When is GEO NOT worth doing?

When your product is not something anyone asks an assistant about — impulse purchases, purely local and immediate needs — or when your access and entity foundations are unresolved. In the latter case the work would sit on a bad base: machines first have to be able to get in and identify you unambiguously. And if you have no capacity to measure, it is also not worth it, because without measurement you cannot tell whether the spend did anything.

How is this different from SEO?

Much of the technical foundation is shared: crawlability, server-rendered content, unambiguous facts, consistent data. The difference is in the goal and the unit. Search optimization works for a position within a list of results; this works for being named inside a generated answer, where there is no tenth place — you are either in it or you are not. The detailed terminology comparison has its own page.

What does it cost, roughly?

We deliberately do not put a figure on this page, because prices are maintained in one place and a price repeated in several places eventually goes stale. The order of magnitude and what each package contains are on the pricing section of our GEO service page. What matters for the decision: the entry point is a measurement rather than a long commitment — first establish where you stand.

How long before there are results?

It depends which layer you fix, and this is the most useful practical fact. Retrieval-side fixes — crawlability, citable content — are measurable in days to weeks. What the model knows from memory refreshes only with a new model generation, where months is the unit. Anyone quoting a single turnaround time is conflating two different mechanisms.

Is one language enough?

If your buyers ask in one language, that one matters most, and competition there is usually far thinner. But languages build from different corpora and partly different sources, so presence in one does not arrive by itself from the other. If you also have an international market, the second language has to be written rather than translated.

Cookbook

Generative engine optimization explained — and when it is worth it

GEO is not a new button next to SEO: the unit of success is different, the feedback loop is different, and it does not pay off for every company.

Updated · August 9, 2026

Generative engine optimization — GEO in market usage — refers to work aimed at getting a brand to appear in the answers AI assistants give. Because the term is still young, two opposite exaggerations circulate: that there is nothing new in it and it is SEO renamed, or that it is a separate piece of magic that gets you into answers. Neither holds. What genuinely differs is the unit of success. In classic search optimization, position and clicks are the currency; here it is whether you get named — and those are not the same thing, because an answer does not contain ten results, only a few names. This page is not trying to restate the definition; the introductory level lives on our blog. What it answers is the question that comes before the work: when it makes sense to start, when it does not, and roughly what it involves.

When it is worth doing

When your buyers genuinely consult an assistant before deciding — typically true for considered, higher-value purchases: professional services, industrial procurement, software, healthcare, education. The second condition is having something to say: a method, measured results, published knowledge give a surface something to quote. The third is a sound technical base — your pages are crawlable, the content is present server-side, and your name resolves unambiguously to one thing. With those three in place, this work makes your existing content reachable on a new surface rather than building from nothing.

When it is NOT worth doing

When your product is an impulse purchase, or serves a purely local and immediate need — an assistant is not on that decision path. Also when access or entity foundations are unresolved: money spent here then sits on a bad base, and the correct order is crawlability first, unambiguous identity second. And, rarely said out loud, when you have no capacity to measure. Without measurement this becomes a matter of faith: you will not know whether a change came from your work, from a model refresh, or from ordinary variance. A fourth case: if your market is small enough that only a handful of people per month ask your questions at all, presence stays negligible even when everything is technically right.

What the work actually involves

Most of it is unglamorous. Measurement first: real buyer questions, repeated, per surface separately, so there is something to compare against. Then access and entity work — the part that most often produces change on its own, because a great many companies are simply not present when asked. Then content: quotable, sourced claims, saying the category words out loud, describing methods. And at the end the same measurement again on the same prompt set, because only a difference measured on an identical set says anything. What is NOT part of it: no secret markup, no keyword stuffing, and no file that gets you into answers.

What it costs, roughly

We deliberately do not print a figure here: prices are maintained in one place, because a price copied to several places eventually goes stale, and a stale price is worse than none. The order of magnitude and package contents live in the pricing section of our GEO service page. Two price-independent anchors are useful for the decision, though. First, the entry step is always a measurement rather than a long commitment — establish where you stand and whether there is anything to fix. Second, because return cannot be guaranteed, this work is rational where the value of a single won customer materially exceeds the cost of measuring and fixing. Where that ratio does not hold, it is better said upfront.

What it solves, and what it doesn't

This page is decision support: when it is worth doing and when it is not. What it does NOT offer is a guarantee. No amount of work compels an assistant to name your brand — answers are not deterministic, vendors publish no ranking rules, and anyone offering an inclusion guarantee is promising something with no mechanism behind it. What can responsibly be promised: a measurable baseline, identification of the gaps, specific fixes, and re-measurement on the same prompt set. If someone offers more than that — a position, a mention, "getting into AI" — it is worth asking what they measured it with and what happens when it does not land. An honest word on cost too: because return cannot be guaranteed, this work is rational at companies where customer value is high enough that a handful of good encounters repays it.

Frequently asked questions

When is GEO NOT worth doing?: When your product is not something anyone asks an assistant about — impulse purchases, purely local and immediate needs — or when your access and entity foundations are unresolved. In the latter case the work would sit on a bad base: machines first have to be able to get in and identify you unambiguously. And if you have no capacity to measure, it is also not worth it, because without measurement you cannot tell whether the spend did anything.
How is this different from SEO?: Much of the technical foundation is shared: crawlability, server-rendered content, unambiguous facts, consistent data. The difference is in the goal and the unit. Search optimization works for a position within a list of results; this works for being named inside a generated answer, where there is no tenth place — you are either in it or you are not. The detailed terminology comparison has its own page.
What does it cost, roughly?: We deliberately do not put a figure on this page, because prices are maintained in one place and a price repeated in several places eventually goes stale. The order of magnitude and what each package contains are on the pricing section of our GEO service page. What matters for the decision: the entry point is a measurement rather than a long commitment — first establish where you stand.
How long before there are results?: It depends which layer you fix, and this is the most useful practical fact. Retrieval-side fixes — crawlability, citable content — are measurable in days to weeks. What the model knows from memory refreshes only with a new model generation, where months is the unit. Anyone quoting a single turnaround time is conflating two different mechanisms.
Is one language enough?: If your buyers ask in one language, that one matters most, and competition there is usually far thinner. But languages build from different corpora and partly different sources, so presence in one does not arrive by itself from the other. If you also have an international market, the second language has to be written rather than translated.

Sources

Google — AI features and your website. https://developers.google.com/search/docs/appearance/ai-features
Perplexity — Crawlers (the retrieval side of sourced surfaces). https://docs.perplexity.ai/docs/resources/perplexity-crawlers