I run lans.cloud, a collection of roughly 65 small, fast, privacy-first web tools. We're pre-launch on traffic — the pages are live, indexed, and clean, but the flywheel hasn't started spinning yet. That's actually the ideal moment to think hard about a question I kept dodging: when someone asks ChatGPT how to calculate their test grade, does my calculator show up in the answer, or does the model just do the math itself and move on?
That question is the whole reason this series exists. It has a name now — GEO — and the honest version of the story has two halves, one of which almost nobody is talking about. Let me lay out the terms first, then the framing this whole series is built on.
GEO, AEO, and the vocabulary you need
GEO stands for Generative Engine Optimization: optimizing your content so that AI answer engines — ChatGPT, Perplexity, Google's AI Overviews, Gemini — cite it, quote it, or lean on it when they generate an answer. You'll also see AEO (Answer Engine Optimization), which is the same idea under a different acronym. I use GEO throughout; treat them as synonyms.
The thing GEO is defined against is the SERP — the Search Engine Results Page. That's the classic list of ten blue links you get after a Google search: title, URL, snippet, repeat. For twenty years, SEO has been the craft of climbing that list. The SERP still exists and still matters. But an AI answer engine does something different: instead of handing you ten links and letting you pick, it reads a bunch of sources and synthesizes one answer, sometimes with citations, sometimes not. GEO is the craft of being one of the sources that answer is built from.
GEO is a layer on SEO, not a replacement for it
Here's the part the breathless "SEO is dead" takes get wrong. GEO is not a new game that replaces the old one. It's a second game played on the same board.
Every foundation that made you rank on the SERP still applies. The model's retrieval layer has to crawl you, so you need to be crawlable. It rewards fast, well-structured pages, so speed and clean HTML still matter. If your page was invisible to Googlebot, it's invisible to Perplexity's retriever too. This is exactly why our SEO guide is the prerequisite for this one — GEO builds directly on those crawlable foundations, it doesn't route around them.
What GEO adds on top is a different set of instincts:
- Citation-friendliness — content a model can lift a clean, self-contained claim from without dragging in three paragraphs of context.
- Front-loaded answers — the answer in the first sentence, not buried under a 600-word preamble about the history of grading.
- Machine-readable structure — the schema and markup that make your claims unambiguous to a parser, which is all of Part 2.
Notice the pattern: the brands winning GEO right now are, overwhelmingly, the ones who already had strong SEO. GEO is a multiplier on a foundation, not a shortcut past it. If you haven't done the boring work, there's nothing for the multiplier to multiply.
The two games a tools site is actually playing
This is the framing the rest of the series hangs on, and it's specific to what we are — a tools site, not a content site. There are two distinct games, and they reward completely different things.
Game A: content citation. Someone asks a model, "how do I calculate my test grade if I got 8 wrong out of 40?" The model explains the method, and ideally cites or links a page that walks through it. Here, we're competing as content — my test grade calculator page, with its explanation and worked examples, is up against every big education and homework-help site on the planet. This is the game everyone means when they say "GEO." It's real, and it's winnable in the right niches, but it's crowded.
Game B: tool invocation, a.k.a. the agentic web. Someone — or, increasingly, their agent — doesn't want an explanation. They need to actually flip a coin or generate a password, right now, as a step in some larger task. The model reaches for a tool, or recommends one, or (in an agentic setup) calls one. When an assistant decides it needs to hand a user a working coin flip or a real password generator, which tool does it reach for?
Game B is much closer to what a tools site actually is. We don't sell articles about coin flips; we sell the coin flip. And here's the striking part: almost nobody is deliberately optimizing for Game B. The whole industry is piled into Game A, writing citation-friendly prose, while the "which tool does the agent invoke" question sits wide open. That gap is the entire premise of Part 5, and it's the most interesting strategic bet we're making.
The honest strategic read: don't fight for the head
Whichever game you're playing, the terrain is the same as it always was, just steeper.
On head terms — "password generator," "coin flip," "dice roller" — the incumbents own the AI answer for the exact same reason they own the SERP: age, authority, and a mountain of links. A model asked for a password generator in the abstract will name the sites it has seen ten thousand times. You will not dislodge them by trying harder. Don't chase the head.
On the specific long tail, the math flips. Take a query with a real qualifier baked in — a fandom, an intent, a format: "classroom timer fullscreen no ads," "grade calculator how many wrong." That specificity is your opening. Our classroom timer exists precisely because "fullscreen, no ads, for a classroom" is a concrete need that the generic timers don't cleanly serve.
And here's the crucial bit that makes long-tail better under GEO than it ever was under classic SEO: on the SERP, ranking #4 for a long-tail query still gets you some clicks — ten links, ten chances. In an AI answer, the model synthesizes one answer. There's often no #4. Winning the retrieval slot doesn't mean placing well; it means you are the answer. The long tail is where AI results are far more differentiated than ten blue links, because the entire response collapses onto the handful of sources the model actually pulled. Win the retrieval, win the whole thing. Lose it, and you're invisible — there's no consolation traffic from page two.
So the strategy writes itself: concede the head, and go win specific long-tail queries where a young, fast, clean site can genuinely be the source a model chooses.
Where this series goes next
That's the map. The rest is how you actually do it.
- Part 2 — Structured data & on-page answers for AI. The schema, markup, and front-loaded page structure that make a model confident enough to cite you.
- Part 3 — llms.txt and the agentic web. The emerging conventions for telling AI crawlers and agents what you offer and how to use it.
- Part 4 — Measuring AI search. You can't tune what you can't see, and AI referrals are genuinely hard to measure. What actually works.
- Part 5 — The tool-invocation moat. Game B in full: how a tools site becomes the thing an agent reaches for, and why that's a deeper moat than any citation.
Foundations first. On to Part 2 →.