ChatGPT vs Google: How Customers Are Finding Local Businesses Now

In January 2025, AI Overviews appeared on about 6.5 percent of tracked Google queries. By late 2025, that number had grown to 15 to 16 percent, based on Semrush's 2025 analysis of over 10 million keywords (Semrush, "AI Overviews Study"). That's the shift this post is about, a real, measured, fast-moving change in ordinary search behavior, not a forecast.

It shows up in ordinary moments. Someone in your own household probably did this last month: needed a plumber, a divorce attorney, or a decent Thai place nearby, and instead of opening Google, opened ChatGPT and typed something like "who's a reliable plumber near me that won't overcharge for a weekend call." Or asked Perplexity to compare three project management tools for a five person team. Or asked Gemini, right inside an Android phone, to just handle it.

The shift matters more for local and service businesses than almost any other category, because these are exactly the high stakes, comparison heavy, help me decide questions that people have started outsourcing to a conversational AI instead of a search bar. Nobody wants to read four blog posts to figure out which HVAC company to call. They want an answer.

Google search and AI chat answer questions in fundamentally different ways, and that difference is the whole game.

Google shows you a ranked list. You, the human, do the filtering. You skim titles, glance at review stars, maybe click two or three links before deciding. Your business could rank fourth and still get plenty of business, because humans browse.

An AI assistant doesn't hand you a list to browse. It reads across the web, decides what it believes is true and relevant, and gives you a synthesized answer, often naming two or three specific businesses or products by name. If your business isn't one of those two or three it names, you don't get a consolation click. You get nothing.

This is the core insight behind a body of research now often called Generative Engine Optimization, or GEO. Academic work out of Princeton, Georgia Tech, and IIT Delhi formally tested how content characteristics affect whether AI systems cite a source in their generated answers, and found that traditional SEO tactics don't map cleanly onto what gets a business mentioned by an AI model (arXiv 2311.09735, "GEO: Generative Engine Optimization"). The paper's central finding, in plain terms: what makes content easy for a search engine to rank and what makes content easy for a language model to trust, extract, and repeat are related but distinct skill sets.

Nobody outside OpenAI, Google, or Anthropic has the exact recipe, and anyone who tells you they've cracked the algorithm is selling something. But a few patterns show up consistently in independent research and in what we see running weekly AI visibility scans for OMG customers.

Clarity beats cleverness. A page that plainly states "We are a licensed plumbing company serving Austin, Texas, specializing in emergency water heater repair" gets picked up more reliably than a page built around branded taglines and vague value propositions. AI models are pattern matching and fact extracting machines. Give them facts to extract.

Structured data acts like a translation layer. Schema markup, the behind the scenes code that tells machines "this is a business, this is its address, this is its service, this is its hours," makes it dramatically easier for an AI crawler to understand your page without having to guess. Google's own documentation on structured data explains this exact function: it exists to remove ambiguity for machines reading your content (Google Search Central, "Introduction to structured data").

Being mentioned elsewhere still counts, maybe more than ever. AI models often pull from and cross reference multiple sources rather than just your own site. Directory listings, review sites, local press, and other third party mentions of your business feed into what the model considers established fact about you.

Freshness and specificity matter. Generic "About Us" pages written years ago and never touched since are much less useful to a model than a page updated recently with specific service details, current pricing ranges, and clear service area information.

You don't need a total rebuild. Start with the boring, structural stuff. Get your basic facts machine readable with LocalBusiness, Service, and FAQPage schema. Write pages that answer specific questions plainly, so that if people ask AI tools "who does X near me," a page on your site answers that exact question in those exact terms. Keep an eye on where you're mentioned outside your own site, since reviews, directories, and local press all feed the models' sense of who you are. And check your visibility regularly, not once, since AI models update what they know constantly and a one time audit goes stale fast.

This is exactly the gap OMG was built to close. Our weekly AI visibility scans check how ChatGPT, Gemini, Claude, and Perplexity currently see your business, then generate a concrete to-do list: blog posts, schema fixes, FAQ pages, llms.txt files, the whole list. If you'd rather not do it yourself, our Do It For Me option lets our team, or full autopilot if you opt in, actually execute the fixes for you.