Your buyers have stopped scrolling through ten blue links. They ask ChatGPT, Google AI Overviews, Perplexity, Claude or Copilot who to hire, and they act on the two or three companies it names. GEO is the work that makes you one of them.
Who is the best supplier for this in my area? Give me two or three companies you would actually recommend.
Based on what I can verify, I would look at a competitor you know first. Their site sets out their services, service area and pricing clearly, and their customer reviews are consistent.
A second option is a company founded three years ago, which publishes detailed answers to exactly this question.
Your company was not mentioned, and no page of yours was cited.
Most companies have never checked, and a surprising number are invisible to AI answer engines without knowing it. Put in your address and get a GEO score out of 100, with the eight things to fix in order.
We read the way an AI engine does: crawler access, structured data, answers, entity signals. The report with your score has been sent to . Check your spam folder if it does not show up within a few minutes.
While you wait, the eight signals we score are explained below, and so is what each band of the score actually means.
See what we measure →+ 4 more categories in the full report
These are the kinds of sentences your buyers type. Not keywords. Whole questions, with a decision attached to the answer.
This is the part that catches established companies off guard. Brand equity has always been a moat. Inside a generated answer, it is not, because almost none of it exists in a form a machine can read.
Decades of word of mouth, a name everyone in the province recognises, a sign on the busiest road in town. A model cannot see any of it. It reads pages. If your reputation lives in people’s heads and your website is a brochure from 2018, you are starting from zero against companies that started last year.
In most cases the block is accidental: a security plugin, a bot filter at the CDN, a site that only renders once JavaScript runs. The company never decided to keep AI crawlers out. It simply never checked, and the engines quietly moved on to a source they could fetch.
Every question a competitor answers first becomes a source for the next related question. Their citation share grows while yours stays at nothing. Catching up in a year means arguing against an answer the engine has already settled on, which is far harder than filling an empty slot today.
A model cannot ask around. It can only read. Everything you are famous for offline counts for nothing until it is written down where the engines look.
GEO does not replace SEO and it is not a separate budget. AI Overviews and AI Mode run on the same index as classic search, and the assistants lean on web search to retrieve. Good SEO puts you in the pool. GEO decides whether you are the passage that gets quoted out of it.
| Classic SEO | GEO | |
|---|---|---|
| What you compete for | A position in a list of links | Being the source that gets named |
| What competes | Page against page | Passage against passage |
| What the buyer sees | Ten results to compare | One answer, one to three companies |
| Room at the top | The top three split the clicks | There is no second page |
| What it rewards | Keywords, links, domain authority | Clarity, structure, verifiable facts |
| How the buyer lands | Early, still comparing | Late, already shortlisted |
| How you measure it | Rankings and impressions | Prompt coverage and citation share |
They differ in how they retrieve and how they cite, but what they need from you overlaps almost completely: pages they can fetch, structure they can parse, facts they can verify.
OAI-SearchBot, GPTBot, ChatGPT-UserGooglebot, Google-ExtendedPerplexityBotClaudeBot, Claude-SearchBotBingbotGoogle-ExtendedThis is exactly what the free scan scores, and exactly what we fix in a GEO engagement. No secret sauce: the engines are fairly open about what they need, and almost nobody supplies it.
Whether robots.txt, your firewall or your CDN quietly blocks OAI-SearchBot, GPTBot, ClaudeBot, PerplexityBot, Googlebot or Bingbot.
Why it decides the answer. An engine that cannot fetch your pages cannot cite them. This is the most common reason a company is missing from answers, and usually nobody chose it: a security plugin or a bot filter did.
Which schema.org types you publish: Organization, LocalBusiness, Service, FAQPage, BreadcrumbList, Product, Review.
Why it decides the answer. Structured data is the one place where you state the facts about your company in a form a machine cannot misread. Without it an engine has to infer them from prose, and it infers conservatively.
Whether your pages contain question-shaped headings with short, self-contained answers that survive being lifted out of the page.
Why it decides the answer. Generative engines quote passages, not pages. A paragraph that only makes sense after reading the two above it will never be the quote.
Consistency of your company name, address, phone number, founding date, services and profiles across your own site and the wider web.
Why it decides the answer. An engine has to be confident that the company on your site is the same one it has read about elsewhere. Contradictions lower that confidence, and low confidence means it names someone else.
Whether your content is in the HTML that gets served or only appears after JavaScript runs, plus response time and redirect chains.
Why it decides the answer. Several AI crawlers do not execute JavaScript. If your text arrives after the script does, they fetch an empty page and move on.
Titles, meta descriptions, canonical tags, Open Graph data and heading hierarchy.
Why it decides the answer. This is the raw material for how a model summarizes you in one line. Leave it to chance and that line will be written from someone else’s wording.
Dates, named authors, cases with numbers, reviews, and how recently the pages that matter actually changed.
Why it decides the answer. Models lean on sources they can verify and date. Undated, unattributed marketing copy reads as unverifiable, and unverifiable sources get left out.
Whether you publish an llms.txt that tells AI systems what your company is, which pages matter and how to describe you.
Why it decides the answer. It is new, it is cheap, and almost nobody in the average category has one. It costs an afternoon and it is the closest thing there is to writing your own entry.
Eight groups, 100 points in total. The free scan scores all eight.
It is a diagnostic, not a ranking promise. Nobody controls what a generative engine decides to say. What you can control is whether it is able to read you, understand you and verify you, and that is what the number measures.
The engines either cannot read you or cannot verify you. In most answer sets for your category a competitor gets named and you are not mentioned at all. Everything here is fixable, and the first fixes are usually technical and fast.
You are readable, so you turn up when someone asks for you by name. You are not the answer when someone asks for your category, because nothing on your site is shaped like an answer. This is where most established companies sit today.
You get read, understood and cited. The work shifts from catching up to holding the position: new questions, fresh proof, and watching the prompts where a competitor starts appearing next to you.
Half of it is unglamorous technical repair that moves the score within weeks. The other half is publishing answers good enough to be worth quoting, which is slower and is where the lasting advantage sits.
We write down the questions your buyers actually type into an assistant, per market and per language. Not keywords: sentences. Usually between 150 and 400 of them.
We run every prompt across the engines and record what comes back: who gets named, which pages get cited, and where you do not appear at all. That is your starting citation share.
Crawler access, structured data, server-rendered HTML, canonicals, llms.txt. The unglamorous half, and the half that moves the score fastest.
The pages that deserve the citation: one clear question per page, the answer inside the first hundred words, the evidence underneath it.
Your facts made identical everywhere a model reads them: your own site, your profiles, directories, press, partner pages. Confidence is a ranking factor here.
Named experts, dated cases, real numbers, reviews. The material that turns a claim into something a model will repeat with your name attached to it.
The full prompt set re-runs on a schedule. You get citation share over time, per engine, next to the competitors who appear beside you.
Every question you answer first becomes the source for the next related question. The lead widens, which is exactly why starting late is expensive.
Book your free GEO audit. We run your category questions across ChatGPT, Google AI Overviews, Perplexity, Claude and Copilot while you watch, and show you who gets named instead of you and why. You keep the findings whether you work with us or not.
We take on a limited number of GEO clients per month, because the work is hands-on.
Fill in the form and we’ll get back to you within one business day.
We’ll get back to you within one business day to plan your GEO audit.