is.
• Topical depth across a connected set of useful pages.
• Unambiguous location and specialty signals.
• Fresh content and pages that AI crawlers can actually access.

How we got this data

On 7 October 2026 we asked ChatGPT (gpt-5.5), Perplexity (sonar-pro), Gemini (gemini-3.8-flash) and Claude (claude-sonnet-5-5), all with web search on, the same question: "Who is the best {moving company | wedding photographer | solar installer} in {city}? Give me your top picks." We asked it in Denver, Austin, Phoenix, Charlotte, Seattle, Tampa, London and Toronto, and ran the matching "best {service} in {city}" search in Google AI Mode, for 120 answers in total, 24 of them from ChatGPT. We matched every recommended business against the Google Maps top 20 for the same industry and city (480 listings) and classified every cited website by type. Each prompt ran once per engine, so this is a snapshot: AI answers vary from run to run, and the Maps top 20 reflects a search from the city center. The full method and results are in our AI recommendation study.

Frequently Asked Questions

How does ChatGPT decide which businesses to recommend?

It combines what it learned during training with live web results for current questions, then favors businesses it can recognize and verify. Consistent information, frequent independent mentions, third-party reviews, and clear, structured pages all make a business easier to recommend. The exact internal weighting is not public, so treat any specific percentage you see as an estimate, not a fact.

Why does ChatGPT recommend my competitor instead of me?

Usually because the competitor gives the model more to work with: more mentions, more reviews, clearer pages, or stronger location signals. It can also reflect training data lag if your business is newer or recently rebranded. Closing the gap is rarely about doing better work and more about being documented as clearly and widely as they are.

Can I pay to be recommended by ChatGPT?

There is no paid slot inside organic ChatGPT recommendations. You influence recommendations indirectly, by improving the real-world signals the model reads: reviews, mentions, consistent information, and clear content. Be wary of anyone promising guaranteed placement. That is not how AI engines work.

How can I check whether ChatGPT recommends my business?

Ask it the questions your customers ask, in their words, and note which businesses it names. Repeat the same set of prompts every month so you can see change over time. Our guide on how to track ChatGPT citations shows how to set up a repeatable prompt panel.

Conclusion

How ChatGPT recommends businesses is less mysterious than it first appears. The model is trying to name businesses it can confidently recognize and verify, and it leans on the same signals a careful person would: consistent details, independent mentions, genuine reviews, and clear pages. You cannot rewrite a model's training data, but you can make your business the easiest one to recognize and the safest one to recommend.

If you want help getting recommended by ChatGPT and other AI engines, take a look at our generative engine optimization services. We focus on exactly these signals for moving companies, wedding businesses, and solar installers.