How to Get Your Business Recommended by ChatGPT

We asked ChatGPT to recommend a web development firm in Calgary. It gave us three names. Two of them don't exist. The third is a marketing agency that doesn't build websites.

That's the state of AI-powered business recommendations right now. Large language models like Claude, ChatGPT, Gemini, and Perplexity are already answering the questions your customers used to type into Google. "Best accountant in Calgary." "Who can build me a WordPress site in Alberta." "What company does WCAG audits near me." And the answers they give are a mix of real data, training-set memory, and confident guessing.

Your job is to make sure the real data wins.

This isn't a future problem. According to OpenAI's own documentation, ChatGPT search now partners with search providers including Bing to surface live web results directly inside conversations. Google's AI Overviews appear on over 13% of all searches as of early 2025, and that number keeps climbing. When someone asks an AI assistant about your industry, you're either part of the answer or you don't exist.

How AI Assistants Actually Find Businesses

There's a common misconception that LLMs only know what was in their training data. That was true in 2023. It's not true now.

Modern AI assistants pull from three distinct sources:

Training data — the massive text corpus the model was built on. This is months or years old and includes whatever the web said about your business at crawl time. You can't directly control this, but you can influence what future training runs find.

Live search results — ChatGPT uses Bing's Web Search API. When a user asks a question that needs current information, ChatGPT rewrites that query into one or more targeted searches and pulls a structured JSON payload back from Bing: URLs, titles, snippets, ranking positions. Perplexity does something similar with its own crawlers. Google's AI Overviews pull from Google's search index. The common thread: if you don't rank in traditional search, you probably won't appear in AI answers either.

Structured data from indexes — this is where it gets interesting. Both Google and Microsoft publicly confirmed in 2025 that they use schema markup for their generative AI features. Google was explicit: structured data is efficient, precise, and easy for machines to process. When an AI system builds an answer about local businesses, it isn't just reading your About page. It's pulling from indexed schema markup to understand what your business is, where it operates, what services it offers, and how those services relate to the user's question.

That third source is where most businesses have zero presence. And it's the one you have the most control over.

Entity Establishment: Teaching AI What You Are

AI systems don't think in keywords. They think in entities. An entity is a defined thing — a person, a business, a service, a location — with properties and relationships. Google's Knowledge Graph has been mapping these relationships for over a decade. AI systems now rely on that same graph to ground their answers in verified facts.

When your business has a well-established entity, AI systems can reference you with confidence. When it doesn't, they guess. And guessing is how you end up as "a Calgary-based web design firm" that also apparently offers pet grooming.

Establishing your entity means building a consistent, connected web of information about your business across multiple signals. Here's what matters:

Your Google Business Profile is the anchor. If you don't have one, stop reading and go set it up. If you have one and the information doesn't match what's on your website, fix that first. Name, address, phone number — these need to be identical everywhere. Not "close enough." Identical. AI systems cross-reference your GBP with your website schema, your social profiles, and directory listings. Inconsistencies create doubt, and doubtful entities don't get recommended.

Schema markup is how you tell machines what you are in their language. We covered this in depth in our structured data guide for local businesses, but the short version: JSON-LD markup in your page's <head> tells search engines and AI crawlers your business type, location, services, operating hours, and how you relate to other entities. Without it, machines have to infer all of that from your prose — and inference is where errors creep in.

Consistent mentions across the web reinforce your entity. Citations in business directories, mentions in local news, backlinks from industry-relevant sites. These aren't just SEO signals anymore. They're corroboration that AI systems use to decide whether your entity is real, active, and authoritative enough to mention.

The businesses that show up in AI recommendations aren't necessarily the biggest or the best. They're the ones whose identity is clearest to a machine.

Schema Markup: The Part Most Businesses Skip

We audit local business websites regularly. The pattern is almost always the same: decent content, reasonable SEO, and either zero schema markup or whatever their WordPress theme auto-generated — which is usually a bare-bones Organization schema missing half the required fields.

Here's what a proper schema setup looks like for a local business that wants AI visibility:

LocalBusiness (or a specific subtype) — Schema.org offers dozens of business types. A dental clinic should use Dentist, not the generic LocalBusiness. An HVAC company should use HVACBusiness. The more specific your type, the clearer the signal. Include your full address, geo coordinates, phone, email, opening hours, price range, and area served. Every field that matches your Google Business Profile strengthens the connection between your web entity and your GBP entity.

Organization — connects your business to its founders, logo, social profiles, and contact information. This is where you establish the "who" behind the business.

Service and hasOfferCatalog — list every service you provide with descriptions. This is how AI systems know whether to recommend you for "WordPress development" versus "email marketing" versus "accessibility audits." If your services aren't in your schema, the AI has to guess what you do based on your page copy. We've seen businesses that offer six distinct services but only show up in AI results for one — because the others weren't structured.

FAQPage — write questions and answers the way a human would ask an AI assistant. Not "What are our values?" but "What does a web developer in Calgary charge?" or "How long does a website redesign take?" These get pulled directly into featured snippets and AI Overviews. Write the answers as if someone will read them aloud — because an AI assistant might do exactly that.

There's an important caveat here. A study by SearchVIU in October 2025 tested whether AI chatbots actually read JSON-LD schema during direct page retrieval. The answer: they don't. Claude, ChatGPT, Perplexity, and Gemini all extracted visible HTML content only — information that existed solely in JSON-LD was invisible to them during live browsing.

But that's only half the story. Schema markup gets indexed by Bing and Google during their regular crawling. When ChatGPT later queries Bing's search API, it receives structured data that Bing extracted during indexing — including schema markup. So schema doesn't help during a direct page fetch, but it absolutely helps during search-mediated AI interactions, which is how most AI recommendations actually work.

The practical takeaway: put your most important information in visible HTML on the page AND reinforce it with schema markup. Don't hide anything in schema alone.

Content Strategy for AI Visibility

Schema tells machines what you are. Content tells them what you know.

AI assistants prefer sources that demonstrate topical authority — deep, specific knowledge about a subject, not surface-level overviews. A single page that says "We do web development, SEO, and accessibility" gives an AI nothing to work with. Twenty pages that each address specific questions within those disciplines — that's a body of evidence.

Write content that directly answers questions people ask AI assistants. Not "What is SEO?" — that market is saturated and you're competing with Wikipedia. Instead: "How much does an SEO audit cost in Calgary?" or "What structured data does a restaurant need?" or "Why is my WordPress site slow after installing Elementor?" Specific questions with specific answers, grounded in real experience.

Structure that content with clean semantic HTML. Proper heading hierarchy. Clear topic sentences at the start of each section. AI crawlers and search indexers process your HTML structure, not your visual design. A beautifully designed page with headings implemented as styled <div> tags instead of actual <h2> elements is invisible to machines in the ways that matter.

Use <article> tags. Use descriptive <title> elements. Write meta descriptions that read like something an AI assistant could quote directly — factual, specific, and location-aware. We've written more about this approach in our AI search optimization guide, which covers the full content architecture strategy.

And publish consistently. AI systems weight recency. A site with its most recent blog post from 2022 signals abandonment. A site publishing useful content monthly signals an active, authoritative entity. ChatGPT's search function explicitly uses recency filters when looking for current information.

Let AI Crawlers In

This one is surprisingly easy to get wrong. Your robots.txt file controls which bots can access your site. Many WordPress security plugins and hosting providers block unknown user agents by default — and some of those "unknown" agents are the very AI crawlers you want reading your content.

OpenAI uses three crawlers: GPTBot for model training, OAI-SearchBot for search functionality, and ChatGPT-User for direct user-initiated requests. Each is independently controllable. You can block GPTBot (if you don't want your content in training data) while allowing OAI-SearchBot (so you appear in ChatGPT search results). That distinction matters.

Other AI crawlers to explicitly allow:

• ClaudeBot (Anthropic / Claude)
• PerplexityBot (Perplexity)
• Google-Extended (Gemini)
• Applebot (Apple Intelligence / Siri)
• Bingbot (Microsoft Copilot — and, by extension, ChatGPT search)

Check your robots.txt right now. If it has a blanket Disallow: / for unknown bots, or if it doesn't mention these agents at all, you might be invisible to AI assistants regardless of how good your content and schema are.

What Doesn't Work (and What Will Get You Ignored)

Keyword stuffing doesn't fool AI. This should be obvious, but we still see it. Cramming "best Calgary web developer" into every paragraph doesn't make an LLM more likely to recommend you — it makes your content read as low quality, which means AI systems are less likely to cite you as a source.

Buying backlinks from link farms. AI systems are built on the same indexes as traditional search engines, and those indexes have been penalising link manipulation for years. A link from a Calgary Chamber of Commerce page is worth a thousand links from a random blog network.

Generating AI content to feed back to AI. This circular approach produces the kind of generic, hedged, keyword-adjacent content that LLMs already have in abundance. They don't need more of it. They need specific, authoritative, experience-backed content that adds something their training data doesn't already contain.

Creating dozens of thin location pages. "Web Development in Airdrie." "Web Development in Cochrane." "Web Development in Okotoks." Each with 200 words of near-identical copy. Search engines flagged this tactic years ago. AI systems ignore it entirely because there's no substantive content to reference.

If your content wouldn't be worth reading without AI in the picture, it's not worth creating for AI either. Machines are getting better at identifying genuine expertise — and that's the one thing you can't fake.

Bing Matters More Than You Think

Here's something most businesses haven't caught up with: ChatGPT search runs on Bing's infrastructure. When ChatGPT needs to answer a question with current information, it queries Bing's Web Search API. Bing powers Yahoo, DuckDuckGo, Microsoft Start, and several AI tools beyond just ChatGPT.

Most SEO effort goes into Google. That's reasonable — Google still dominates traditional search. But if you're specifically trying to get recommended by AI assistants, Bing's index matters. Submit your site to Bing Webmaster Tools. Make sure Bing has crawled your pages. Check that your structured data renders correctly in Bing's tools, not just Google's Rich Results Test.

This isn't about choosing sides. It's about covering the infrastructure that actually feeds AI recommendations.

The Minimum Viable Checklist

Where This Goes From Here

We're still in the early innings. AI recommendations are messy, inconsistent, and sometimes flat-out wrong. The models are getting better fast. Six months from now, the bar for appearing in AI answers will be higher than it is today.

The businesses that start building their machine-readable identity now — proper schema, clean structured content, consistent entity signals — are the ones that will already be established when the models improve. Everyone else will be playing catch-up against entities that AI systems already trust.

If you're not sure where your business stands right now, we do exactly this kind of AI search readiness assessment. And if you want to go deeper on the technical side, our guides on structured data for local businesses and AI search optimization cover the implementation details.

The thing to remember is this: AI assistants don't recommend businesses. They recommend entities. Make sure yours is one.

Sources

• OpenAI — ChatGPT Search Documentation
• OpenAI — Overview of OpenAI Crawlers
• Google Search Central — Intro to Structured Data
• Google Search Central — Local Business Structured Data
• Schema.org — LocalBusiness Type
• SearchVIU — Schema Markup and AI in 2025: What Claude, ChatGPT, Perplexity & Gemini Really See
• Schema App — What 2025 Revealed About AI Search and the Future of Schema Markup
• Search Engine Land — Schema and AI Overviews: Does Structured Data Improve Visibility?
• Search Engine Journal — Structured Data's Role in AI and AI Search Visibility