Someone needs a bank branch. Or a hospital. Or a school for their kid. Where do they look first — and has that changed because ChatGPT exists now?
A 2026 research project presented at a webinar hosted by Duda, a website-building platform for agencies, went looking for a real answer. The findings were shared by researchers Myles Anderson and Shawn Davis. Two studies, run side by side. One was a roughly 1,200-person consumer panel asking people how they actually search for local businesses. The other was bigger and more technical: about 1,300 local businesses, tested with roughly 200,000 “unbranded” prompts fired across ChatGPT, Google’s AI Mode, and AI Overviews. (“Unbranded” means searches that describe what someone wants — “best hospital near me” — rather than naming a specific business.) Those prompts generated close to 2 million citations pulled out of the answers.
Two terms worth pinning down before the numbers start: AI Overviews are the short, AI-written summary boxes that sometimes appear above Google’s normal search results. AI Mode is different — a separate, full-page, conversational AI search experience inside Google, more like a back-and-forth chat than a quick summary. Both are Google products; they just behave, and draw on data, differently.
That’s enough data to say something real about how AI is reshaping local search. And, just as usefully, where it isn’t.
This isn’t only about restaurants and plumbers, either. “Local” covers anything tied to a nearby or specific location — a bank branch, a diagnostic center, a school or college campus. Run any kind of multi-location business, and this data is about you, whether you’ve thought about it that way or not.
Where People Actually Start Looking
Start with the number that keeps the hype honest: 61% of consumers still start a local search on Google. Barely moved — it was 60% a year earlier. Despite everything written about AI rewriting search behavior, the starting point for most people hasn’t budged much at all.
What did move is the AI chat share — from 2% to 8% in a year. Four times larger. Still a small slice of total behavior. But the direction, and the speed, are hard to wave off.
Here’s the more useful finding, though: almost nobody searches on just one channel. 75% of consumers use more than one per search. Among people who start on Google, 67% go check at least one other channel before deciding anything. And about 12% of Google-starters cross over to AI chat before they’re done — with a similar share moving the other direction, starting on AI chat and finishing on Google.
Google and AI chat, in other words, aren’t really two competing journeys for most people. They’re often the same journey, with a stop at both.
And it’s a fast journey. 55% of consumers pick a business within 15 minutes of starting. 75-80% decide within half an hour. Whatever channel a business shows up on, the window to be found, look credible, and get picked is short.
AI’s Last-Mile Problem
Here’s the catch buried in that 12% crossing from AI chat back to Google. Most of them aren’t leaving because AI chat gave them a bad answer. They go to Google for what the research calls the “last mile” — confirming hours, getting directions, checking whether a place is actually open right now.
AI chat tools are good at narrowing a list down and explaining why one option might fit. They’re trusted less, for now, with the real-time operational facts that matter in the final minutes before someone walks into a branch or a clinic. That’s a real limitation. It’s also an opening: businesses that keep hours, locations, and live details accurate and easy to verify hold up better at that last-mile check, no matter where it happens.
What’s Actually Inside an AI Engine’s Citations
This is where the research earns its keep for anyone trying to actually influence what an AI says.
But domain count isn’t citation volume. Count every single citation — all roughly 2 million — and business websites account for about 42% of the total. The rest comes from directories, review platforms, and other third-party sources, repeated again and again across different answers.
So two things are true at once. A business’s own website is the single most common type of source cited. And a smaller set of directory and review domains get reused so often that they still add up to a large share of total citation volume.
That second point shows up again elsewhere: in traditional Google Maps ranking, business listings are estimated to influence results by around 6%. In AI search, that roughly doubles — to about 13%. Listings and citations simply carry more weight in what AI engines choose to recommend than they do in how Google Maps ranks a pin on a map.
It’s not only Google and Yelp doing this work, either. Researchers found more than 415 other directories and review sites showing up in citations — together, roughly 30% of everything collected. Being listed, accurately and consistently, across a wide spread of these sites is clearly feeding into what AI engines surface.
ChatGPT and Google AI Mode Are Reading From Different Books
If one finding here should actually change how you work, it’s this.
Google Business Profile is the free listing a business sets up directly with Google, showing its hours, location, photos and reviews on Google Search and Maps. In ChatGPT’s results, it barely shows up as a citation source — researchers found it isn’t a meaningful source there at all. Yelp dominates instead — a review and business-directory site, most widely used in the US, where people rate and write about local businesses. The likely reason: ChatGPT leans heavily on Yelp for local information, while Google isn’t handing its own Business Profile data to a competing AI product.
Google’s AI Mode runs the opposite way. There, Google Business Profile alone accounts for roughly 53% of citations — the clear dominant source.
So there’s no single “AI SEO” lever to pull. A business with a fully optimized Google Business Profile and a thin, stale Yelp presence is set up well for Google’s AI Mode and poorly for ChatGPT. Flip it around, same problem in reverse. Showing up consistently across AI engines means treating the two data sources as genuinely separate workstreams — not one combined task you check off once.
Reviews: The Second-Biggest Signal
Same split applies to reviews. Google reviews feed Google’s AI products. Yelp reviews feed ChatGPT’s citations, for the same partnership reason. Strong Google reviews sitting next to a neglected Yelp profile is, again, only half the job done.
Freshness compounds it. Content published or updated within the last month showed up roughly 7.6x more visible in AI results than older content. Whatever it covers specifically, the direction lines up with everything else in this research — AI engines appear to weight recency harder than classic Google Maps ranking ever has.
The Bar Is Higher Than It Looks
One more number worth sitting with. AI engines typically recommend only 2 to 4 businesses per answer. Compare that to Google Maps, where a user can scroll an endless list. That’s a much smaller stage and a correspondingly harder bar to clear than ranking somewhere on a scrollable map.
The research put a sharp number on just how hard. Tracked businesses showed up in Google Maps results in 66% of searches tested but across AI surfaces, that dropped to roughly a third, in the 32-38% range. Being in the underlying data and being recommended turned out to be two very different things.
Consistency was also uneven. In a round of repeated testing over time, less-optimized businesses bounced around — present in one check, gone in the next. Businesses with the fundamentals solidly in place stayed far more consistent.
That gap — consistent versus random — is really the whole argument for the playbook below. AI recommendations reward the businesses that have done the complete, unglamorous, foundational work everywhere. Not the ones chasing a single AI-specific trick.
The Five-Pillar Local AI Visibility Playbook
None of these five are exotic. That’s the point.
- Broad, deep content covering every service and every location. Clean structure. Proper schema markup — hidden code added to a page that tells search engines and AI systems exactly what the content means (this is a service page, this is a review, this is a phone number), rather than making them guess. Visible proof — reviews, testimonials, ratings — displayed where both people and AI systems can see it. Pricing and service details out in the open, not hidden behind a “contact us” form. Increasingly: built so an AI agent — a tool that can complete a task on someone’s behalf, like checking availability or starting a booking — can actually use the site, not just read it.
- Google Business Profile. Fully filled out. Kept current — hours, services, photos, posts, all of it. Given how heavily Google’s AI Mode leans on this data, a stale profile is a direct visibility cost, not a cosmetic one.
- Fresh, flowing consistently into 4-5 relevant platforms rather than piled onto just one. Given the Yelp/Google split above, which platforms matter most depends on which AI engine actually matters most to you — but leaning on a single source leaves real gaps either way.
- Citations and listings. As wide and as consistent as possible across the directory landscape. Not just Google and Yelp — the hundreds of other sites that collectively make up a meaningful share of AI citations.
- Digital PR. Real “best of” mentions and coverage on genuine local blogs and news outlets. Not fabricated lists. Not paid placement dressed up as editorial. Actual third-party recognition that search engines and AI systems alike can find and trust.
Nothing here is new advice on its own. What’s new is seeing AI engines weight all five together and sometimes weight them differently than Google Maps always has.
The Trap: AI Content Farms
One clear warning from this research is worth repeating plainly. Don’t try to shortcut any of this by mass-producing AI-written pages to cover every possible search variation — generating a separate near-identical page for every way someone might phrase the same question, sometimes called “fan-out query” coverage.
The pattern showed up consistently. Sites generating large volumes of thin AI content to chase every long-tail query got a real traffic spike at first. Then, months later, once nobody kept updating those pages and both traditional search systems and AI engines caught on, the traffic collapsed.
Given how hard freshness and real depth showed up everywhere else in this data, that’s not surprising in hindsight. Volume without upkeep isn’t a visibility strategy. It’s a temporary illusion of one.
Branches, Hospitals, Campuses: This Is a Per-Location Job
This is where the data connects most directly to the kind of businesses that live or die on being found locally, over and over, at scale.
Take an NBFC — a non-banking financial company, a lender that offers loans and similar financial services without holding a full banking license. A bank or NBFC with branches across a dozen cities isn’t one “local business” in AI’s eyes. It’s dozens. Same for a hospital network with multiple facilities, or a university running several campuses — each one judged on its own. The five-pillar logic has to run at the level of each individual branch, hospital, or campus: location-specific website content, a fully optimized Google Business Profile per site, fresh reviews flowing into the right platforms for each location, consistent and accurate listings everywhere that location shows up.
That’s heavier than optimizing one brand-level presence. It’s also exactly where a lot of multi-location brands currently fall short — most local-SEO effort has historically concentrated at the brand level, not the branch level.
Worth being honest here, too: this space is still early and volatile. It lines up with what shows up in Ridure’s own anonymized AI-visibility research across BFSI, healthcare, and education brands — visibility in ChatGPT and similar tools varies enormously, with most organizations sitting in the low single digits and a handful reaching far higher. There’s no shortcut here, and no guarantee. This is directional evidence of where the fundamentals pay off, not a promise of a specific result.
A Starting Checklist
Pulled straight from the five pillars. Work through it location by location, if you run more than one:
- Does every location have its own clear, complete page — services, pricing, proof — rather than just pointing back to a brand homepage?
- Is schema markup in place so search engines and AI systems alike can parse what each location actually offers?
- Is every location’s Google Business Profile complete and current — hours, services, photos?
- Are reviews coming in consistently — not once a year — across at least a few relevant platforms, not just one?
- Is each location listed accurately across the major directories, not only Google and Yelp?
- Has the business earned any genuine, verifiable “best of” or local press coverage recently?
- Is anything on the site stale enough that it needs refreshing rather than being left alone?
FAQs
Is AI chat replacing Google for local search?
Not yet, and not quickly. 61% of consumers still start their local search on Google — essentially unchanged from the year before. AI chat’s starting share grew from 2% to 8%, a real jump, but still a small slice of overall behavior.
What’s the difference between AI Overviews and AI Mode?
AI Overviews are short, AI-written summary boxes that sometimes appear above Google’s normal search results. AI Mode is a separate, full-page, conversational AI search experience inside Google — more like a back-and-forth chat than a quick summary box. Both are Google products, but per this research, they draw on different mixes of underlying data.
Does Google Business Profile matter for ChatGPT visibility?
Not much, per this data. ChatGPT’s local citations lean heavily on Yelp rather than Google Business Profile. Google Business Profile matters far more for Google’s own AI Mode.
How many businesses does an AI engine actually recommend at once?
Usually just 2 to 4 — far fewer than the scrollable list you’d get on Google Maps, which is what makes consistent visibility such a high bar.
Can a multi-location business — bank branches, a hospital network, school or college campuses — use this same playbook?
Yes, but only if it’s applied at the individual-location level: its own website content, its own Google Business Profile, its own review flow and listings, repeated consistently across every branch or campus.
Is mass-producing AI content pages a shortcut to AI visibility?
No. The research found sites doing this saw an initial traffic spike, then a collapse once the content went stale and search systems caught on. Depth and ongoing upkeep mattered more than raw page count.
The Core Takeaway
AI chat is still a small fraction of how local searches start. But it’s growing fast, and it rewards a different — sometimes opposite — mix of signals than classic Google Maps ranking ever did. The businesses that show up consistently in AI answers aren’t chasing an AI-specific trick. They’ve already done the fundamentals well, everywhere: deep, well-structured website content; a fully optimized Google Business Profile; fresh reviews flowing to the right platforms; broad, accurate listings; real digital PR. For multi-location brands — bank branches, hospital networks, education campuses — that work happens one location at a time. Not once, at the brand level, and done.
Learn more about how Ridure approaches generative AI search optimization across platforms, or see how this plays out specifically for ChatGPT visibility and broader AI search visibility work.
