Why ChatGPT Cites Some Local Businesses 2.4× More Than Google Does

AI Search Hasn’t Killed SEO — It’s Reweighted It

Here’s a fact that should change how every local business thinks about being found online: pages with structured data are cited by ChatGPT at more than twice the rate they appear at the top of traditional Google results. AirOps’s 2025 analysis of more than 15,000 URLs found that 61% of pages cited by ChatGPT included schema markup, compared to just 25% of Google’s top traditional search results.

That’s not a small shift in preference. That’s a structural reweighting of how the tools your customers actually use are deciding who gets recommended.

SEO Foundations Still Matter — They Just Got Rebalanced

Let’s get one thing out of the way first: AI search hasn’t replaced SEO. ChatGPT, Gemini, Perplexity, and Google’s AI Overviews all rely on the same web your traditional SEO has been optimizing for. They crawl pages, index content, and weigh authority signals like backlinks, brand mentions, and content quality. ChatGPT Search runs against Bing’s index. Perplexity does live web searches. The fundamentals — a real website, actual content, legitimate backlinks, and genuine authority in your niche — are still the price of entry. Skip them and AI engines won’t find you any more reliably than Google would.

What’s changed is the relative weighting of one specific input: structured data, also known as schema markup.

For Google, schema has always been a useful bonus. It powers rich results, star ratings, FAQ accordions, and the local map pack. Roughly 72% of pages on the first page of Google use some form of schema, per Backlinko, and Schema App’s research has shown that pages with schema enjoy about 40% higher click-through rates. But for traditional Google, you can rank perfectly well without it. Plenty of authority sites do.

For AI engines, that calculus has shifted dramatically. The 61% vs. 25% gap isn’t a stylistic decoration. It reflects something architectural about how AI engines consume the web that doesn’t apply to classic Google.

Why AI Engines Are Structurally Pulled Toward Schema

The shift isn’t a preference inside OpenAI or Google’s AI division. It’s an inevitable consequence of how large language models actually work. There are at least four reinforcing reasons AI engines lean on schema more heavily than Google ever did, and all four compound.

1. Token Economics

Every LLM has a fixed budget of tokens it can process per query — its “context window.” When an AI engine is answering a question like “best plumber in Tucson,” it doesn’t read the entire web. It pulls a handful of candidate pages into context and reasons across them.

Parsing prose to extract entities is expensive. The model has to identify the business name within a paragraph, classify it as a business at all, locate the address somewhere on the page, recognize the services mentioned, and infer the relationships between all of those things. Every one of those inferences burns tokens, and every token costs money and latency.

Structured data delivers all of that pre-extracted and pre-labeled. The same context window that fits eight prose-only pages can fit eight schema-rich pages plus dramatically more usable entity information. Schema is, in a real sense, information density. When AI providers are trying to deliver fast, accurate answers at massive scale without their compute costs spiraling, density wins. That’s not aesthetic preference. That’s economics.

2. Hallucination Reduction

When an AI extracts information from prose, it can misread, misattribute, or in the worst case invent details that aren’t actually there. AI companies care intensely about this because hallucinations damage user trust, trigger negative press cycles, and erode the willingness of users to rely on AI for real decisions about money and time.

Structured data is explicit. When a page declares — via a standardized format — that the business is named “Smith Plumbing,” located at a specific address, offering a specific list of services in a specific service area, the AI doesn’t have to guess. It reads, it cites, it moves on. Anything that lets a model assert facts with declared confidence rather than inferred confidence is a quality and safety win, and AI providers will increasingly steer their systems toward sources that make accurate citation easy.

3. Citation Accuracy

AI search outputs frequently include citations, links, and named references. With structured data, those discrete fields are right there in machine-readable form: canonical URL, business name, address, hours, services. Without it, the model has to grab whatever it can from the page title, the H1, the meta description, or scattered prose, and hope the result is correct.

For platforms like ChatGPT, Perplexity, and Google’s AI Overviews that cite sources visibly in their answers, this matters enormously. A wrong citation looks worse than no citation. Structured data makes citations cheap and reliable, which is exactly what you want in production at scale.

4. Cross-Source Consistency

When an AI engine reasons across multiple candidate sources for a single query, having those sources in a common format is enormously valuable. Prose is wildly inconsistent — every website writes about its services differently, places its address differently, formats its hours differently, names its categories differently. Structured data is standardized. The same fields appear in the same shape across every site that implements it. That makes comparing, ranking, and synthesizing answers from multiple sources radically easier for the model.

For Google’s classic algorithm, this consistency mattered less, because Google wasn’t deeply reasoning across the content of multiple pages. It was ranking pages by signals like inbound links and authority. For AI engines, which actually read and synthesize content from several sources to produce a single answer, format consistency is a force multiplier.

What This Means for Local Businesses

If you’re a local service business — a roofer, an HVAC contractor, a plumber, a landscaper, a fence installer, an electrician — here’s the practical takeaway.

The traditional SEO foundations still matter. A real website, decent backlinks, an accurate Google Business Profile, fresh content, fast page speed, and a clean technical setup are still the entry ticket. Skip them and AI won’t find you any more reliably than Google would.

But the lever that’s been quietly reweighted underneath everyone’s feet is structured data. The businesses winning in AI visibility today are not always the ones with the biggest backlink profiles or the longest-running websites. They’re the ones whose websites are most machine-readable, with rich and accurate structured data describing the business and what it does in a form AI engines can consume directly.

This is where most local businesses are leaving the most ground uncontested. Schema implementation, when it happens at all, is usually a one-time site-build task. It gets put in place when the website is launched, then quietly forgotten. The vocabulary at schema.org continues to evolve, AI consumption patterns continue to mature, and the original schema sits there unchanged for years.

Meanwhile, businesses that maintain fresh, recurring structured data tied to actual ongoing operations — every job, every project, every location served, every customer interaction — accumulate entity signals across their websites that AI engines can extract and cite. The gap between those two groups is widening, and it’s widening fastest in industries where AI search adoption is moving quickest: home services, professional services, healthcare, and other locally-oriented categories where customers increasingly ask AI for recommendations rather than typing into a search box.

There’s a practical wrinkle. Implementing modern, ongoing structured data is genuinely difficult. The format is a developer skill. Most business owners can’t write it, most office staff can’t, most general-purpose web designers don’t track the evolving recommendations from schema.org, and most agencies deploy structured data once during initial site build and rarely revisit it. The businesses pulling ahead on this lever almost universally have either an in-house technical resource keeping schema fresh, or a tool that handles ongoing structured data deployment as part of normal business operations.

The Bottom Line

SEO is still king. But the throne has been reshaped.

For Google, the kingdom was won with links, content, and authority, with schema as a helpful bonus on the side. For AI search, structured data has stepped from supporting cast to leading role — not because anyone decided to make it that way, but because the architecture of large language models pulls them toward standardized, declared information for hard economic and quality reasons that aren’t going away.

The local businesses that figure this out in 2026 — and especially the ones who treat structured data as an ongoing operational practice rather than a one-time technical task — will own AI search recommendations in their markets long after their slower-moving competitors notice the shift.

The reweighting has already happened. The question is whether your SEO strategy reflects it.

Sources