How AI Engines Decide Who to Cite

The most common question I get from Brooklyn business owners after showing them their AI audit results is some version of: "Why isn't my business coming up? I have Google reviews. I have a website. I've been here for 12 years." The answer is always some variation of the same thing — the signals that get you found on traditional Google are not the same signals that get you cited by AI engines. Understanding the difference is the entire premise of what Signal does.

There is no algorithm to game here in the way people think about Google SEO. AI engines are making a fundamentally different decision: not "which of these pages is most authoritative for this keyword?" but "which real-world entity is the best answer to this question, and can I state that with enough confidence to put it in my response?" That's a retrieval and confidence problem, not a ranking problem. And it responds to different inputs.

These are the four signals I've identified, through direct work with Brooklyn businesses, that most reliably determine whether and how AI engines cite a local business.

Signal One: Entity Clarity

Before an AI engine can cite your business, it needs to know your business exists as a distinct, identifiable entity in the world — not just as a webpage. This sounds obvious, but it's where most local businesses fall short in the AI era.

An entity, in the context of knowledge graphs and AI retrieval, is a real-world thing with a stable identity that can be recognized and verified across multiple data sources. For a local business, entity clarity means: the AI can confidently say "this is a specific optometry practice in Crown Heights, Brooklyn" rather than "this is a webpage that mentions optometry and Crown Heights."

What builds entity clarity:

• Consistent NAP (name, address, phone) — the exact same business name, address, and phone number appearing verbatim across your website, Google Business Profile, Yelp, and every directory where you're listed. Any variation — "Ave" vs. "Avenue," different suite numbers, old phone numbers — creates ambiguity that reduces AI confidence.
• Schema.org identifiers — LocalBusiness markup with sameAs links to your GBP, Yelp, and other authoritative profiles. This creates machine-readable entity connections that AI systems can traverse to verify your identity.
• Structured database presence — a claimed and complete Google Business Profile is the most important single asset here. Yelp, Apple Maps, Bing Places, and Wikidata entries (where relevant) add verification layers. The more places your entity is confirmed with consistent data, the more confidently an AI can name you.

Nostrand Optical's AI visibility improvement over 60 days was substantially driven by entity clarity work done before we touched a word of website copy. We audited and corrected NAP consistency across 24 directories, completed the GBP to 100% including all service attributes, and added Schema markup with sameAs links to all authoritative profiles. That infrastructure gives the AI something solid to point to.

Signal Two: Retrieval-Grade Content

This is the signal I spend the most time explaining to clients, because it's the most counterintuitive relative to everything people have been told about web content for the past 15 years. AI engines aren't looking for content that ranks — they're looking for content that extracts cleanly into a citation.

The difference in what that content looks like is significant. Keyword-optimized content tends to use vague superlatives, passive constructions, and repeated keyword phrases. It's written to match a search query string. Retrieval-grade content is written in declarative entity-prose — direct, specific, attributable statements about who you are, what you do, where you do it, and for whom.

AI engines extract facts. If your site doesn't contain clearly stated facts, it doesn't get cited.

Compare these two sentences:

Version A: "We offer comprehensive eye care services to the Crown Heights community with a focus on personalized patient experiences and the latest optical technology."

Version B: "Nostrand Optical is a full-service optometry practice on Nostrand Avenue in Crown Heights, Brooklyn, offering eye exams, contact lens fittings, pediatric vision care, and prescription frames, with same-week appointments available for new patients."

Version A is marketing copy. Version B is a citation. Only one of these sentences will appear in a Perplexity or ChatGPT answer. When I rewrote the Nostrand Optical homepage with this structure — the declarative entity statement in the first paragraph, specific services named plainly, location stated precisely, audience described clearly — the citation frequency in AI answers improved materially within weeks.

How to write retrieval-grade content

Every key page on your site should open with a statement that answers, in one or two sentences: who you are (business name and type), where you are (neighborhood and city), what you do (services), and who you serve (audience). Then each service should have its own page or well-defined section with the same structure: service name, what it includes, who it's for, any relevant specifics (price range, duration, approach). FAQ sections written in natural question-and-answer format are direct candidates for AI citation for conversational queries.

Signal Three: Citation Consistency

This signal is related to entity clarity but operates at a different level. Where entity clarity is about establishing that you are a distinct, verifiable entity, citation consistency is about ensuring the facts AI engines retrieve about you are the same regardless of where they look.

AI systems, especially those with live web retrieval like Perplexity, triangulate facts across multiple sources before committing to a claim. If your website says your hours are 9am–6pm Monday through Friday, but your Google Business Profile says 10am–7pm, and a 2022 Yelp listing says 9am–5pm, the AI has contradictory data. It resolves contradictions either by declining to state the fact, hedging it, or picking one source arbitrarily — none of which is good for you.

Citation consistency work means auditing every place your business information lives online and bringing it into alignment. This is unglamorous work. It involves claiming old directory listings on platforms you didn't know you were listed on, updating hours that haven't been touched since 2019, and correcting address formatting in a dozen different places. But it is load-bearing for AI confidence. An AI that sees the same facts repeated verbatim across 20 sources cites them with confidence. An AI that sees contradictions hedges or omits.

Signal Four: Topical Authority

The fourth signal operates over a longer time horizon than the first three. Topical authority is what happens when an AI engine's retrieval model has encountered your business in the context of a specific topic enough times, across enough sources, that it indexes you as a relevant entity for that topic — not just a business that happens to have a website.

For a local business, topical authority is built through depth of content on your primary service area, external citations from relevant sources, and a consistent track record of being mentioned in the context of your topic. A single-page website with an about section and a contact form has essentially no topical authority. A site with dedicated pages for each service, FAQ content answering real customer questions, case study content, and some form of regular publishing — a blog, field notes, client updates — builds it over time.

The minimum viable content footprint

For a Brooklyn solo practitioner or small business, topical authority doesn't require 200 blog posts. It requires depth on the specific services you offer and the specific questions your prospective customers ask. For the Brooklyn BJJ Lessons site, that meant: a dedicated page for private adult lessons, a dedicated page for kids' private lessons, a detailed about page establishing the instructor's credentials and teaching philosophy, a FAQ answering the 8 most common questions beginners ask before their first class, and enough prose on each page to demonstrate expertise beyond the basics. That's achievable in a one-time content build. It's not a years-long content marketing project.

External citations — being mentioned by name in relevant contexts outside your own site — also contribute to topical authority. A mention in a local Brooklyn publication, a reference in a neighborhood forum, a feature on a relevant industry site: these all add data points that AI systems use to verify your relevance to a topic. You can build this organically over time, or you can accelerate it through targeted outreach and local press, which is part of the Premium tier work Signal does for clients.

How the Four Signals Work Together

The four signals aren't independent. They compound. A business with strong entity clarity but no retrieval-grade content is recognizable to AI but doesn't give it anything specific to say. A business with excellent content but inconsistent citation data will be found by some AI engines and missed by others. A business with all four signals working together is the one that shows up first, with confidence, across all four major AI engines.

Here's how the signal picture looked for Nostrand Optical before and after Signal's build:

• Entity clarity: Before — NAP inconsistencies across 11 directories, GBP at 60% completion, no sameAs schema. After — NAP consistent across 24 directories, GBP at 100%, full LocalBusiness schema with entity links.
• Retrieval-grade content: Before — marketing-copy homepage, no service-specific pages. After — declarative entity prose on homepage, dedicated pages for each service, 9-question FAQ with natural language Q&A.
• Citation consistency: Before — 6 directory listings with varying data. After — 24 listings with verified, matching NAP.
• Topical authority: Before — single-page site, no depth. After — 8 pages of service and FAQ content, structured to demonstrate expertise in Crown Heights optometry specifically.

The result was first-position AI citations across the majority of tested prompts within 60 days. Not all of that is attributable to any single signal — it's the combination that moved the needle. The businesses that still don't appear in AI answers typically have one or two signals working and two or three neglected. The fix is never "just add more content" or "just fix your schema." It's identifying which signals are weak and closing the gaps systematically.

The upshot for any Brooklyn business owner: your 12 years in the neighborhood don't automatically translate to AI visibility. AI engines don't have institutional memory of your presence on the block. What they have is a snapshot of the structured, retrievable data that exists about you right now. If that data is sparse, inconsistent, or buried in generic marketing language, the AI passes you over — not because it doesn't know you exist, but because it can't say something confident and specific about you with enough certainty to stake a recommendation on it. That's a solvable problem. It just requires working the four signals deliberately, not hoping your existing web presence carries over.