If you ran the visibility check from last week's post, you probably noticed something: the same business, the same queries, five different sets of results. Maybe ChatGPT named you but Perplexity named a competitor. Maybe Google AI Mode had you in the summary but Claude did not know you existed. Maybe Grok gave results that felt completely disconnected from everything else.
That is not random. Each engine has a different data diet, a different retrieval mechanism, and a different opinion about what makes a local business worth recommending. If you treat them as one monolithic thing called AI search, you will optimize for some of them and miss others entirely. This post breaks down how each engine actually works for local queries, what they read, what they weight, and where the specific leverage points are.
Why the five engines behave differently
The underlying reason they diverge is architectural. Some engines answer from training data, a snapshot of the web at a point in time. Some do live web searches every time a question is asked. Some do both. Some weight structured databases heavily. Some weight social signals. Some are deeply integrated with existing search infrastructure and some are building from scratch.
For local businesses, this means there is no single fix that moves all five at once. There is a set of foundational signals that help every engine, and then there are engine-specific things worth knowing. Start with the foundation, then read the rest of this.
The foundation that helps every engine
Before the engine-specific breakdown, the signals that matter everywhere:
Google Business Profile completeness: Every engine either reads GBP directly or reads sources that mirror it. A complete, accurate, regularly maintained GBP is the single highest-leverage move across all five.
NAP consistency: Name, address, phone number, and hours matching across your website, GBP, Yelp, Facebook, and the major directories. Inconsistency creates hedging across every engine.
Review volume and recency: Every engine reads reviews as third-party validation. The threshold varies but the direction is universal: more recent reviews from more platforms beat fewer older ones.
Website content clarity: A homepage and services pages that say clearly what you do, where, for whom, and since when. This is the source of truth every engine checks against.
Everything below builds on this foundation. If the foundation is broken, engine-specific optimization will not rescue you.
Google AI Mode
Who it affects most
Everyone. This is where the volume is.
How it decides who to recommend
AI Mode is built on Google's existing infrastructure, which means it inherits Google's massive index of the web plus its local data from Maps, GBP, and the Local Pack. Unlike the other four engines, AI Mode already knows your business in detail. It has your GBP data, your reviews, your website, your local citation history. The question it asks is not whether your business exists. It is whether your data is clean and complete enough to stake a recommendation on.
This makes AI Mode the engine most sensitive to data quality. Wrong hours, mismatched categories, incomplete services, and inconsistent citations across platforms all create friction that AI Mode resolves by citing someone else. A competitor with less impressive credentials but cleaner data frequently beats you.
The specific leverage points
The services section of your GBP: Underused by most businesses and disproportionately weighted by AI Mode. Every individual service should be listed as a separate item with a short description in plain language. This is how AI Mode matches "same-day plumbing near me" to your plumbing company rather than to a general contractor.
The Q&A section of your GBP: Parsed directly by AI Mode to answer conversational queries. If someone asks "does [your business name] have parking," the answer in your Q&A shows up in AI Mode. If the Q&A is empty, the engine guesses or hedges.
Google Reviews volume and recency: The dominant third-party signal. The 4.3-star competitor with 340 reviews will consistently outrank the 4.8-star competitor with 28 reviews for most query types. Review velocity (how frequently new reviews arrive) matters more than average score above a certain threshold.
The mistake to avoid
Treating AI Mode as a separate thing from regular Google SEO. They share infrastructure. Fixing your GBP, your reviews, and your NAP consistency does double duty: it helps your organic rankings and your AI Mode citations simultaneously.
Perplexity
Who it affects most
High-research categories: legal, finance, home services, accounting, specialty retail, and medical or dental practices. Perplexity users tend to be information-seeking rather than immediately transactional, which means they are earlier in the decision cycle and more open to being influenced.
How it decides who to recommend
Perplexity does a live web search every time a query comes in. It retrieves current pages, synthesizes them, and cites its sources. This means its results are more up-to-date than any training-data-based engine, but it also means the results shift based on what is currently ranking well and what pages it can most cleanly extract useful content from.
For local queries, Perplexity tends to pull from Google Business Profiles (through third-party directories that mirror GBP data), major review platforms like Yelp, Houzz, Avvo, and TripAdvisor, local press and blog coverage, and your own website when it has well-structured content.
The specific leverage points
Third-party mentions: Matter more on Perplexity than on any other engine. A mention in a local publication, a listing in a curated industry directory, a guest quote in a relevant blog post. These are what Perplexity's retrieval finds and cites. If the only pages talking about your business are pages you published yourself, Perplexity has thin material to work with.
FAQ content on your website: Heavily surfaced by Perplexity. A page that directly answers "how much does [your service] cost in [your city]" or "what should I bring to my first [your service] appointment" gets pulled and cited when customers ask those questions. Perplexity is essentially doing the research your customer would have done themselves. Give it the answers.
Review platform presence beyond Google: Yelp, Houzz, TripAdvisor, Avvo, Angi, and industry-specific directories all feed into Perplexity's retrieval. A business with ten reviews on Google and nothing elsewhere looks thin. A business with reviews spread across three or four relevant platforms looks established.
The mistake to avoid
Publishing only on your own website and ignoring the broader content ecosystem. Perplexity is specifically looking for what others say about you, not just what you say about yourself.
ChatGPT
Who it affects most
Broad consumer categories: restaurants, entertainment, fitness, beauty, and retail. Local queries make up a significant portion of daily ChatGPT use. "Where should I take my mom for her birthday dinner," "a good gym near my new apartment," "best barbershop in the East Village." These are the kinds of questions people are already comfortable asking ChatGPT.
How it decides who to recommend
ChatGPT uses a combination of training data and live web search through its Browse feature, which is increasingly active by default. Its training data is a snapshot of the web up to its knowledge cutoff, and it has been trained on a massive corpus that includes review sites, directories, local blogs, and city guides. Its live search supplements that when queries seem to call for current information.
This makes ChatGPT the engine where your historical web presence matters most. The businesses that appear reliably in ChatGPT recommendations tend to be ones that have been mentioned consistently across the web for years: in Yelp reviews, in city guides, in best-of lists, on food blogs and local press, in community discussions.
The specific leverage points
Longevity of mentions: A unique ChatGPT signal. A business that has appeared in five local blog posts over three years accumulates weight in training data that a newer business or one with a thinner online footprint has not built yet. This is not a fast fix. But it means every press mention, every local blog feature, every city guide listing is worth pursuing.
Best-of list appearances: In local publications, aggregator sites, and curated city guides all tend to transfer into ChatGPT citations. These listings carry more weight per citation than a single directory entry. Local food blogs, neighborhood guides, and annual best-of roundups are all worth targeting.
Well-structured website content: When ChatGPT does live web searches, it reads your website the same way Perplexity does. Well-structured content with clear service descriptions and FAQs gives it material to cite. Make sure your site does not make ChatGPT work to understand what you offer.
The mistake to avoid
Expecting ChatGPT to know about recent changes. If you rebranded, moved locations, or significantly changed your services in the last year, ChatGPT's training data may be showing customers an outdated picture. The fix is ensuring current, accurate information is well-established across sources its live search will find.
Claude
Who it affects most
Professional services, law firms, financial advisors, medical and dental practices, and customers doing longer research sessions. Claude has a reputation for being more careful and more caveat-heavy about local recommendations than other engines, which is actually useful information for how to approach it.
How it decides who to recommend
Claude has web access and increasingly uses it for local queries, but it is more inclined than other engines to acknowledge uncertainty and recommend the customer verify with Google Maps or another source. This means it will not confidently name a business unless it has high-confidence signals. That is a higher bar to clear, but it also means that when it does name you, the customer is more likely to act.
Claude weighs verifiable, structured information heavily. A business with complete schema markup, a detailed GBP, consistent directory presence, and clear website content is a business Claude can describe with confidence. A business with sparse, inconsistent signals is one Claude hedges about.
The specific leverage points
Schema markup on your website: Matters more for Claude than for most other engines. JSON-LD LocalBusiness schema, Service schema, and FAQPage schema give Claude structured, machine-readable facts to draw on. The richer and more accurate your schema, the more confidently Claude can describe your business.
Named human authorities: Improve Claude's confidence. If your website names the owners, practitioners, or specialists at your business with titles, credentials, and bio pages, Claude has more to verify against and more reason to cite you specifically rather than generically recommending a category.
Longer-form content: Because Claude is used for longer research sessions, detailed service pages, FAQ pages, and process explanations give it material that short-form and list-based content does not. A law firm with a page explaining exactly what to expect during a first consultation is one Claude can cite helpfully. One with a one-paragraph homepage cannot be described usefully.
The mistake to avoid
Writing exclusively for humans and ignoring machine readability. Claude does well with well-structured content and struggles with sites that are visually polished but structurally thin. Heavy on design, light on actual words describing what the business does and for whom.
Grok
Who it affects most
Businesses with active social presence and businesses in categories with community discussion on X (formerly Twitter). Grok is the newest entrant and the most socially-weighted of the five engines.
How it decides who to recommend
Grok uses both its training data and live access to X. For local queries, it tends to surface businesses that are actively discussed on X. Mentioned by customers, recommended in threads, talked about in local community accounts. A business with no social presence on X is nearly invisible to Grok relative to its competitors. A business with regular customer mentions and community discussion has a natural advantage.
Grok's training data also includes standard web content, so GBP, website content, and directory presence still apply. But the X signal is what differentiates it from the other four.
The specific leverage points
Customer mentions on X: The unique signal. Customers who tweet about great experiences, screenshots of specials posted to local community accounts, recommendations in neighborhood discussion threads. All of these feed Grok's retrieval in ways that do not affect any other engine.
A low floor for participation: You do not need to run a full X marketing strategy. You need to make it easy for customers who are already on X to mention you: a handle on your receipts, a reply when someone tags you, occasional content worth sharing.
Local community accounts on X: Neighborhood accounts, city-specific lifestyle accounts, and local event pages are worth building genuine relationships with. A single post from a community account with 10,000 local followers can produce more Grok citation weight than ten directory listings.
The mistake to avoid
Dismissing Grok because it is newest and smallest. It is the fastest-growing of the five, especially in markets with active local discourse on X. Getting cited early, while the field is less crowded, is easier than catching up once everyone has figured it out.
How to use this across your engine strategy
Running the visibility check told you which engines you are showing up in and which ones you are not. This breakdown tells you why.
The problem is almost always GBP completeness and review volume.
The problem is thin third-party presence. You need more external pages talking about you.
The gap is usually historical web presence. Not enough mentions over time from sources outside your own domain.
The work is schema markup, named authorities, and structured content depth.
Social mentions on X are the missing piece.
Most businesses underperform in two or three of these for distinct reasons. The fix for each is different. Knowing which engine has which gap is what makes the check worth running, and what makes running it again in 60 days worth doing, to see which gaps closed.
Frequently asked questions
Which AI engine matters most for local businesses?
Google AI Mode has the highest volume because it sits at the top of Google Search results where customers already are. If you can only focus on one engine, focus on AI Mode. But each engine reaches different customers at different stages of the research process, so a complete strategy addresses all five.
Do all AI engines pull from the same data sources?
No. Google AI Mode draws on Google's full index including GBP and Maps. Perplexity does live web searches every time. ChatGPT uses training data plus Browse. Claude uses training data and increasingly live search. Grok has the full X firehose plus training data. The overlap is that all of them benefit from a complete GBP, consistent NAP, and clear website content.
Can I fix all five engines with the same changes?
There is a foundational set of changes that helps every engine: GBP completeness, NAP consistency, review volume, and clear website content. Beyond that, each engine has engine-specific leverage points. Fixing the foundation moves all five. Then the engine-specific work moves the stragglers.
How long does it take for AI citations to improve after making changes?
Typically 4 to 8 weeks, depending on the engine. Perplexity updates fastest because it does live searches. Google AI Mode and ChatGPT take longer because changes need to propagate through Google's index or appear in enough sources for training data to reflect them. Run a visibility check 60 days after making changes to measure the difference.
Does Grok matter for most local businesses?
It depends on the category and market. Grok is the fastest-growing of the five and is especially relevant for businesses with active social communities: restaurants, bars, fitness studios, local events venues, and businesses in cities with active local discourse on X. Getting cited early, before the field is crowded, is easier than catching up later.
Track all five automatically
Seen tracks your citation rate across all five engines weekly
Against the queries that matter for your category and location. You get your citation rate, your competitors' citation rates, and the exact gaps. Without opening a single chat window.
Start your free audit