Section 1
First, understand the machine you're optimizing for
You cannot game a system you do not understand, so start with the mechanism. Most AI assistants answer using retrieval-augmented generation (RAG): when asked a question, the system retrieves relevant pages from the live web (or its index), extracts the passages that best answer the question, and cites the pages those passages came from (2). The implication is liberating, you are not trying to win an abstract popularity contest; you are trying to be the page with the most extractable, quotable, trustworthy passage for a specific question. The two big assistants behave differently, and the difference shapes your tactics. Perplexity performs real-time web searches, visiting roughly 10 pages per query and citing only 3 to 4 (2), so freshness and live crawlability matter enormously. ChatGPT leans more on training data updated periodically, plus live search, so durable authority and broad corroboration matter more (2). A service business should optimize for both: be freshly crawlable for Perplexity, and broadly authoritative for ChatGPT. Stop thinking "how do I rank?" and start thinking "if an AI pulls one passage to answer my buyer's question, why would it be mine?" That question reorganizes everything you do.
Section 2
The seven moves that get you cited
The research has become specific enough to prescribe. Here are the seven highest-leverage moves, in priority order. 1, Let the AI crawlers in. The most embarrassing way to lose is to be invisible by accident. Many sites block, or fail to allow, the AI crawler bots, and a page that cannot be crawled cannot be cited. Ensure your robots configuration permits the major AI agents, including GPTBot, OAI-SearchBot and ChatGPT-User (OpenAI), ClaudeBot and Claude-SearchBot (Anthropic), PerplexityBot, and Google-Extended (2). This is a five-minute fix that is a precondition for everything else. 2, Answer the question immediately, under a question-shaped heading. This is the single most evidence-backed tactic. Nearly 72.4% of pages cited by ChatGPT contained a short, direct answer immediately after a question-based heading (2). The pattern is explicit: phrase a heading as the question your buyer asks ("How much does a brand refresh cost?"), then answer it directly in the first sentence beneath, before any elaboration. You are handing the AI a pre-packaged, extractable answer. This is the answer-first structure, made literal. 3, Structure with clean heading hierarchy. Pages with proper H1/H2/H3 nesting are 40% more likely to be cited by AI engines than unstructured content (2), because the hierarchy tells the machine what the page covers and how to find the relevant section. Disciplined, logical headings are not just for readers anymore; they are extraction scaffolding for AI. 4, Add statistics and cite sources. The foundational academic work here, the Princeton-led "GEO" study, found that adding statistics and adding source citations each improved a page's AI visibility by 30–40% (3). This is one of the most actionable findings in the field: data and references do not just persuade humans, they materially increase the odds an AI quotes you. Service copy thick with specific numbers and credible citations is structurally favored. 5, Build third-party authority profiles. AI systems cross-reference, and certain platforms carry outsized weight. Brands with active profiles on review and authority platforms like Trustpilot, G2, and Capterra have a 3x higher chance of being cited by ChatGPT, because those platforms aggregate the credibility signals AI uses (2). For a service business, this means claiming and building out the relevant review and directory profiles in your category. Citation is partly earned off your own site. 6, Publish an llms.txt file. An emerging standard, llms.txt, is a file that tells AI models what your site is about, what you offer, and how to reference you, giving them a clear, authoritative source and reducing the risk of hallucinated or wrong information about you (2). It is early and not universally consumed yet, but it is low-cost insurance and a signal of GEO maturity. (The degree to which engines currently use llms.txt is still evolving; treat it as a cheap forward bet, not a guaranteed lever.) 7, Use the right schema. Structured data, particularly FAQ and "speakable" schema, helps engines identify and extract your answers (2). Mark up your questions and answers so the machine can parse them unambiguously.
Section 3
The citation moves, in one view
The seven moves are roughly ordered by leverage-to-effort: crawler access and answer-first structure are fast and high-impact; review profiles and schema take longer but compound. None requires a developer-heavy rebuild, most are content and configuration changes a founder or a capable freelancer can make. And because the Ahrefs data shows over 90% of even top-ranking pages are never cited (1), the businesses that do this work are competing in a nearly empty field.
Section 4
Execute This With AI
Here is a workflow to engineer your citations with any capable AI model, and the AI engines themselves. Step 1, Baseline your citation status. Ask ChatGPT, Perplexity, and Google's AI mode the real questions your buyers ask. Record whether you're named, what's said, and who's cited instead. This is your starting line. Step 2, Run the citation-optimization prompt on a key page: You are a GEO specialist optimizing my page to be CITED by ChatGPT and Perplexity. Apply seven evidence-based moves: 1. Crawlability (note which AI bots must be allowed: GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, Claude-SearchBot, PerplexityBot, Google-Extended). 2. A direct answer immediately under a question-shaped heading (72% of cited pages do this). 3. Clean H1/H2/H3 hierarchy (+40% citation likelihood). 4. Statistics and source citations (each +30–40% AI visibility, Princeton GEO study). 5. Third-party authority profiles (Trustpilot/G2/Capterra = 3x citation chance). 6. An llms.txt self-description. 7. FAQ/speakable schema. My page and the buyer question I want to be cited for: """ [PASTE PAGE TEXT + THE QUESTION] """ Output: 1. Rewrite my page's key section as a question-heading + immediate direct answer. 2. Restructure my headings into clean H1/H2/H3. 3. Suggest 3 specific statistics or citations I could add (mark any needing a real number from me as [NEED INPUT], do not invent data). 4. Draft an llms.txt for my business. 5. List the exact review/authority profiles I should build in my category. Be concrete and prioritize by impact. Step 3, Re-test after changes. Wait for re-crawl (faster for Perplexity), then re-ask the buyer questions and track whether your citations improved. Tools and expected output. The AI engines (for testing), any model (for optimization), plus access to your robots/schema settings and review profiles. Expect a rewritten answer-first section, restructured headings, suggested data points, an llms.txt draft, and a profile-building list. The QA discipline, critical for GEO: never let the model fabricate the statistics in move 4. The entire mechanism rewards verifiable facts that get cross-checked off-site; a fabricated number that an AI cites and another source contradicts destroys the authority you're building. The model structures and prompts; every fact must be real and yours. Word-of-mouth used to be the most powerful and least controllable way a service business got clients. AI recommendation is its successor, equally powerful, and, unlike word-of-mouth, engineerable. Let the crawlers in, answer the question right under the question, thicken your pages with real data and citations, and build the off-site profiles AI trusts. Do that while over 90% of your competitors remain uncited, and you become the name the assistant gives the next buyer who asks.
Section 5
Keep reading
Keep reading in the GEO & AI Search cluster and across the library: [Generative Engine Optimization: The Service Founder's Guide to Getting Recommended by AI](/blog/generative-engine-optimization-the-service-founders-guide-to-getting-recommended-by-ai), [Topical Authority for AI: The Content Cluster Strategy That Gets You Cited](/blog/topical-authority-for-ai-the-content-cluster-strategy-that-gets-you-cited), [Getting Into AI "X vs. Y" Answers as the Recommended Option](/blog/getting-into-ai-x-vs-y-answers-as-the-recommended-option). Also relevant: [Making Your Pricing and Offer Legible to an AI Recommender](/blog/making-your-pricing-and-offer-legible-to-an-ai-recommender), [The Service-Business Website Priority Stack: What to Fix First When Everything Needs Work](/blog/the-service-business-website-priority-stack-what-to-fix-first-when-everything-needs-work), [Attention Is Not a Lead: Why Service Businesses Confuse Audience With Pipeline](/blog/attention-is-not-a-lead-why-service-businesses-confuse-audience-with-pipeline).