Guide · GEO
How to be the answer.
Your buyers ask ChatGPT, Perplexity and Google before they ask you, and read one answer assembled from sources they never visit. Here is how to become a source those engines quote, what actually moves it, and what does not.
This guide, in 6 parts
The click is being replaced by the answer.
For twenty years the work was to rank: be the link Google listed and earn the click. That outcome is eroding in plain sight. When an AI summary appears, Pew found Google users click a result on roughly 8% of visits, against 15% when there is no summary, and they click a source inside the summary just 1% of the time.
The buyer reads one answer, assembled from sources they never open. If the machine does not name you, you were not in the room.
We say this as operators who fought it, not analysts who read about it. We led the defense against an AI-Overviews traffic collapse inside a business with tens of millions of annual visits, and we built this site to be cited too. This guide is the method we use. It is also our Be the Answer work, written down.
« Being cited is not a click. It is the recommendation the buyer trusts before they ever reach you. »
Engines cite what they can identify, quote and trust.
A generative engine does not rank pages and hand you a list. It reads a question, retrieves passages it considers relevant and reliable, and writes one answer, naming the sources it leaned on. So the unit of competition is no longer the page, it is the passage, and the prize is no longer the click, it is the citation. Three things decide whether that citation is yours.
Identity. The engine has to resolve you to a single, unambiguous entity before it will recommend you. Two trading names, a category that reads differently on three pages, and no structured data linking your site to your external profiles, and the model cannot be confident who you are. Confusion does not get cited.
Shape. The engine lifts passages that stand on their own. Content written as a buyer's question with the answer in the first two sentences, in self-contained paragraphs, is content it can quote. AI Overviews are short, roughly 90 to 120 words by Semrush's measurement of 200,000 of them, and dominated by how and what questions. Long preambles and buried answers do not survive that compression.
Substance. The engine favours sources that say something specific and verifiable. The peer-reviewed GEO study (Aggarwal et al., KDD 2024) tested this directly: adding citations, quotations and statistics raised a source's visibility in AI answers by up to 40%, while keyword stuffing, the reflex of old SEO, performed poorly and sometimes hurt. Precision earns the citation. Volume does not.
What to change on the page.
Concrete edits, grouped by what they fix. Not principles, changes you can make this week.
Entity clarity, so the engine knows who you are
- One canonical name for your company, used identically everywhere. Pick the legal or the trading name and stop alternating between them.
- An About page that states, in plain sentences, what you do, for whom, and where. Not a mission statement. A definition the model can lift.
- Organization and sameAs structured data that connects your site to your LinkedIn, Crunchbase and registry profiles, so the engine resolves you to one entity, not three.
- Consistent facts across the web. If your category, founding, or location reads differently on three pages, the model cannot trust any of them.
Answer-shaped content, so the engine can quote you
- Headings written as the question a buyer asks, then the answer in the first two sentences below it. Lead with the answer, support it after.
- A specific claim with a number, a date, or a named source in the opening lines. Generative engines lift sentences that can stand alone.
- Self-contained paragraphs. A model extracts a passage, not your page. If a sentence only makes sense after three others, it will not travel.
- Coverage of how and what questions in your category. AI Overviews are short, roughly 90 to 120 words, and dominated by how and what queries.
Specificity over volume, so you earn the citation
- One original number a competitor cannot copy: your own benchmark, survey, or operating data. Cited sources travel further than clever ones.
- Quotations and named references, attributed. The GEO study found citations and quotations among the strongest moves; stuffing keywords among the weakest.
- Structured data that mirrors the visible text exactly. Schema that claims more than the page shows is a flag, not a boost.
- Fewer, deeper pages over many thin ones. Volume was the old game. Being the most precise answer to a real question is the new one.
One paragraph, before and after.
The single highest-leverage change is shape. Here is a typical B2B paragraph, then the same facts rewritten so an engine can lift the answer, identify the source, and trust the claim. Nothing was invented. It was reordered and made specific.
Before · written to read
In today's fast-moving landscape, our innovative platform leverages cutting-edge technology to help businesses of all sizes unlock their full potential and drive meaningful results across the entire customer journey. We are passionate about delivering best-in-class solutions tailored to your needs.
No question answered. No entity stated. No number. Nothing here can be quoted, and nothing can be trusted.
After · written to be cited
How long does warehouse automation take to pay back? For mid-market retailers, a goods-to-person system typically reaches payback in 18 to 30 months. In our 2025 deployment for a European retailer, picking throughput rose 42% and error rates fell by half within the first quarter. The variables that decide the timeline are order volume, SKU count, and how clean the inventory data is on day one.
A buyer's question, answered first. A specific, original number. Sentences that stand alone. Citable, and checkable.
The honest caveat: the after version only works if the numbers are real and yours. Inventing a statistic to look citable is the fastest way to lose trust with a buyer who checks, and increasingly with engines that cross-reference. Specificity is a commitment to having done the work, not a writing trick.
GEO is not a hack, and three metrics will mislead you.
Before the method, the warning. Generative engine optimization is being sold as a switch you flip. It is not. It is entity work, content work and measurement work, compounding over months, and anyone promising citation by next week is describing something that will not hold. Worse, the dashboards you already trust will tell you the wrong story while it happens.
Keyword rankings
You can hold position one and still be omitted from the AI summary sitting above it, which is exactly where the click is lost. Rank now measures a race that fewer buyers are running.
Organic sessions
Traffic understates both the damage and the progress. A buyer can read your answer, your name, and your recommendation inside ChatGPT, decide you are the firm, and arrive later by typing your name. Your analytics will record none of the influence, only the eventual direct visit.
Impressions and clicks in Search Console
These count Google's traditional surface. They are blind to whether ChatGPT or Perplexity quoted you at all. Optimising to a number that cannot see most of the new battlefield is how teams declare victory while losing.
The metric that does not lie is share of answer: for the questions your buyers actually ask, are you named, quoted and recommended across the engines, and is that improving. It is harder to measure and it is the only one that reflects where the decision now happens.
Frequently asked.
How do I get cited by ChatGPT?
Make your company easy to identify, easy to quote, and worth quoting. In practice: one canonical name and Organization structured data so the model resolves you to a single entity; headings written as buyer questions with the answer in the first two sentences; and at least one specific, original fact per page, a number or a named source it can lift.
The peer-reviewed GEO study found citations, quotations and statistics lifted a source's visibility in AI answers by up to 40%, while keyword stuffing performed poorly. There is no submission form. You earn it by being the clearest, most citable answer to a real question.
Is GEO just SEO with a new name?
No, though they share foundations. SEO optimized to be the link a buyer clicks. GEO optimizes to be the source an AI quotes when the buyer never clicks at all. Pew found that on Google pages with an AI summary, a source inside that summary is clicked just 1% of the time.
So the work shifts from ranking a page to making a passage quotable, from keywords to entity clarity, and from traffic to share of answer. Much of good SEO still helps. But the goal, and the measurement, are different.
How long does it take to get cited?
Longer than a campaign, shorter than a rebrand. This is not a two-week hack. Entity signals and content changes take time to be recrawled, re-embedded and reflected in answers, and engines vary in how fast they update. Expect movement on specific questions over weeks, and a durable position over a few months of consistent work. Anyone promising citation by Friday is selling you the version that does not last.
We rank well on Google already. Are we safe?
Not necessarily. Ranking and being cited are now two different outcomes. A page can hold position one and still be skipped by the AI summary above it, which is exactly where the click is being lost. Gartner expects traditional search volume to fall about 25% by 2026 as buyers ask assistants instead. Strong rankings are an asset to build on, not a reason to stop looking.
What should we measure if clicks are disappearing?
Share of answer: for the questions your buyers actually ask, are you named, quoted and recommended across ChatGPT, Perplexity and Google AI Overviews, and how does that change over time. Sessions and rankings will understate the damage and the progress, because the value now happens inside an answer the buyer reads without visiting you. Measure presence in the answer, not just traffic to the page.
Being cited is the start of the journey, not the end of it. Once the engine sends a buyer your way, what they meet next decides whether the citation converts. We map that whole path with a framework we call the Hub Map, and the better question is usually not "how do we get cited" but "what happens after we are." We do not publish the Hub Map.
You can run much of this yourself. The entity fixes, the answer-shaped rewrites and the schema are all in this guide, and a capable team can ship them. Bring us in when organic traffic is eroding faster than you can rewrite, when you need to know exactly what the engines say about you and your category today, or when a competitor is being cited in your place and you want it to be you.
We have done this under fire, and we stay until it holds. See Be the Answer.
Want to know what the machines say about you?
No deck, no gate. A working session on your AI visibility, with the operators who would fix it.