There is one type of search query that breaks most AI-visibility playbooks: "near me." A user in Sandton asking "best Italian restaurant near me," a homeowner in Fourways asking ChatGPT for "an emergency electrician nearby," a parent in Cape Town asking Copilot for "the closest pharmacy that does after-hours prescriptions" — these are queries where the AI assistant cannot bluff. Either the geographic data is right, or the answer is useless.
For these queries, AI assistants don't replace Google Maps. They quietly defer to it. Understanding why — and what that means for your business — is the difference between investing in the right local-SEO work and chasing AI-visibility theory that never moves a single booking.
What proximity actually requires
"Near me" looks like a simple query. Computationally, it is not. To answer it you need:
The user's location (lat/long), inferred from device GPS, IP, or stated context ("Sandton").
A geographic dataset of businesses with verified addresses, categories, hours, and other attributes.
A ranking function that combines relevance, prominence and proximity to choose three to five candidates.
Verification signals — recent reviews, recent posts, recent photos — that prove the business is still operating.
A language model has access to none of this by default. It is trained on text; it does not maintain a lat/long index of every plumber in Johannesburg. So when a user asks a "near me" question, the assistant does what it always does — it delegates to a system that does have the dataset.
In practice, that system is Google Maps (or a derivative of it). Even AI assistants that use Bing as a primary index typically fall through to Maps-style data for high-intent geographic queries. The Google Maps dataset is the most complete, the most actively maintained, and the most demanded by users — which is why every AI assistant ends up reading from it for local intent.
Why AI almost never sends users far away
There is a structural reason AI assistants do not "discover" a great business 30km away and recommend it over the mediocre one next door: the prompt itself includes a proximity constraint. The user said "near me." The model has to honour that constraint or the answer is wrong.
This protects local independents in ways the AI-doomer narrative does not acknowledge. A well-optimised neighbourhood plumber with a complete GBP, 30 recent reviews, and the right primary category will beat the nearest national chain for "plumber near [suburb]" queries regardless of which AI assistant the user is on. The prompt geography overrides national brand authority.
There is one exception: high-prominence categories with low proximity sensitivity. "Best restaurant in Cape Town" weights prominence (reviews, press, brand) much more heavily than proximity, because the user is willing to travel. For everyday local services — plumbing, electrical, medical, school holiday programmes, vet — proximity dominates.
The local pack is the AI shortlist
The single best predictor of AI visibility for a local query is your position in Google's local pack (the map + three businesses block at the top of a local SERP).
This is not a coincidence. AI assistants read the local pack as their primary source for "near me" queries because it is already a ranked, filtered, location-aware shortlist. The top three local results are disproportionately likely to appear in the AI's source set; positions 4–10 see a noticeable drop-off; below that, citations are rare.
The implication is unambiguous: work on local pack rank, and AI visibility follows. The two are not separate channels.
The local optimisation checklist
This is the audit we run when a South African business wants AI visibility on local intent. Most of it is fundamental local SEO, executed with more discipline than usual.
Google Business Profile
Claim and verify the listing. Use a real address, not a PO Box. Service-area businesses (no walk-ins) can hide the address but still need one for verification.
Pick the most specific primary category. Add secondary categories sparingly — only those you genuinely offer.
Fill every field: services with descriptions, attributes, hours including special hours, service area, payment methods, languages spoken.
Publish a GBP post weekly. Offers, updates, events. Empty post history reads as inactive.
Upload fresh photos monthly. Geotagged where possible.
Use the Q&A section actively — seed it with the questions your customers actually ask and answer them yourself, before competitors do.
Reviews
Build a steady, organic flow. Recency matters as much as volume — 30 reviews in the last 12 months beats 200 from 2019.
Reply to every review, positive and negative. Reply within 48 hours.
Encourage reviewers to mention the specific service they used and the suburb — this strengthens topical and geographic relevance signals.
NAP consistency
Audit every directory listing your business appears on. Yellow Pages, Brabys, Apple Maps, Bing Places, industry-specific directories.
The exact phone number format, business name, and address must match across every listing.
Kill duplicate listings — they fracture authority and confuse the retrieval layer.
On-site signals
A dedicated location page for each physical or service location. Embedded Map, opening hours, services, parking notes, transport.
LocalBusinessschema on the home page;Serviceschema withareaServedon each service page;FAQPageon the location page with genuinely useful questions.A clear single-sentence brand answer in the homepage hero and meta description.
Local authority
Citations from local sources — chambers of commerce, industry bodies, local press, partner businesses, school newsletters, community Facebook groups (these often get indexed).
Sponsor or partner with one local event per quarter and get the event organiser to link/cite you.
Build a small backlink profile from genuinely local domains. Five high-relevance local links beat fifty generic directory submissions.
Where AI does add something new
There is one place AI assistants behave differently from a Google local pack, and it is worth optimising for: the assistant often paraphrases your business in its own words.
Where Google shows your business name, category, rating, and a snippet, an AI assistant might say "For an emergency electrician near Fourways, you could try [Business Name], which handles after-hours work and has strong recent reviews." The phrase "handles after-hours work and has strong recent reviews" is paraphrased from your GBP, your reviews, and your service page.
You can influence this paraphrase by being explicit about what you do. A service page that opens with "We are a Fourways-based electrical contractor specialising in after-hours emergency call-outs, fault-finding, and DB board upgrades for residential customers" gives the assistant a tight, citable sentence. A page that opens with "Welcome to our website — we are passionate about electrical excellence" gives it nothing usable.
The bottom line
"Near me" is not an AI-visibility problem. It is a local-SEO problem with an AI layer on top. Google Maps still owns the geographic dataset, the local pack is still the shortlist, and the businesses that win the local pack are still the ones that win in AI answers.
If you have been chasing exotic AI-visibility tactics for local queries, stop. Spend the next month on your Google Business Profile, your reviews, your NAP, and your location pages. The AI citations will follow — because the engines you are trying to please are reading the same data.