Somewhere between the search box and your website, a new gatekeeper appeared. Your next customer asks ChatGPT "who's a good family lawyer in Columbus?" — and acts on the answer. If the engines don't know your business, you were never in that conversation. Optimizing to be the answer is called GEO: Generative Engine Optimization. Here's how it actually works, minus the hype.
What GEO is (and what it isn't)
GEO is the practice of making your business easy for AI answer engines to understand, trust, and cite — across ChatGPT, Perplexity, Google's AI Overviews, and Gemini. It is not a new algorithm to trick, and it doesn't replace SEO. Think of it as the layer on top: SEO gets you eligible for classic rankings; GEO makes you quotable inside generated answers.
| Classic SEO | GEO | |
|---|---|---|
| Goal | Rank in the ten blue links + map pack | Be named, quoted, or linked inside AI answers |
| Who reads your site | Googlebot, searchers | LLM crawlers and the models themselves |
| Core signals | Relevance, authority, experience, speed | Entity clarity, consistency, quotability, third-party corroboration |
| Output | A position you can point at | A mention you can screenshot |
How AI engines decide who to cite
Generative engines assemble answers from sources they can parse and trust. In client work, four factors keep deciding who gets named:
- Machine-readability. Clean HTML, descriptive headings, valid schema markup (LocalBusiness, Service, FAQ), and an llms.txt file that states plainly what you do, where, and for whom.
- Entity consistency. Your name, address, phone, services, and descriptions must agree everywhere — site, Google Business Profile, directories, socials. Contradictions make the model hedge, and hedged models don't name you.
- Third-party corroboration. AI engines cross-check. Businesses that appear in the directories, listicles, and review sites the models read get corroborated; businesses that exist only on their own website get skipped.
- Quotable answers. Pages that answer a real question in one confident paragraph — pricing, comparisons, "how does X work" — give the model something safe to lift. Vague marketing copy gives it nothing.
The six-step starter plan
- Audit your AI visibility. Run your five money questions through ChatGPT, Perplexity, Gemini, and an AI Overview search. Note who gets cited and for what. That's your baseline.
- Fix entity basics. One consistent NAP, a plain-language About page, an Organization schema block with founder, area served, and contact info.
- Publish an llms.txt. A short file that introduces your business, services, and key pages in plain text — increasingly the front door for AI crawlers.
- Add schema where it matters. LocalBusiness/Organization, Service per offering, FAQ for the questions you just audited.
- Write the citable page. For each money question, one page that answers it directly in the first 100 words, then supports it.
- Track monthly. Re-run the prompts, log mentions by engine, and iterate on the sources that beat you.
How long does GEO take?
Faster than classic SEO in our experience — entity and schema fixes get picked up in weeks, and citation content in one to two months. But it compounds the same way: businesses that start now are building the corroboration trail the models will still be reading in two years. The cheap time to become the default answer was last year. The second-cheapest is this quarter.