Field Notes
AI Search Is Taking Over. Here's What Your Local Business Should Do About It.
A growing share of your future customers will never see a page of Google results. They'll type "who's the best plumber near me?" into ChatGPT, or ask Perplexity for "a good family dentist in Mesa that takes new patients," and they'll get a short, confident answer naming two or three businesses. If yours isn't one of them, you don't rank lower — you simply don't exist in that conversation.
This guide explains, in plain English, how those answers actually get made, what the credible evidence says about how big this shift is, and what a local business owner can do about it — including a 15-minute test you can run yourself today, for free.
What actually happens when someone asks an AI for a local recommendation
When someone asks an AI assistant for a recommendation, one of two things happens behind the scenes — and the difference matters for you.
1. The model answers from memory ("training knowledge")
Large language models are trained on an enormous snapshot of the public web — websites, directories, review discussions, news articles, Wikipedia. If your business appeared consistently across that material, the model may "know" you the way a well-read newcomer to town might: it has absorbed your name, what you do, and roughly how you're regarded. This memory is fuzzy, has a cutoff date, and can't be directly edited. You influence it the slow way: by being genuinely well-documented across the web for years.
2. The model looks things up live ("retrieval")
Increasingly, assistants don't rely on memory alone. Perplexity, ChatGPT with search, and Gemini can run live web searches behind the scenes, read a handful of pages — review sites, maps listings, local guides, your website — and compose an answer from what they just read, often with citations. This is the part you can influence on a normal business timescale, because it draws on sources you can improve this quarter: your Google Business Profile, your review presence, your website's actual text.
Why this is different from ranking on Google
Traditional local SEO is a ranking contest: ten blue links plus a map pack, and being #4 still gets you seen. AI answers work differently in three ways that change the strategy:
- It's a shortlist, not a list. Assistants typically name two to five businesses and stop. There is no page two. The gap between "mentioned" and "not mentioned" is the whole game.
- It's synthesized, not linked. Google shows sources and lets the user judge. An assistant reads the sources and delivers a verdict. If three review platforms and a local news mention all describe you as the reliable option for emergency calls, that consensus is what tends to surface — no single "ranking factor" hack replicates it.
- It's conversational and specific. People ask assistants longer, more detailed questions than they type into Google — "a dog groomer near downtown that handles anxious dogs." Businesses whose web presence actually answers specific questions in plain text tend to be the ones the model can match to specific requests.
How big is this shift, really?
Ignore the hype threads. Two widely-cited, attributable data points are enough to justify taking this seriously:
Gartner's forecast. In February 2024, the research firm Gartner publicly predicted that traditional search engine volume would drop 25% by 2026, with search marketing losing share to AI chatbots and other virtual agents. Forecasts are forecasts — but Gartner is not a company that profits from AI-search panic, and mainstream coverage of the shift has only grown since.
The zero-click reality. SparkToro and Datos published a 2024 study of real user clickstream data finding that for every 1,000 Google searches in the US, only 374 clicks went out to the open web. The majority of searches already ended without a click to anyone's website — before AI answers became mainstream. The lesson: the era of "just get traffic to your site" was already fading. Visibility increasingly means being named in the answer, wherever the answer appears.
Neither source says Google is dying — it isn't, and your Google presence still matters enormously (not least because AI assistants read Google-adjacent sources). The evidence suggests something more specific: a meaningful and growing slice of discovery is moving into answer engines, and businesses that are invisible there are invisible to that slice.
The signals AI assistants tend to draw on
No one outside these companies knows the exact recipe, and anyone who claims to is selling something. But from how these systems are built, what they cite, and what practitioners observe when testing, a consistent picture emerges. Assistants tend to draw on:
- Entity clarity. Can a machine tell, unambiguously, what your business is, where it is, and what it does? A consistent name, address, phone number, and category across your website, Google Business Profile, Apple Maps, Yelp, and the major directories makes you a clean "entity." Contradictions (old addresses, three spellings of your name) make you noise.
- Structured data. Schema.org markup — LocalBusiness type, hours, services, service area — is machine-readable labeling on your website. It's cheap to add and removes guesswork for crawlers.
- Review platforms. Google reviews, Yelp, industry-specific platforms (Avvo, Healthgrades, Houzz, TripAdvisor). Assistants cite these constantly. Volume, recency, and the actual text of reviews all appear to matter — reviews that mention specific services in words give the model matchable language.
- Third-party mentions. "Best of" lists in local publications, chamber of commerce pages, news coverage, supplier and association listings. Independent sources saying you exist and are good carry more weight than you saying it yourself.
- Wikipedia/Wikidata-tier sources. Most local businesses won't have (and shouldn't chase) a Wikipedia page — but the principle generalizes: the more authoritative and neutral the source describing you, the more it tends to shape the answer.
- Crawlable, answer-shaped content. Pages that state facts in plain text: "We serve Gilbert, Chandler, and Queen Creek. Emergency calls answered 24/7. Licensed, bonded, ROC #123456." Text locked inside images, PDFs, or JavaScript-heavy designs may never make it into an answer. If your site can't be read aloud by a machine, assume it isn't being read at all.
Test your own AI visibility in 15 minutes
You don't need a tool or a consultant to get a first read. Open ChatGPT (free tier is fine), Perplexity, and Gemini, and paste in prompts like these — edited for your trade and city. Use a private/incognito session so your history doesn't color results, and note that answers vary run to run; what you're looking for is the pattern, not any single response.
- "What are the best [your trade] in [your city]?"
- "I need a [your trade] in [your city] for [your most profitable service]. Who should I call?"
- "Recommend a [your trade] near [neighborhood or landmark] that [key differentiator — e.g., offers same-day service / takes new patients / is open weekends]."
- "Compare [your business name] and [top competitor] in [city]. Which would you choose and why?"
- "What do customers say about [your business name]?"
- "Is [your business name] in [city] legit? What are its reviews like?"
- "What services does [your business name] offer, and what are their hours?"
- "Who are the most trusted [your trade] in [city] for [second service or specialty]?"
- "I'm new to [city]. How do I find a reliable [your trade], and can you suggest a few?"
- In Perplexity specifically: any of the above — then read the citations. Which sites is it pulling from? Are you on them?
Score yourself honestly across the three assistants: How often are you named at all? Are your details (hours, services, phone) correct? Who shows up instead of you, and which sources get cited when they do? Those cited sources are, quite literally, your to-do list.
If the results are weak: what to do, in order
Work top to bottom. The early items are cheap, fast, and foundational; the later ones compound over time.
- Fix your facts everywhere. One exact name, address, phone, website, and category on Google Business Profile, Apple Maps, Bing Places, Yelp, Facebook, and the top directories for your industry. Kill duplicates and dead listings. This is unglamorous and it is the foundation.
- Complete your Google Business Profile fully. Every service listed, real photos, accurate hours, attributes filled in. Assistants and the sources they read lean heavily on it.
- Build a steady review engine. Ask every happy customer, make it one tap, and reply to reviews. Steady beats bursts, and reviews that naturally mention specific services are worth more than "great job!"
- Put plain-text answers on your website. A page per major service and per service area, stating who you serve, what it costs (even ranges), how fast you respond, and what makes you different — in sentences a machine can quote.
- Add structured data. LocalBusiness schema with hours, geo, services, and review markup where legitimate. A competent developer can do this in an afternoon.
- Earn third-party mentions. Local press, "best of" roundups, chamber and association listings, community sponsorships that come with a real webpage link. One genuine local news mention tends to outweigh dozens of junk directory entries.
- Re-run the 15-minute test quarterly. Keep a simple note of which prompts mention you. This is your scoreboard, and it's free.
What not to waste money on
- "Guaranteed #1 in ChatGPT" services. No one can guarantee placement in a probabilistic system they don't control. Anyone promising it is describing a service that cannot exist as promised.
- AI-generated content farms. Hundreds of thin, machine-written pages don't create the independent, consistent evidence assistants draw on — and they can degrade your standing with Google, which you still need.
- Fake or incentivized reviews. Beyond being an FTC violation with real penalties, review platforms are actively filtering them, and a filtered profile is worse than a modest one.
- Exotic "LLM optimization" tricks. Hidden prompts on your webpage, keyword-stuffed pages "for the AI." There's no credible evidence these work, and the trust cost if a customer notices is real.
- Panic-buying every new "AEO/GEO platform" (the industry's new acronyms for AI-answer optimization). Monitoring tools have their place later. But if your listings are inconsistent and your review flow is a trickle, a dashboard just measures the problem in higher resolution.
The checklist
- ☐ Ran the 10-prompt test on ChatGPT, Perplexity, and Gemini; recorded who gets mentioned
- ☐ Name, address, phone, and category identical across Google, Apple, Bing, Yelp, Facebook, key directories
- ☐ Google Business Profile 100% complete with services, photos, and current hours
- ☐ Review request built into your routine; responding to reviews weekly
- ☐ One plain-text page per major service and per service area on your website
- ☐ LocalBusiness structured data installed and validated
- ☐ At least one credible third-party mention pursued this quarter
- ☐ Calendar reminder set to re-test in 90 days
None of this is exotic. It's the same discipline that built good local reputations for decades — being findable, consistent, and well-spoken-of — applied to a new kind of reader: one that reads everything, remembers imperfectly, and answers out loud on your behalf. The businesses that treat that reader seriously now tend to be the ones it names later.
Want to know where your business actually stands?
Apply for a free visibility audit: your rankings, your AI-assistant mentions, and where competitors are ahead. Free, and yours to keep either way.