Key takeaways
- Google still routes most literal "near me" queries to the traditional Local Pack / Map Pack, not AI Overviews. BrightLocal data cited by Tarasaka puts the overlap between AI Overviews and "near me" queries at just 4-8%.
- ChatGPT handles "near me" differently. Since its March 26, 2026 location-sharing rollout, it can use precise device location (GPS/address) in addition to the IP-based approximation it already relied on, and it names this exact use case ("best coffee shops near me") in its own documentation.
- Across ChatGPT citations generally, product and landing pages dominate commercial-style prompts (about 70% combined per Promptwatch's content-type data), while Reddit's citation share in ChatGPT collapsed from roughly 4% to 0.5% overnight on August 14, 2026.
- Whitespark's 2026 local ranking factors survey shows a reallocation away from Google Business Profile weight and toward website signals and citations when the target is AI Search visibility rather than the classic Local Pack.
- 45% of consumers now use AI tools for local recommendations, up from 6% a year earlier (BrightLocal LCRS 2026), and ChatGPT is the single most-used AI tool for that purpose.
Two systems, two different jobs
Here's the thing that trips up most local business owners trying to figure out "AI SEO": ChatGPT and Google AI Mode are not competing to answer the same query the same way. They're often not even trying to answer the same query at all.
Type "best pizza near me" into Google and in most cases you still get the old-school Local Pack: three pinned businesses on a map, star ratings, a "directions" button. According to the BrightLocal data referenced in Tarasaka's 2026 local SEO guide, AI Overviews overlap with genuine "near me" queries only 4-8% of the time. Google appears to have decided, at least for now, that hyper-local transactional intent is better served by Maps data than by a generated paragraph. AI Overviews instead show up for the research layer around local decisions: "how much does a roof repair cost," "best type of flooring for a rental," the kind of question that needs synthesis rather than a pin on a map.
ChatGPT works from the opposite direction. It has no native map pack to fall back on, so when someone asks "best coffee shop near me," it has to simulate what a map pack would give you, using web search results plus, now, your actual location.
How ChatGPT actually figures out "near me"
For most of 2024 and 2025, ChatGPT inferred location purely from IP address, baked quietly into its system prompt. People on Reddit and Hacker News documented it giving city-level "near me" answers without ever asking permission, and sometimes contradicting itself about whether it had location data at all.
That changed on March 26, 2026. OpenAI shipped opt-in device location sharing on iOS and web (Android "coming soon" at the time), with two tiers: an approximate location signal and a precise one pulled from GPS or a typed address. OpenAI's own example of when precise location kicks in is, almost word for word, the query type we're discussing here: "best coffee shops near me." The precise data itself gets deleted after generating the response, though if the response names specific places, those names stick around in your chat history. It's off by default, sits in Settings > Data Controls, and the industry writer Glenn Gabe nicknamed it the "Near Me ChatGPT Update," which is a decent description of what it is.
Worth sitting with for a second: this isn't a minor UX tweak. OpenAI has also started testing in-chat ads (February 2026) and signed Criteo as its first ad-tech partner in March, with a reported $200K minimum pilot spend and roughly 17,000 advertisers already in the door. Location data is described internally as "foundational" to future local ad targeting. So the "near me" feature isn't just about helpfulness, it's infrastructure for a local ads business that doesn't exist yet but clearly will.
What ChatGPT actually cites when it answers commercial queries
Once ChatGPT has a location, it still needs sources to recommend from. Promptwatch's own citation-type data (week of September 28 to October 4, 2026) shows a heavy lean toward pages that describe what a business sells: product pages are the single most common citation type at 22.4%, landing pages add another 16.8%, and together they make up 39% of all citations. Narrow that to commercial-intent prompts, the closest proxy for "best X for Y" and "best X near me" style questions, and the combination of product pages, landing pages, listicles, and comparisons accounts for roughly 70% of citations.
That's a meaningful signal for a local business. It means your service page, not your blog, not your Instagram, not a forum thread about you, is the thing most likely to get pulled into an answer. If your plumbing company's "emergency drain repair" page reads like filler copy instead of a genuine description of the service, area covered, and price range, you're handing ChatGPT nothing usable.
The mechanics of how ChatGPT searches got more aggressive too. On August 8, 2026, its use of the site: operator inside fanout queries jumped from about 0.4% to roughly 17% of all fanout queries overnight, a roughly 46x increase in a single day, and average searches per response nearly doubled at the same time. In plain terms: ChatGPT now frequently runs a search scoped directly at a business's own domain (site:yourdomain.com plumbing austin) in addition to open-web queries. If your site isn't properly crawlable or your service/location pages aren't indexed, that direct lookup comes back empty.
At the same time, ChatGPT's fanout queries have gotten shorter and blunter. Average query length dropped from about 117 characters in December 2025 to roughly 53 by April 2026, per Promptwatch's query fanout data. The system is now firing off terse keyword strings like "best plumber austin" rather than conversational sentences. Practical consequence: write your page headings the way a tired person types a search, not the way a chatbot talks.
Reddit is fading, directories and mid-tier sites are not
If your local SEO plan has been leaning on "get mentioned in a local subreddit," it's time to rethink that. Reddit's share of ChatGPT citations fell from about 4% to 0.5% on a single day, August 14, 2026, a collapse rather than a gradual slide, per Promptwatch's reddit citation tracking. Google's surfaces pulled back from Reddit too, just more slowly over the same window.
Social platforms overall barely register in ChatGPT: just 0.46% of citations in the first week of October 2026. Compare that to Google AI Overviews, which drew 12.2% of citations from social platforms in the same window, and you've got a real strategic fork. If you're optimizing specifically for ChatGPT visibility, time spent chasing social mentions is largely wasted; if AI Overviews matter more to your category, it's a different calculation.
What does still work for ChatGPT is domain authority in the middle of the pack, not the extreme top. Promptwatch's August 2026 domain-rank data shows sites rated DR 46-60 and DR 61-75 together pulled in about 46% of citations, the single most reliable range, while the very top tier (DR 91-100) fell from roughly 7% to about 3% over the course of the month. That's good news for a local business: you don't need to out-rank Yelp or TripAdvisor site-wide. A solid local directory listing, a regional news mention, or an industry association page with decent authority can carry real weight.
Ranking factor weights are shifting under local SEO's feet
Whitespark's 2026 Local Search Ranking Factors survey (47 experts, 187 ranked factors, published November 2025) introduced a separate scoring category this year specifically for AI Search visibility, distinct from the classic Local Pack and organic rankings. One analysis of that survey data (Digital Applied) breaks out how weight shifts between the two goals:
| Ranking factor | Weight for Local Pack | Weight for AI Search visibility |
|---|---|---|
| Google Business Profile signals | 32% | 12% |
| On-page / website signals | 15% | 24% |
| Reviews | 20% | 16% |
| Citations | 6% | 13% |
| Links | 8% | 13% |
The biggest loser is Google Business Profile, cut by 20 points once you're optimizing for how AI systems answer rather than how the Map Pack ranks. The biggest winner is your own website, on-page content, service-area pages, schema, NAP consistency, topical depth, which jumps to become the single heaviest factor for AI visibility. Citations and links both roughly double in weight too, with Local Falcon's 2026 analysis pointing specifically at placement on expert "best of" lists and prominence on industry-relevant domains as two of the top five AI visibility factors.
Reviews stay important either way, but recency now matters more than raw volume. BrightLocal's 2026 Local Consumer Review Survey found 74% of consumers want to see reviews from the past three months before trusting a recommendation.
The schema question: genuinely unresolved
There's a lot of confident-sounding advice floating around claiming LocalBusiness schema produces specific, large lifts in AI citations, numbers like "40% more AI Overview appearances" or a "2.5x higher chance" of being cited. I'd treat those specific percentages with real skepticism. They come from agency blog posts without published methodology, and at least one controlled test, run by Doublebricks, found no statistically meaningful increase in ChatGPT or Google AI Mode citations from adding schema, and even a small, statistically significant decrease in AI Overview appearances.
That doesn't mean skip schema. LocalBusiness markup (name, address, phone, geo coordinates, hours, service area, price range) still helps with rich results and gives search engines an unambiguous read on your entity, which matters for a different reason: clarity, not a citation hack. If you run multiple locations, a parent Organization entity with individual LocalBusiness entities underneath, matched to your Google Business Profile data, is the structurally sound way to do it. Just don't expect it to single-handedly move your AI visibility numbers.
Why consumers are actually trusting this now
The adoption curve here is steep. BrightLocal's March 2026 Local Consumer Review Survey found 45% of consumers now use AI tools for local recommendations, up from just 6% a year prior, with ChatGPT as the single most-used tool and Google's AI surfaces (AI Mode, Gemini) in second place. AI tools have actually overtaken Yelp and TripAdvisor for local discovery in this survey. Millennials lead the adoption curve; people over 60 remain the most cautious.
BrightLocal's CEO Myles Anderson put it bluntly: a Google-only reputation strategy now risks invisibility, because "ChatGPT and other LLMs can't see inside Google's walled garden of reviews. If your reputation only exists on Google, you are effectively invisible to the millions of people using ChatGPT to find local services." That's a real structural gap, not scaremongering. Also telling: 88% of consumers say they fact-check AI-cited reviews before trusting them, so AI recommendation isn't replacing skepticism, it's adding a new first filter before the skepticism kicks in.
How many businesses actually get named
If you're wondering how competitive the "best X near me" slot actually is: per BrightLocal data on average citation behavior, Google AI Overviews name about 2.5 businesses per answer, AI Mode names about 3.5, and ChatGPT names about 4.1. ChatGPT is the most generous with name-checks but also, by some accounts, more selective about which businesses make that cut in the first place, a PushLeads analysis (worth flagging as a secondary, non-peer-reviewed source) claims only 1.2% of local business locations ever get recommended by ChatGPT, attributing roughly 44% of visibility to the business's own website, 42% to listings and directories, 28% to Google Business Profile, and just 2% to forums. Treat the exact percentages loosely, but the direction matches everything else here: your own site and your directory presence carry most of the weight, not social chatter.
What this means in practice for a local business
A few concrete, non-glamorous moves actually matter based on everything above:
- Write real service pages, not filler. Product and landing pages are what ChatGPT cites most for commercial queries, so a thin "our services" page with no specifics is wasted real estate.
- Make sure your own domain is crawlable and indexed properly. ChatGPT now runs site: scoped searches directly against your domain roughly 17% of the time in fanout queries, so a blocked or poorly indexed site means it finds nothing there to cite.
- Get listed on mid-authority directories and local press rather than chasing the biggest national sites. DR 46-75 domains earn the largest citation share.
- Keep reviews fresh, not just plentiful. Recency now outweighs raw count for AI Search visibility weighting.
- Don't bet heavily on Reddit or social for ChatGPT specifically, its social citation share is under half a percent. If Google AI Overviews matter more for your category, that calculus changes since social makes up over 12% of its citations.
- Add LocalBusiness schema for clarity and multi-location structure, but don't expect it alone to move your citation rate.
Tracking whether any of this is working
The honest problem with all of the above is that it's genuinely hard to know, without tooling, whether your business is showing up when someone nearby asks ChatGPT or Google AI Mode for a recommendation. You can run the query yourself a few times, but that only shows you what you see, from your IP, your account history, your region. It doesn't show patterns across hundreds of prompt variations or track whether you dropped out of an answer after a model update.
This is the specific gap that Promptwatch is built to close. It monitors how brands and local businesses actually get cited across ChatGPT, Gemini, Perplexity, Google AI Overviews, and AI Mode, not through API sampling but by watching the real user-facing interfaces, which is important because what the UI shows can differ from what an API call returns. For a local business, the useful parts are the crawler logs (so you can see whether ChatGPT's bots are even reading your service pages), citation tracking by page and domain, and the content gap analysis that shows which competitor pages are getting cited instead of yours for a given prompt.
If you want to compare options more broadly, the GEO software directory at bestgeosoftware.com lists other AI visibility platforms worth evaluating depending on your budget and team size.
A quick comparison of what to optimize for which platform
| Signal | Matters for ChatGPT | Matters for Google AI Mode / Overviews |
|---|---|---|
| Precise device location | Yes, since March 2026 rollout | N/A for AI Mode text answers, handled by Maps layer |
| Literal "near me" queries | Answered directly by the AI | Mostly routed to Local Pack, not AI Overview |
| Product/service pages | ~39% of all citations, highest category | Still important, but social and listicles carry more weight |
| Reddit mentions | Collapsed to ~0.5% of citations | Declining, but more gradually |
| Mid-tier domain authority (DR 46-75) | ~46% of citation share | Less documented, but directionally similar |
| Reviews | Secondary signal via cited pages | Direct factor, feeds Map Pack and AI Overview prominence |
The bottom line
Treating "ChatGPT SEO" and "Google AI Mode SEO" as the same task is the most common mistake I see local businesses make in 2026. Google still mostly answers literal "near me" queries with its decades-old Local Pack, meaning Google Business Profile, reviews, and proximity still rule there. ChatGPT, lacking that infrastructure, is building its own version from device location plus web citations, and it's citing service pages and mid-tier directories far more than social chatter or big-name domains. Optimize for both, but don't assume a tactic that works on one will move the needle on the other.