What ChatGPT, Perplexity, and Google AI Overviews Actually Look At Before Recommending a Business

---

title: "What ChatGPT, Perplexity, and Google AI Overviews Actually Look At Before Recommending a Business" slug: how-ai-recommends-businesses date: 2026-01-10 category: AI Visibility author: Formula Won Labs image: /blog/how-ai-recommends-businesses.jpg summary: "When someone asks ChatGPT for a plumber or Perplexity for a dentist, these systems pull from specific data sources. Here is what they look at and what makes your business citable." tags: [ai-visibility, chatgpt, perplexity, google-ai-overviews, structured-data] pillar: ai-visibility status: published externalLinks:

• label: "OpenAI - ChatGPT with Browsing" url: "https://openai.com/index/chatgpt-browsing/"
• label: "Google AI Overviews and Search" url: "https://blog.google/products/search/generative-ai-google-search-may-2024/"
• label: "Schema.org LocalBusiness" url: "https://schema.org/LocalBusiness" internalLinks:
• label: "AI Visibility for Plumbers" url: "/for/plumbers/ai-visibility"
• label: "AI Search Visibility in Denver" url: "/services/ai-search-visibility/denver-co"
• label: "AI Visibility for Med Spas" url: "/for/med-spas/ai-visibility"
• label: "AI Search Visibility in Atlanta" url: "/services/ai-search-visibility/atlanta-ga"
• label: "AI Visibility for Dentists" url: "/for/dentists/ai-visibility" faqs:
• question: "Does ChatGPT always give the same recommendation for a local business query?" answer: "No. ChatGPT responses vary based on the specific phrasing of the query, the browsing results it retrieves, and inherent variability in how large language models generate responses. You may see different businesses recommended if you ask the same question multiple times. This is different from Google, where rankings are more deterministic."
• question: "Can I block AI systems from reading my website?" answer: "You can block specific AI crawlers using your robots.txt file. However, doing so means those AI systems will not recommend your business, which is increasingly a competitive disadvantage. For local service businesses, visibility to AI systems is almost always beneficial."
• question: "Is Bing Places really important now?" answer: "More than it used to be. ChatGPT uses Bing for web browsing, which means your Bing Places listing affects whether ChatGPT finds and recommends your business. Claiming and completing your Bing Places profile takes about 15 minutes and mirrors much of what you already have on Google Business Profile."
• question: "How do I add structured data to my website?" answer: "Structured data is JSON-LD code added to the HTML of your web pages. For local businesses, the most impactful types are LocalBusiness schema, Service schema, and FAQPage schema. The key is ensuring the markup is accurate, complete, and matches the visible content on your pages."

---

Ask ChatGPT "Who is the best plumber in Denver?" and it will give you an answer. It will name specific businesses, explain why it recommends them, and sometimes include addresses, phone numbers, and ratings.

Ask Perplexity the same question and you will get a different answer, pulled from different sources, with footnoted citations.

Ask Google and, depending on the query, an AI Overview will synthesize information from across Google's index and present a summary above the traditional search results.

Three different systems. Three different answers. Three different data pipelines. If you are a local business owner, understanding exactly what each system looks at, and what makes it include one business over another, is the difference between getting recommended and getting skipped.

Google AI Overviews: the data it pulls from

Google AI Overviews have a significant advantage over every other AI search system: they pull from Google's own proprietary data. This includes sources that no other AI system can access directly.

Google Business Profile data

Your GBP is the primary structured data source for Google AI Overviews when it comes to local queries. The system reads:

• Primary and secondary categories. If someone searches "emergency plumber near me" and your primary category is "Plumber" with a secondary category of "Emergency Plumbing Service," you are significantly more likely to appear than a business categorized only as "General Contractor."
• Services listed in your profile. Google now allows businesses to list specific services with descriptions. AI Overviews reference these directly when matching a query to a business.
• Business attributes. Hours, payment methods, accessibility features, service options (like "offers online estimates"). These attributes help Google answer specific sub-queries within a broader search.
• Photos and visual content. Google's AI can analyze photos on your profile. Businesses with photos of actual work (completed roofing jobs, finished dental work, installed HVAC systems) provide visual evidence that supplements text data.

Google Reviews

Reviews are not just a ranking signal for AI Overviews. They are a content source. Google's AI reads the text of your reviews and uses them to generate summaries.

When a review says "They fixed our AC on a Sunday afternoon and the tech explained everything clearly," Google AI can use that information to recommend your business for emergency HVAC service and note that customers praise your communication. This is why review content matters, not just star ratings.

The recency, volume, and specificity of reviews all factor in. A business with 200 vague "Great service!" reviews gives the AI less to work with than a business with 80 detailed reviews that mention specific services, locations, and experiences.

Your website

Google AI Overviews crawl and reference your website content. They look at:

• Service pages. Dedicated pages for each service you offer give the AI specific content to reference. For plumbing companies working on AI visibility, having separate pages for drain cleaning, water heater installation, and pipe repair means the AI has three distinct sources it can cite for three different query types.
• Location pages. If you serve multiple cities, pages specific to each service area help Google's AI connect your business to location-specific queries.
• Structured data markup. Schema.org LocalBusiness markup, Service schema, and FAQ schema provide machine-readable information that the AI can parse directly rather than inferring from page content.

The Google Maps knowledge graph

Behind the scenes, Google maintains a knowledge graph that connects businesses to locations, categories, competitors, and related entities. This is not something you can directly edit, but your GBP data, website content, and directory citations all feed into it. When AI Overviews generate local recommendations, they draw from this graph to understand relationships between businesses, services, and geographic areas.

ChatGPT: where it gets its local data

ChatGPT's approach to local business recommendations is fundamentally different from Google's. It does not have access to Google's proprietary data. Instead, it relies on a combination of web browsing, the Bing search index, and its training data.

Bing search index

When ChatGPT browses the web to answer a query, it uses Bing as its primary search engine. This means your visibility on Bing directly affects whether ChatGPT finds and recommends you.

Most local businesses have never thought about Bing. They focus entirely on Google. But with ChatGPT now handling search queries, your Bing presence matters. This includes your Bing Places listing (Microsoft's equivalent of Google Business Profile), your website's indexing in Bing, and your presence on sites that Bing ranks highly.

Website content

ChatGPT reads your website. When it browses to answer a local query, it can land on your homepage, service pages, about page, or blog posts. What it finds there determines whether it recommends you.

The difference between websites that ChatGPT cites and websites it ignores comes down to specificity and structure. Consider two HVAC company websites:

Website A: A homepage that says "We are a full-service HVAC company serving the Denver metro area. Call us today for a free estimate."

Website B: Dedicated pages for AC installation, furnace repair, duct cleaning, and heat pump service. Each page includes the specific service area (Denver, Aurora, Lakewood, Arvada), pricing guidance, the process a customer can expect, and answers to common questions.

ChatGPT can extract useful, quotable information from Website B. It has nothing to work with from Website A. For businesses in Denver trying to appear in AI search results, the content depth of your website is the deciding factor.

Review aggregator sites

ChatGPT frequently references Yelp, BBB, Angi, and other review aggregators when answering local business queries. These sites rank well in Bing, and their structured review data is easy for ChatGPT to parse and summarize.

This means your presence on review platforms matters beyond the reviews themselves. If Yelp has a complete, accurate listing for your business with recent reviews, ChatGPT may cite that listing as a source when recommending businesses in your category.

Training data

ChatGPT's training data includes a snapshot of the internet. Businesses that had a strong web presence when the training data was collected may appear in ChatGPT's responses even without real-time browsing. However, this data has a cutoff date and becomes stale over time. Real-time browsing is where current recommendations come from.

Perplexity: the citation engine

Perplexity operates differently from both Google and ChatGPT. It is built specifically as a research and answer tool, and its defining feature is that it cites every claim with a linked source.

Real-time web crawling

Perplexity crawls the web in real time for every query. It does not rely primarily on a pre-built index like Google or Bing. When someone asks "best med spa in Atlanta," Perplexity sends out crawlers that read current web content, aggregate information, and generate a sourced answer.

This real-time approach means your current web presence matters more than historical presence. A website updated last week is more likely to be crawled and cited than one last updated two years ago.

Source selection criteria

Perplexity does not just find information. It evaluates sources for credibility and relevance. Based on observable patterns in its recommendations, Perplexity appears to prioritize:

• Authoritative domain sources. Business websites with detailed content, review platforms with substantial review counts, and industry directories with editorial standards.
• Specificity of content. Pages that directly answer the query with specific information are cited more often than general pages that tangentially relate.
• Structured information. Pages with clear headings, structured data, and organized content are easier for Perplexity to parse and cite accurately.

For med spas building AI visibility, this means your service pages need to be specific enough that Perplexity can extract and cite individual claims. "We offer Botox, fillers, and chemical peels" is too thin. A page that covers what each treatment involves, who it is for, expected results, and recovery time gives Perplexity citable content.

Citation link behavior

When Perplexity cites your business, it links to the specific page it pulled the information from. This means every service page, every location page, and every informational page on your site is a potential citation target.

This is a fundamentally different model from traditional search, where your homepage is often the primary landing page. With Perplexity, your deep content pages are the ones that get cited and linked.

What makes a business "citable" across all three systems

Despite their differences, all three AI systems converge on a set of characteristics that make a business more likely to be recommended. These are the common signals.

Specificity over generality

Every AI system struggles with vague content. "We provide quality service" tells an AI nothing. "We install Carrier and Trane central air conditioning systems for homes in the Denver metro area, typically completing installation in one day" tells the AI exactly what you do, what brands you work with, where you operate, and how long it takes.

AI systems recommend businesses they can make specific claims about. If your web presence does not give them specific claims to make, they will recommend a competitor whose presence does.

Consistency across sources

When Google sees one set of information on your GBP, a different set on your website, and a third set on Yelp, its confidence drops. When ChatGPT browses multiple sources and finds conflicting information about your business, it is less likely to make a definitive recommendation.

Consistency does not mean identical content everywhere. It means your business name, address, phone number, service descriptions, and service areas should align across all platforms. Contradictions create doubt for AI systems just as they do for human customers.

Structured data as a competitive edge

Schema.org markup is the technical layer that most local businesses ignore entirely. LocalBusiness schema tells AI systems your business type, address, hours, and service area in a machine-readable format. Service schema describes your individual services with structured properties. FAQ schema provides question-and-answer pairs that AI can use directly.

For businesses in competitive markets like Atlanta, structured data is often the differentiator. When two businesses have similar review profiles and similar website content, the one with proper schema markup gives AI systems an easier path to recommendation.

Most local business websites have zero structured data. Adding it is not difficult, but it requires someone who understands the technical implementation. The payoff is disproportionate to the effort.

Review depth and recency

All three systems use reviews as evidence. Google reads Google Reviews directly. ChatGPT reads reviews on Yelp, Google (via Bing), and other aggregators. Perplexity cites review platforms as sources.

The reviews that help most are recent, detailed, and mention specific services. A review that says "John from ABC Plumbing fixed our slab leak in Buckhead. He was here within an hour and explained the repair options clearly" gives AI systems specific, quotable information about your service area, response time, and customer communication.

Generic five-star reviews with no text provide almost no value to AI recommendation systems.

The practical framework for AI citability

Based on how these three systems actually work, here is the priority order for making your business citable:

Layer 1: Data foundation. Complete your Google Business Profile with accurate categories, services, and attributes. Claim your Bing Places listing. Ensure your website has accurate, consistent business information.

Layer 2: Content depth. Create dedicated service pages with specific, detailed content for each service you offer. Build location-specific pages for each city or area you serve. Write content that an AI system could quote directly as a recommendation.

Layer 3: Technical structure. Add LocalBusiness, Service, and FAQ schema markup to your website. Ensure your site is crawlable by all major search engines and AI crawlers. Check that your robots.txt does not block AI crawler user agents.

Layer 4: Validation signals. Build a consistent flow of recent, detailed Google reviews. Maintain accurate listings on major review platforms and directories. Earn mentions and citations from authoritative local sources.

For dentists navigating AI visibility, this framework applies directly. Your GBP categories and services, your website's procedure pages, your schema markup, and your review velocity all feed into whether AI systems recommend your practice over the one down the street.

Each layer builds on the previous one. Structured data without good content underneath it has limited value. Great content without a solid data foundation can get lost. The businesses that appear consistently across all three AI systems are the ones that have built all four layers.

Frequently asked questions

Does ChatGPT always give the same recommendation for a local business query?

No. ChatGPT's responses vary based on the specific phrasing of the query, the browsing results it retrieves in that session, and some inherent variability in how large language models generate responses. You may see different businesses recommended if you ask the same question multiple times. This is different from Google, where rankings are more deterministic.

Can I block AI systems from reading my website?

You can block specific AI crawlers using your robots.txt file. However, doing so means those AI systems will not recommend your business, which is increasingly a competitive disadvantage. Blocking AI crawlers is only advisable if you have a specific reason, such as protecting proprietary content. For local service businesses, visibility to AI systems is almost always beneficial.

Is Bing Places really important now?

More than it used to be. ChatGPT uses Bing for web browsing, which means your Bing Places listing affects whether ChatGPT finds and recommends your business. Claiming and completing your Bing Places profile takes about 15 minutes and mirrors much of what you already have on Google Business Profile. Given that ChatGPT handles hundreds of millions of queries, this is a low-effort, high-potential return action.

How do I add structured data to my website?

Structured data is JSON-LD code added to the HTML of your web pages. For local businesses, the most impactful types are LocalBusiness schema (your business details), Service schema (your individual services), and FAQPage schema (common questions and answers). If you use WordPress, plugins like Rank Math or Yoast can generate some of this automatically. For custom sites, a developer can implement it from Schema.org documentation. The key is ensuring the markup is accurate, complete, and matches the visible content on your pages.