How to rank on ChatGPT: the contractor's GEO guide

You can't "rank" on ChatGPT the way you rank on Google, because there's no page one. When a customer asks who to hire, the engine writes an answer that names a short list of companies, and everyone else is invisible. Getting into that answer is called GEO (generative engine optimization), and for a local contractor it comes down to five moves: be retrievable, be liftable, be corroborated, be consistent, and be measured. This guide covers all five.

For local hiring questions, ChatGPT runs a live web search, retrieves a handful of pages it trusts, and writes an answer grounded in them. So every recommendation is two decisions: which pages get retrieved, and which companies those pages support. The retrieval skews third-party: Ahrefs found 88% of ChatGPT's cited URLs come from its general web-search retrieval rather than from news, Reddit, YouTube, or academic sources, and Yext's study of 6.8 million citations found 48.73% of ChatGPT's citations point at sites like Yelp, with directory sources rising to 46.3% for subjective "best X near me" queries. You influence the directory bucket, and you fully control your own pages.

Publish the pages the engines are starving for: a priced cost guide for each major service in each city you serve, and a real service page per city. Directories dominate AI citations for local services, but no directory publishes what a job costs in your city; a priced local guide is information only you can supply, and engines retrieve it for exactly the cost questions directories can't answer. Structure matters as much as substance. Phrase headings as the questions customers type, and answer each one directly in the first sentence or two of the section, because engines lift self-contained passages, not whole pages.

• Lead every section with the answer: a direct, 40–60 word response to the heading's question, then the supporting detail.
• Use real numbers. "Tile roof replacement in Miami runs $18,000–$35,000" is citable; "costs vary" is not.
• Name your cities and neighborhoods. Engines resolve "in Fort Lauderdale" literally, and a page that says the words beats one that doesn't.
• Keep dates visible and current. Engines discount stale pages, so update your guides and show the updated date.
• Add FAQ and Article schema (JSON-LD) so the engines can parse your structure without guessing.

Two ways. First, citations: when ChatGPT, Perplexity, or an AI Overview cites your cost guide as a source, that link carries real referral traffic from high-intent buyers. Second, the mention itself: a customer who's told "call these three companies" searches for you by name next, so branded search and direct calls rise with AI visibility even when no link is clicked. Both start from the same work: liftable pages plus a trust layer the engines can verify.

Engines minimize the risk of a wrong recommendation by looking for consensus across independent sources, and the platforms carry serious citation weight: in Local Dominator's analysis of 267,280 AI citations, Yelp (71,512), Google (55,977), and Reddit (36,678) led the field, with Facebook, Angi, YouTube, BBB, and HomeAdvisor behind them. Reddit matters far more than its size suggests, because Perplexity draws 46.7% of its top citations from it (Profound). A company that appears on its own site, on Google and Yelp with fresh reviews, on Angi and the BBB with matching details, and in a community thread reads as a safe answer. So keep review velocity up, keep listings consistent, and earn a few genuine third-party mentions rather than many thin ones.

Not by typing your best question into ChatGPT once. Answers change run to run: the same question produces different company lists on different draws, and each engine keeps its own list because each trusts different sources. Real measurement means a fixed set of buying questions, asked repeatedly on every engine your customers use, tracked as a frequency over time: your share of answers, your average position when named, and the sentiment of how you're described. That's the loop a GEO tool automates.

1. Week 1, baseline: list your 20–50 buying questions, run them across engines (or run our free report), and record who gets named.
2. Week 2, publish: one priced cost guide for your highest-value service, question-formatted headings, real numbers, visible dates, FAQ schema.
3. Week 3, trust layer: fix name/phone/service-area mismatches across your site, Google Business Profile, and directories; restart review velocity with a simple post-job ask.
4. Week 4, corroboration: one genuine community contribution (a helpful answer in a local or trade thread) and one local mention (supplier, association, or news).
5. Ongoing: re-measure weekly and let the per-engine deltas tell you what to publish next.