GEO is Generative Engine Optimisation: the practice of structuring a business so that AI systems cite and recommend it when somebody asks a relevant question. Where SEO aims to rank a page in a list of links, GEO aims to have the business named inside the answer itself.
That distinction sounds small and it changes the economics completely. A ranked list has a position ten. A generated answer names two or three sources, and everything else is not lower down, it is absent.
What Changed, and Why It Matters Commercially
For twenty-five years the deal was the same: search engines returned links, and the job was to be high in that list. The user chose from ten options and clicked.
Increasingly they do not. They ask a question and receive an answer with two or three businesses named in it. There is no list to be tenth in.
This changes the shape of the loss. In classic search, poor visibility means a small share of clicks. In AI search, poor visibility means the buyer never learns you exist, and there is no impression to count, no lost bid to review, and nothing in your analytics that says a decision happened without you. It is the quietest possible way to lose a pipeline.
The flip side is that being named carries an implicit endorsement a ranked link never had. The assistant is not showing options, it is answering. Being the answer is worth more than being an option.
How GEO Differs from SEO
| Traditional SEO | GEO | |
|---|---|---|
| Objective | Rank in the top 10 | Be named in the answer |
| Output | A list of links | A synthesised answer with citations |
| Dominant signal | Backlinks and keyword relevance | Entity clarity and corroboration |
| Content shape | Comprehensive, keyword-targeted | Answer-first, specific, extractable |
| Measurement | Position, traffic, CTR | Citation share by platform and query |
| Failure mode | Page three | Absent, with no position to improve |
| Compounding | Links accumulate | Corroboration accumulates, more slowly |
The honest caveat, which most agencies leave out: roughly eighty per cent of the underlying work is shared with good SEO. Entity clarity, structured data, technical accessibility, credible third-party mentions. None of that is new.
What is new is the twenty per cent, and it decides outcomes. A site with excellent classic SEO and an entity no model can resolve will rank and never be cited.
The Four Signals That Decide Citations
1. Entity authority
Whether a model can confidently identify your business, and whether it associates you with a subject. These are two separate things and they fail differently.
Identification is where most problems live and where almost nobody looks. A company trading under three variants of its own name, described differently on its website, its LinkedIn and its Companies House filing, is not one clear entity. It is a cluster of weak signals that may or may not be the same organisation, and a model that cannot resolve you will generally decline to name you.
Association is whether that entity is linked to a topic. You can be perfectly identifiable and never cited, because nothing connects you to the subject.
Detail: what is entity authority · entity signals checklist
2. Claim density
Models cite content that commits to something checkable. “We have extensive experience in this area” cannot be quoted. “We have advised on 40 transactions in this sector since 2019” can.
This is the habit with the highest return and the lowest cost, because it does not require new content. It requires editing existing content to say something specific. Most professional services copy is vague not because the firm lacks specifics but because vagueness feels safer to sign off.
3. Source architecture
Whether an answer can be lifted cleanly from your page. The answer in the first sixty words, question-shaped headings, one idea per section, structured data that matches what is visible.
A page that builds an argument across five paragraphs before arriving at the point contributes nothing liftable, however good the argument. That is the single most common reason good content goes uncited.
Detail: how to write content AI models cite · schema guide
4. Cross-platform consistency
The same facts about you, everywhere. Site, profiles, directories, filings, professional bodies, review platforms. Inconsistency here does not merely fail to help, it actively weakens everything else, because it gives a model reasons to doubt what it has assembled.
Where the Real Work Is
An uncomfortable truth about the four signals: the first three are largely within your control and can be fixed in weeks. The fourth is partly outside it. And the thing that most decides competitive outcomes is not on the list at all, because it does not happen on your website.
Third-party corroboration. Independent, credible sources saying things about you, on the domains a model already draws on for your category. Trade press, professional bodies, industry directories, genuine expert commentary, review platforms.
This is the slow, expensive, unglamorous half. It is also the half that separates a firm that is identifiable from a firm that gets recommended, and it is the first thing to quietly disappear from an engagement when an agency is stretched, because content is visible and easy to report while placements are hard.
If you take one diagnostic from this page: ask any agency to list every third-party mention they earned you last quarter. The answer tells you what you are actually buying.
Who GEO Is For
The variable is not your sector. It is how much research happens before somebody contacts you.
Strongest fit. Considered, comparison-heavy purchases where a buyer builds a shortlist before speaking to anyone. B2B services, professional services, regulated sectors, specialist suppliers, higher-value consumer decisions.
Weakest fit. Impulse purchases, pure proximity trade, and anything bought without research. Nobody asks an assistant which corner shop to walk into.
The middle, where most of the confusion is. A local business with a genuine specialism is more exposed than a generalist, because the specialist query is exactly what an assistant answers well. “Emergency plumber near me” is a maps query. “Which UK plumbers are certified for unvented cylinder work in listed buildings” is an AI query, and it is worth considerably more.
Detail: who needs GEO · is GEO only relevant for B2B
By sector: SaaS · financial services · legal · healthcare · e-commerce · insurance · manufacturing · logistics · higher education
What It Costs
| Step | Investment | Term |
|---|---|---|
| Free AI Visibility Audit | Free | None |
| Synaptic Audit | £2,950 one-off, credited against your first retainer month | None |
| Foundation | £2,950 a month | 3 months |
| Authority | £5,950 a month (£6,500 regulated) | 12 months |
| Dominance | From £12,950 a month | 12 months |
We publish these because most of the market does not. Of the seven UK agencies we track, only two publish any figures at all, which makes honest comparison harder than it should be.
More: GEO pricing guide · what is a fair monthly price · GEO versus PPC spend
How to Start
- Measure a baseline. Ask the assistants the questions your buyers ask. Record who is named, and who is named instead of you. Without this you cannot judge anything later.
- Fix the entity. Consistent naming,
Organizationschema withsameAs, one canonical statement of the facts. Ours is at /entity-map/, and the schema generator is free. - Make the pages that matter extractable. Answer first, specifics over adjectives, question-shaped headings.
- Confirm the crawlers can reach you. A permissive
robots.txtproves nothing if a firewall is returning 403. See the AI crawler guide. - Then build corroboration, on the sources already cited in your category.
- Report citation share, not just clicks. Otherwise you will grade the work on the wrong exam.
Steps one to four are genuinely doable in-house. Step five is where most firms need help.
Honest alternatives: can I do GEO myself · in-house versus agency · how much of my team’s time it takes
What a Programme Actually Looks Like
Buyers consistently ask what happens in the months they are paying for before anything visible moves. The honest shape:
Month one. Baseline measurement and entity work. Which competitors are named for your queries, what a model currently believes about you, and fixing the naming, schema and consistency problems that block everything downstream. Almost nothing visible happens in citation terms, and that is correct rather than concerning.
Months two and three. Extractability across the pages that matter, and the beginning of corroboration. This is where structural wins appear: pages that were retrieved but never used start being used, because there is finally something liftable in them.
Months four to six. The inflection, if it comes. Corroboration has had time to register and citation share on your priority queries starts to move. This is the point the first three months were building toward, and it is also the point at which most abandoned programmes were abandoned just before.
Beyond six months. Compounding, and defence. Positions won have to be held, because competitors publishing into the same queries erode them.
The commercial expectation to set internally: structural wins in weeks, citation share in three to six months, and the compounding only after that. Detail: what a GEO programme looks like week by week · GEO results timeline
What GEO Will Not Do
Worth stating, because the category attracts overclaiming.
It will not produce guaranteed citations. Nobody controls model outputs. Any guarantee is a promise about somebody else’s system. See do you guarantee results.
It will not win head terms held by DA 80+ publishers quickly. For a specialist firm the winnable ground is specific, and specific queries convert better anyway.
It will not fix a weak proposition. Being cited more often for something people do not want is not progress.
It will not work in three months. Structural wins, yes. Citation share on contested queries, no. Stopping at three months is the most expensive point to stop, because you have bought the foundations and skipped the compounding. See what happens if I pause after three months.
GEO: Common Questions
What is GEO in digital marketing?
GEO is Generative Engine Optimisation: the practice of structuring a business so AI systems cite and recommend it when somebody asks a relevant question. Where SEO aims to rank a page in a list of links, GEO aims to have the business named inside the answer itself, on ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews.
Is GEO replacing SEO?
No, it is being layered on top. AI Overviews are built on Google’s ordinary index, and analyses consistently find a large majority of AI-cited URLs already rank on page one. Strong SEO is close to a precondition for AI citation rather than an alternative to it. What is changing is that ranking is no longer sufficient on its own.
How quickly does GEO work?
Structural wins can surface in weeks, because fixing an entity a model cannot resolve or content it cannot extract removes something actively blocking you. Citation authority typically takes three to six months, because it depends on independent sources corroborating you and that accumulates on its own schedule. Anyone promising citations in thirty days is describing something outside their control.
Do I need different content for GEO and SEO?
Usually not different content, but differently structured content. The same expertise works for both if the answer comes first, the claims are specific enough to quote, and the page is marked up properly. Most firms find restructuring existing pages delivers faster than commissioning new ones, because the hard part, having something true and specific to say, is already done.
Which AI platforms should a UK business focus on?
Start by measuring rather than assuming, because it varies by category. Google AI Overviews usually matter most by volume in the UK, ChatGPT by mindshare, and Perplexity is the fastest to show movement, which makes it a useful early indicator. Copilot is worth attention specifically because it grounds in Bing, which most UK firms have never optimised for.
Can a small business compete with large brands in GEO?
On specific queries, yes, and more easily than in classic search. Models reward depth on a narrow topic over breadth across many, so a specialist with genuine expertise on one subject can be cited alongside far larger competitors. What a small business cannot do is win broad category head terms held by high-authority publishers. Specificity is the whole strategy.
What is the difference between GEO and AEO?
GEO is being cited inside a longer generated answer from ChatGPT, Perplexity, Gemini or Claude. AEO is winning the direct answer: Google AI Overviews, featured snippets, voice results. The underlying engineering is roughly eighty per cent shared, which is why MarGen delivers them as one engagement rather than three invoices.
How much does GEO cost in the UK?
MarGen publishes its figures: a free AI Visibility Audit, a £2,950 Synaptic Audit credited against your first retainer month, then retainers from £2,950 a month on Foundation, £5,950 on Authority (£6,500 regulated) and from £12,950 on Dominance. Only two of the seven UK agencies we track publish any pricing at all, so comparison is harder than it should be.
Where to Go Next
Deeper on the category: the complete GEO guide · GEO vs SEO · GEO vs AEO · LLMO vs GEO vs AEO · what is AI SEO
Diagnose your own position: why am I not in AI search results · the 35 GEO questions, answered individually · the free AI Visibility Score
Choose an agency well: 12 questions to ask · red flags · best UK GEO agencies compared
Have it delivered: MarGen is a UK GEO agency. The method is published in full, the pricing is on one page, and the results are in the case studies.