Hello

Pages

Search

Start a project

GEO — Generative Engine Optimisation

Known by
the models

GEO · Part of search

You cannot rank in a generated answer. You can be the thing it is built from.

Our work has been published by 50 outlets

ForbesElleNational GeographicEurovisionBusiness InsiderGreenpeaceCloudflareMarriott
See the full record

When a buyer asks an assistant who to hire, the answer is assembled from what the model holds about you and what it can retrieve in the moment. Neither is a ranking. GEO is the work of being a clearly defined entity, described consistently everywhere, with material a model can quote without risk — and of knowing, measurably, whether you come up at all.

ChatGPTGeminiClaudeEntity workllms.txtShare of model voice

GEO

Being known, quoted and recommended inside ChatGPT, Gemini and Claude.

A generative engine does not have a results page to put you on. It has a representation of your business — assembled from your site, from third-party sources, and from whatever it retrieves while answering — and it will decline to assert anything it cannot corroborate. So the first job is to make that representation unambiguous: one canonical description, consistent facts, properly marked-up entities, and agreement between what your site says and what the directories, profiles and press say.

The second job is to be quotable. Models reproduce self-contained passages: a definition that stands alone, a figure with its source, a comparison laid out plainly. Pages written as continuous persuasion give a model nothing it can safely lift. We rewrite the commercially important ones so they can be retrieved in pieces.

The third job is access and measurement. Crawler policy decides whether the retrieval-time platforms can read you at all. And because none of this appears in a rank tracker, we benchmark against a fixed set of buying prompts across the major assistants — how often you appear, in what company, and what the model gets wrong. That misstatement list is usually the most useful output of the first month.

01

Entity

One unambiguous definition of who you are, what you sell and where — marked up on the site and matched everywhere else the web describes you.

02

Corpus

Material worth quoting, written so it can be lifted in pieces without losing its meaning or its attribution.

03

Access

Crawler and retrieval policy set deliberately: which AI agents may read you, what they get, and what is held back.

04

Benchmarking

A fixed set of buying prompts, run across the major assistants on a schedule — appearance rate, competitive set, and what the models get wrong.

Capabilities

Everything a
GEO brief covers

The full scope, listed. Engage all of it or the two or three lines that are actually holding you back — most programmes start narrow and widen once the first results justify it.

01 Entity definition: Organization, Person, Product and Service schema
02 Fact-consistency audit across your site, directories, profiles and press
03 Canonical descriptions deployed identically everywhere you appear
04 Wikidata and authoritative-source presence where the business qualifies
05 Retrieval-friendly rewriting: self-contained, chunkable, quotable passages
06 llms.txt and an explicit policy for AI crawlers
07 GPTBot, ClaudeBot, PerplexityBot and Google-Extended access configuration
08 Prompt-set benchmarking across ChatGPT, Gemini, Claude and Perplexity
09 Share-of-model-voice tracking against named competitors
10 Citation-source analysis: which pages and third parties models actually pull
11 Third-party corpus work: reviews, directories, trade press and listings
12 Misstatement remediation when a model has you factually wrong

Common questions

GEO, the
questions that matter

GEO — the next step

Let’s talk
about GEO

Send the domain and what you are trying to rank, be cited or be recommended for. You will get back a candid read on where GEO would and would not move the number — not a template proposal.

Discuss GEO
WhatsApp