AI Search & Answer Engine Visibility
When someone asks ChatGPT for the best plumber near them, or Google drops an AI answer above the map for "emergency electrician", a few businesses get named and the rest do not. This guide is about being one of the named ones, for the queries that actually bring you customers. The honest answer to how is less of a trick than people hope: AI is reverse-engineering who is genuinely a strong, trusted choice for this job in this area, and reading it from everything around you. It rewards real businesses that market themselves well across the board, which is good news, because that is something you can actually build.
Start with the queries that bring you customers
A local business does not need to appear for every question an AI can answer. It needs to appear for the ones attached to money: someone is ready to choose and call. Those are commercial-intent queries, and they should drive your whole approach here. Informational questions matter too, but for a different goal, so this guide stays focused on the commercial ones.
Queries worth winning (someone is ready to buy)
- •"best [trade] in Surry Hills", "[trade] near me"
- •"emergency [trade]", "24-hour [trade]", "open now [category]"
- •"[trade] who can [specific need]" (e.g. handles nervous patients, works on combi boilers)
- •"[category] in [neighbourhood]", "recommend a [category] near [place]"
Queries that inform, not convert (someone is just reading)
- •"how to [DIY task]" (they want to do it themselves, not hire you)
- •"[thing] vs [thing]" comparisons
- •"is [thing] worth it", "what is [thing]"
- •Worth creating, just for a different goal than your direct leads.
For the queries on the left, the AI is doing recommendation, not explanation. It wants a named, trusted local business it can stand behind. That comes down to whether you look like the genuine, well-regarded choice: a clear business record that matches what they asked, a strong reputation, and others vouching for you. Not a how-to article on your own site. The rest of this guide works through those, roughly in order of leverage.
How AI decides who to name
The same "best dentist near me" can come back as a Google Map Pack, an AI Overview above the results, an AI Mode conversation, or a paragraph in ChatGPT or Perplexity. They look different, but the machinery underneath is the same, and so are the signals it reads.
- 1
It works out what you mean
The AI turns the question into structured intent: category (dentist), location (the searcher's area), and any constraints (open Saturday, takes nervous patients, emergency appointment). In Google's AI Mode a single question is often split into several smaller ones, a mechanic called query fan-out.
- 2
It pulls candidates
It gathers possible answers from two kinds of source at once: a business/map index (your record, your reputation, your location) and a web index that includes both your own website and the third-party pages that mention you ("best of" lists, directories, local press). For a recommendation query the business index does most of the work, but your own service pages and the pages naming you both feed in.
- 3
It picks a few and writes the answer
It ranks the candidates by how well they match, how trusted they are, and how recent the signals are; works out which records all refer to the same business; and names a few in the answer. Businesses with a clean record, a strong reputation and trusted mentions win this step. Fragmented or thin ones get skipped.
Lever 1: A clear business record, positioned to what people ask
Underneath the jargon, AI treats your business as a known thing in the world, not a string of words on a page. That known thing has a record (usually several) carrying your name, location, category and details. The cleaner and more consistent those records, the more confidently AI names you.
Google Business Profile
The single most important record for a local business. It carries your name, address, phone, hours, categories, services and reviews, and feeds Google's Knowledge Graph. Claim it, complete every field, keep it current.
The Knowledge Graph
Google's internal record of your business as an entity. You do not edit it directly; it is built mostly from your GBP plus consistent mentions of you across the web. It drives Knowledge Panels, AI Overviews, Gemini and AI Mode.
Consistent name, address, phone (NAP)
The same details, written the same way, everywhere you appear (your site, directories, social profiles). Inconsistency is read as two different businesses and waters down your record.
Your site's structured data
LocalBusiness and Organization schema on your own site, with a sameAs list pointing to your other records. This is the one place schema genuinely earns its keep (see the content section below).
The single rule that ties these together: every record that describes your business should agree. Same name, same address, same phone, same category. When they agree, AI is confident it is you and names you. When they conflict, it hedges, or it names a competitor it is more sure about.
Having the records is only half of it. The other half is positioning: making sure what you list actually matches how customers ask. A plumber whose GBP primary category is "Plumber", with services like "Emergency plumber", "Boiler repair" and "Bathroom installation", lines up with real searches; one filed vaguely under "Contractor" does not. A dentist who spells out "nervous patients", "same-day emergencies" and "Invisalign" gets matched to exactly those questions. A restaurant that names its cuisine, its "Sunday roast" and "private dining" gets pulled into those. Say plainly what you do, in the words people use to ask for it, across your GBP categories and services, your description, and your service pages. AI can only match you to a query if you have clearly claimed the thing being asked for.
Wikidata, if you qualify
Wikidata is the open, structured sibling of Wikipedia, and parts of it feed Google's Knowledge Graph and AI training data. Its bar is far lower than Wikipedia's, but not zero: an entry needs at least one independent, verifiable reference to the business (a government register, a reputable directory, real press coverage), not self-published claims. Many established local businesses clear that bar; a brand-new one with no outside coverage will not. If you qualify, create or claim the entry, fill in the core properties (P31 instance-of, P17 country, P159 headquarters location, P856 official website), and add the resulting Q-number to your sameAs list so your site and the Wikidata entry point at each other. It is a one-off job that tends to make established businesses surface more reliably in our testing across ChatGPT, Perplexity and Claude.
Lever 2: Your reputation, what people say about you
Your reputation is what AI leans on to decide you are any good, and reviews are the biggest single part of it, but not the whole of it. It reads the actual review text, not just the star average, and lifts specific phrases ("sorted our boiler the same day", "great with nervous kids") as evidence for naming you. It also weighs your wider standing: sentiment across the places people check, and whether you are spoken about well elsewhere. A strong, consistent reputation is what a genuinely good business accumulates over time, and it is exactly what AI is trying to detect.
- 1
A steady stream, not a one-off pile
HighestRecent reviews count for more than old ones. A business getting a few genuine reviews every week reads as alive and well-used; forty reviews from two years ago reads as stale. Build a simple, consistent ask into your normal workflow.
- 2
Reviews that name the actual service or problem
HighGeneric 'great service, highly recommend' tells AI little. 'Fixed our leaking shower the same morning' ties you to a specific job and a specific intent, which is exactly the phrasing AI extracts when answering 'who can fix X near me'. Prompt for specifics without scripting the words.
- 3
Reputation beyond your Google rating
HighAI does not only read your Google stars. Sentiment on Trustpilot, Facebook and the platforms that matter in your sector (Checkatrade or TrustMark for trades, Doctify for clinics, OpenTable for restaurants), plus unprompted mentions in social and forums, all feed the picture. Consistent positive sentiment across several sources is more convincing than a great score in one place.
- 4
Your replies
UsefulResponding to reviews, the good and the bad, signals an active, real business that is paying attention. It also adds your own words (and service terms) to the listing.
Reputation feeds both halves of the picture: it helps AI decide you are worth naming, and the review text supplies the language it quotes. The operational detail on the biggest piece is in our guide to getting more Google reviews.
Lever 3: Earned media, getting named on pages AI already trusts
For competitive "best of" queries, the page AI quotes is usually not yours. It is a "best plumbers in [city]" round-up, a "top dentists" list, a local restaurant guide, a directory, a news piece, or a trade-body register. Getting your business onto those pages puts your name in the exact place AI looks when it decides who to recommend.
A point traditional SEO understates: these mentions count even when they do not link to you. Across our testing and several independent 2025 and 2026 analyses, the volume of unlinked mentions of your business across the web tracks AI visibility more closely than backlinks do. The model is pattern-matching on how often, and how credibly, your name shows up next to the thing being asked about. Being talked about, in the right places, is one of the higher-leverage things a local business can invest in.
- 1
Earn mentions in recognised local press and trade media
Highest leverageRegional newspapers, city listings, BBC local, sector trade press. Each mention is something AI can quote, and it usually outranks your own content for competitive 'best of' queries.
- 2
Get into curated 'best of' round-ups
Strong (hospitality, beauty, retail, trades)The list articles AI loves to cite. Pitch the writer directly, give them a reason you stand out, and offer a quote or a real data point. Each inclusion is a fresh citation source.
- 3
Claim your professional-body and regulator listings
Strong (and essential for YMYL)Law Society of [state], AHPRA, Medical Board of Australia, CPA Australia, CA ANZ, ASIC's Financial Adviser Register. AI treats these as high-trust sources, especially for medical, legal and financial services.
- 4
Sponsor or contribute locally
ModerateCommunity publications, local events, charity work. Often produces lightly-linked or unlinked mentions that still feed your reputation and give AI more places to find your name.
Your own content: a supporting role, not the main event
This is where most "AI content" advice over-promises. Your own blog and how-to pages are not what win commercial queries, and the evidence that schema markup directly lifts AI citations is weak. That does not make content worthless: it does real work for awareness and top-of-funnel. It just is not the lever for the commercial queries this guide is about, which changes where it sits in the plan.
Where your own content does earn its place:
- Service pages that clearly say what you do, who for, and where. These support both your entity and the commercial queries directly, and they are worth more than any blog post.
- Practical FAQ content answering the real pre-booking questions ("do you charge a call-out fee", "how soon can you come"). It helps customers and gives AI clean, quotable answers tied to you.
- Genuine depth on your subject, which can get you pulled into AI Mode answers on adjacent questions and quietly reinforces that you know your field. Treat this as a long-game extra, not the core of the plan.
If you do write content, the structural rule is simple: make each section answer one clear question, put the answer first, and keep passages self-contained so AI can lift a clean quote without needing the paragraphs around it. Plain, descriptive headings beat clever ones. Write it for a human; the structure is what makes it easy for a machine to use.
Where each assistant gets its answers
The assistants are wired into different search stacks. You do not need to optimise each one separately, but knowing where each gets its data tells you which basics to confirm.
ChatGPT (OpenAI)
ChatGPT search runs mainly on OpenAI's own index, built by its OAI-SearchBot crawler, having moved away from its earlier reliance on Bing. Make sure OAI-SearchBot can crawl you. Crawlers: GPTBot (training), OAI-SearchBot (search), ChatGPT-User (user-triggered).
Gemini, AI Overviews, AI Mode (Google)
Google's index plus the Knowledge Graph plus Maps. The same signals power all three, which is why a strong GBP and clean entity record do most of the work here.
Perplexity
Runs on its own web index (crawled by PerplexityBot). Always cites its sources and weights recent, well-structured pages heavily.
Claude (Anthropic)
Uses Brave Search for live web retrieval (confirmed via Anthropic's published subprocessor list) plus its trained knowledge. Crawlers: ClaudeBot (training), Claude-User (user-triggered), Claude-SearchBot (search).
Two more worth knowing. Microsoft Copilot (which absorbed Bing Chat in late 2023) is grounded in Bing, so Bing Webmaster Tools and a Bing Places listing feed straight into its answers. Apple Intelligence surfaces businesses through Apple's own ecosystem (Apple Maps, Apple Business Connect, Spotlight) and routes some questions to ChatGPT when the user opts in, so for iOS-heavy audiences a complete Apple Business Connect listing is worth the same attention as your GBP.
Crawler access: the audit nobody runs
AI crawlers obey robots.txt, and plenty of sites block them by accident, through wildcard rules, agency defaults, or templates inherited from before these crawlers existed. If they cannot reach you, you cannot be quoted. Check that each of these is intentionally allowed (or intentionally blocked):
OpenAI / ChatGPT
- •
GPTBot: training crawler - •
OAI-SearchBot: search crawler for ChatGPT search - •
ChatGPT-User: fetches a page when a user asks ChatGPT to look
- •
Googlebot: standard crawler (also feeds AI Overviews and AI Mode) - •
Google-Extended: an opt-out token, not a crawler. Controls whether your content trains Gemini and grounds it via Vertex AI. - •Per Google, blocking Google-Extended does not affect Search, and does not remove you from AI Overviews or AI Mode, which use the live Search index
Anthropic / Claude
- •
ClaudeBot: training crawler - •
Claude-User: fetches a page when someone asks Claude to look - •
Claude-SearchBot: indexes content for Claude's search answers - •Note:
anthropic-aiandClaude-Webare old, deprecated names. Blocking only those does nothing to the three current bots.
Apple, Perplexity, others
- •
PerplexityBotandPerplexity-User: indexer plus user-triggered fetches - •
Applebot-Extended: Apple's AI-training opt-out token (paired with the standard Applebot) - •
CCBot(Common Crawl): a widely-used training-data feeder - •
Bytespider,Meta-ExternalAgent: ByteDance and Meta
You may also hear about llms.txt, a file describing your site's content for AI. Be realistic: as of 2026 no major AI search product fetches it when answering queries (server logs show the main crawlers do not even request it, and Google's Search team has compared it to the long-dead keywords meta tag). It is cheap and harmless to add, but it is not a visibility tactic.
Measuring whether it is working
AI search has no single rank to track. You measure it by testing the queries you care about across the assistants and asking two things: are you named, and what is said about you. Lead with the commercial queries; the informational ones are optional.
- Recommendation: 'best [category] in [city]', '[category] near me'
- Recommendation: 'recommend a [category] near [area]', 'a good [category] in [neighbourhood]'
- Specific need: 'I need a [category] who can [specific need]'
- Urgent: 'emergency [category] near me', 'open now [category]'
- Filter: 'highly-rated [category] in [city]', 'affordable [category] near [area]'
- Brand check: '[your business name]' (does your Knowledge Panel look right?)
- Optional, informational: 'what does a [category] cost in [region]?'
- 1
Are you named?
PrimaryAcross your commercial query set, how often does each assistant name you? Track it per assistant and per surface (Map Pack, AI Overview, AI Mode, ChatGPT, Perplexity, Claude), month over month. This is the number that matters most.
- 2
What is said about you?
ImportantWhen you are named, what language is used? Specific, positive phrasing is what turns a mention into a customer; a flat neutral mention does little.
- 3
Who is named instead?
ImportantWhich competitors appear where you do not, and what do they have that you lack (more reviews, a 'best of' mention, a cleaner record)? This is where your next move usually reveals itself.
- 4
Are your pages or earned-media mentions cited?
UsefulWhere the assistant shows sources, is it citing your site or third-party pages that name you? Tells you whether to push on earned media or your own pages.
Running this by hand across many queries every week is impractical. Our Answer Engine Visibility tool automates it, so you can sample dozens of prompts a week across the major assistants without hand-running each one.
Common mistakes
The 90-day plan
A practical sequence for a local business, in priority order: basics first, then the levers that actually move recommendation, then measurement.
- 1
Days 1 to 30: the basics
Claim and fully complete your Google Business Profile. Fix your name, address and phone so they match everywhere. Audit robots.txt for every AI crawler. Add valid LocalBusiness and Organization schema with a
sameAslist. Confirm a Bing Places and Apple Business Connect listing that match your GBP. Nothing clever yet, just a clean, reachable, consistent record. - 2
Days 31 to 60: reviews and earned media
Put a simple, repeatable review request into your normal workflow, prompting for specifics about the actual job. Claim your professional-body and regulator listings. Pitch for one or two mentions in local press or "best of" round-ups. If you qualify, create or claim a Wikidata entry and link it back via
sameAs. This is the phase that moves recommendation. - 3
Days 61 to 90: measure and close gaps
Test your commercial query set across ChatGPT, Gemini, AI Overviews, AI Mode, Perplexity, Claude and Copilot. Note where you are named, what is said, and which competitors appear where you do not. Trace each gap to the underlying lever (record, reviews, or mentions) and fix that. Re-test monthly.
Reference numbers
3 levers
What actually moves AI recommendation
Your entity (GBP, Knowledge Graph, consistent NAP), your reviews, and earned media (third-party pages that mention you). In that rough order.
30-90d
Realistic lag
Between making a change and it showing up consistently in AI answers. Wikidata is faster; Knowledge Graph updates are slower.
10+
AI crawlers to check
GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-User, Claude-SearchBot, PerplexityBot, Perplexity-User, CCBot, plus the Google-Extended and Applebot-Extended opt-out tokens.
Monthly
Re-test cadence
Assistants change their data and answers often. Re-run the same commercial query set every 30 days, across surfaces.
Where this is going
AI search is not replacing Google overnight, but it is taking a growing share of the moment where customers choose. The squeeze is already visible: one 2026 analysis found AI local results named only about a third as many distinct businesses as the equivalent map pack did for the same searches. The pool of named businesses is shrinking, which makes being one of them more valuable, not less. The businesses that win are the ones doing real marketing well: a clear, consistent record that says plainly what they do, a steady flow of specific reviews and a genuinely good reputation, and a presence on the pages AI already trusts. AI is getting better at spotting exactly that, and harder to fool with anything less. The ones that do not get named quietly disappear from the decision, even if their old rankings hold, because the customer never scrolls a list. They just got the answer.
Where to go next
Keep reading