Michael D. Wandzilak
A growing share of your future customers will never see a page of Google results. They’ll ask ChatGPT, Perplexity or Google’s AI mode a question — “who’s a good estate attorney in Fort Wayne?”, “best med spa near me for facials?” — and get a short, confident answer naming two or three businesses. Generative Engine Optimization (GEO) is the work of becoming one of those names. It overlaps with SEO, but it is not the same job, and most businesses haven’t started it.
Where AI answers actually come from
Demystify this first, because everything practical follows from it. When an assistant recommends a business, the answer is assembled from two sources:
- Training data — what the model absorbed about the world when it was built: websites, articles, directories, reviews. Slow-moving, but it’s why established, widely-mentioned businesses get named “from memory.”
- Live retrieval — for anything current or local, assistants search the web in real time and read pages before answering. ChatGPT leans on Bing’s index and its own crawlers; Google’s AI features draw on Google’s index; Perplexity crawls aggressively itself.
Two consequences. First: traditional SEO still matters enormously, because assistants retrieve from the same indexes Google and Bing have always built. If you’re invisible in search, you’re invisible to retrieval. Second: there’s a new, purely technical failure mode — if AI crawlers can’t read your site, you can’t be retrieved at all, no matter how good your reputation is. We’ve found exactly that on a live client site: a firewall setting silently blocking every AI crawler while the owner wondered why ChatGPT didn’t know them. Check that first. It takes five minutes.
The GEO checklist, in priority order
1. Make sure AI systems can read your site
GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot, PerplexityBot and Google-Extended all identify themselves when they visit. Security tools — Cloudflare’s AI-bot blocking in particular — often block them by default, and a robots.txt that says Allow doesn’t override a firewall that says 403. Our step-by-step check shows you how to test every major crawler against your own site in one terminal command each.
2. Tell machines exactly what you are with entity schema
Structured data (Schema.org JSON-LD) is how a machine learns, unambiguously, that your site represents a business — with a name, a location, hours, services, a founder, social profiles. On every site we build, a connected @graph of LocalBusiness (or the specific subtype), Person and Service nodes ships in the head of every page, each fact matching what the page visibly says. Retrieval systems consume this directly; it’s the difference between an assistant parsing prose about you and an assistant knowing facts about you.
3. Publish content that answers questions in plain language
Generative engines quote and synthesize; they reward pages that state things clearly. The formula is unglamorous: one question per page or section, a direct answer in the first sentences, specifics and numbers over adjectives, and a visible FAQ where the questions are real ones. (Notice that this is also just… good writing for humans. The overlap is not a coincidence — assistants were trained on what humans found useful.)
4. Get mentioned in places models trust
Assistants weigh corroboration: reviews on Google and industry platforms, local directories with consistent details, press, professional associations, supplier and partner pages. A business mentioned consistently across twenty independent sources is a safe recommendation for a cautious machine; a business that exists only on its own website is not. This is the GEO version of link-building, and reputation you already have counts — make sure your name, address and details match everywhere.
5. Keep your facts consistent
Conflicting hours, an old address on a directory, three different phone formats — a human shrugs, a retrieval system loses confidence. An annual consistency sweep across your listings is boring and pays off.
6. Add the emerging plumbing: llms.txt
A proposed standard gives AI systems a curated, markdown map of your site. Adoption by the big providers is still unproven — we’re candid about that in our llms.txt guide — but it costs minutes, carries zero risk, and positions you for the agents that already read it.
7. Measure it
Ask the assistants your customers’ questions, monthly, and record the answers. Watch your analytics for referrers like chatgpt.com and perplexity.ai and for the crawlers in your logs. It’s early, the tooling is immature, and anyone promising you a “#1 ranking in ChatGPT” is selling something — but trend lines are already visible and worth tracking.
GEO vs SEO vs AEO — the terms, quickly
SEO optimizes for ranked lists of links. AEO (Answer Engine Optimization) targets direct-answer surfaces like featured snippets and voice. GEO targets generative assistants that compose recommendations. The labels get thrown around interchangeably, and honestly, the work overlaps heavily — the pillars above serve all three. What’s genuinely new in GEO is the crawler-access layer and the citation/corroboration emphasis.
Where to start
Run the crawler-access check today — it’s the only item on the list where a single hidden setting can zero out everything else. Then schema, then content. If you’d rather have it done for you, this is now a standing part of our SEO & performance work: we audit AI-crawler access, build the entity graph, and set up the measurement — the same process we run on our own site.
Frequently asked questions
What is Generative Engine Optimization in one sentence?
GEO is the practice of making your business visible, readable and credible to AI assistants like ChatGPT, Perplexity and Google’s AI mode, so they name you when people ask for what you do.
Is GEO different from SEO?
They overlap heavily — assistants retrieve from the same search indexes SEO targets. GEO adds layers SEO ignores: AI-crawler access, entity schema depth, cross-web corroboration of your facts, and measurement inside the assistants themselves.
Can I pay to appear in ChatGPT's recommendations?
No. There’s no ad placement inside organic assistant recommendations today. Anyone selling guaranteed AI rankings is overpromising — the honest work is access, structure, content and reputation.
How long does GEO take to show results?
Technical fixes (crawler access, schema) can affect live-retrieval answers within days to weeks. Reputation signals move on the scale of months, and training-data presence updates only when models are retrained. Start now precisely because of that lag.