GEO Agency: Generative Engine Optimisation

Being ranked is not the same as being named

Generative engine optimisation is the work of getting your brand named and cited in the answers ChatGPT, Google’s AI Overviews, Gemini, Perplexity and Copilot write. Ranking third for a term still earns a click. Being absent from the answer that replaces the results earns nothing. GEO measures whether you are in the answer.

The two disciplines overlap heavily, which is why we do not sell them apart. Assistants draw their answers largely from pages that already rank, so the authority earned through SEO is the raw material GEO works with. What GEO adds is a different measurement: not position for a keyword, but presence, framing and citation across a fixed set of prompts.

How we work: the 21-prompt method

You cannot improve AI visibility you have not measured, and there is no rank tracker for a conversation. So we build one.

1. Build the prompt set

We write 21 prompts that represent how your buyers ask: comparison questions, “best X for Y” questions, brand questions, and the objections people raise before buying. They are agreed with you and then fixed, because the point is a comparable measurement over time rather than a flattering one.

2. Run the baseline grid

Every prompt runs across the major assistants and is logged: are you named, where in the answer, in what terms, and which sources were cited instead. The grid usually exposes two problems at once. Prompts where you are invisible, and prompts where you are named but described wrongly. They need different fixes.

3. Entity and schema work

Models name brands they can identify with confidence. That means one consistent description of what you are and what you sell, across your site, your profiles and the sources describing you; server-rendered HTML an AI crawler can read without executing JavaScript; and clean Organization and Breadcrumb schema. There is no special file or AI-only markup to add.

4. Source-side content

We then work on the pages the models cite. Sometimes that is your site: the comparison page, the pricing explanation, the table that answers the question in its first sentence. Often it is not your site at all, and the work is earning a mention in the roundups and publications the assistant already trusts.

5. Re-test at day 30, 60 and 90

The same 21 prompts, the same assistants, the same grid. You see movement prompt by prompt, and where a competitor is still named instead of you, we can show which source put them there.

Proof

• Official London Theatre: 85% AI search visibility, and the recommended source for its category. A five-year client, which is roughly how long the underlying authority took to build.
• Koyo Loans: a top three ranking for “loans” against banks and comparison sites. A search-side result rather than an AI-answer one, but it is the kind of position that makes a brand quotable: assistants answering finance questions draw from the pages already holding those results.
• Sons: a men’s wellness brand whose brand-mention footprint was built at scale through creator work, giving models a consistent description to draw on.

More of the work is on the case studies page. You can check your own starting position with our free GEO visibility score.

What GEO costs

GEO is included in every SEO plan at no extra cost. There is no GEO tier, no AI visibility add-on and no separate invoice line. That is deliberate rather than a discount: the entity work, schema hygiene, source-side content and earned authority that get you cited are the same work that gets you ranked, and charging twice for it would be dishonest.

Standalone GEO (an audit, a prompt grid and a remediation plan for a brand that already has an SEO team, or a benchmark before a launch) is scoped to the brief, because the size depends on how many prompts, markets and assistants you need covered. Tell us what you need measured and we will price it.

Who does the work

Patrick Macnamara founded Terrier in 2015 and leads the GEO practice. The prompt set, the baseline grid and every re-test are reviewed by Patrick, so the reading you get comes from the person accountable for the account, not a dashboard export. His background is on his profile; the wider team is on the team page. We work with brands in the UK, Australia and Ireland.