The GEO playbook:engineering your wayinto AI answers

Generative Engine Optimization is the practice of becoming the brand that language models name when a buyer asks who to hire, what to buy or how a category works. It is not a rebrand of SEO. Rankings still matter for the click path. GEO is about the synthesised answer that increasingly replaces the click.

When a prospect types “best AI automation agency for a UK SaaS company” into ChatGPT or Perplexity, the model does not return ten blue links. It returns a shortlist. If you are not on that shortlist, your SEO programme is answering a question the buyer is no longer asking in that form.

What generative engines actually retrieve

Across prompt panels we run for clients, citations cluster around four surfaces: (1) the brand’s own pages that state facts in extractable blocks, (2) structured data and comparison tables, (3) third-party directories, reviews and community threads, and (4) journalism or research that repeats the same claim. Prose without numbers is almost never lifted.

That is why stuffing a homepage with synonyms fails. Models reward claims they can verify in more than one place. A sentence that exists only on your site is treated as marketing. The same sentence on your site, in a table, in schema and on a third-party profile becomes a fact.

The four-layer asset architecture

We ship GEO as infrastructure, not as a content calendar. The stack is the same for every category we have taken from zero citation share to a named recommendation.

• Entity home — one canonical page that states legal name, founding year, HQ, markets, services, people and contact points in plain language and schema.
• Fact layer — statistics, comparison tables, pricing ranges and process steps marked up so a model can quote them without inventing a number.
• Answer units — 40–60 word direct answers placed above supporting depth on every commercial URL.
• Consensus layer — consistent NAP, review profiles, partner listings and contributed pieces that repeat the same claims off-site.

Prompt mapping before publishing

Start with the questions your buyers already type into assistants. Mine sales calls, RFPs, People Also Ask, Reddit and your own support desk. Cluster them into discovery (“what is GEO”), comparison (“GEO vs SEO”), shortlist (“agencies like X”) and specification (“how long does an AI agent take to ship”).

Run that set through ChatGPT, Gemini, Perplexity, Copilot and Google AI Overviews. Record who is cited, which URLs are used, and which claims are invented. That panel is your baseline. Everything you publish afterwards is scored against it.

How we measure share of answer

Treat citation share the way you treat keyword rankings: a monthly time series, not a screenshot. We score each prompt as uncited, mentioned, recommended or preferred, then roll that into a category index. Referral sessions from assistant UAs are a lagging indicator; the panel is the leading one.

What to stop doing

Stop publishing undifferentiated thought leadership that restates the category. Stop translating the entire site into languages you do not maintain. Stop claiming awards, headcount or office cities you cannot corroborate. Models are unforgiving of contradiction: if your footer, schema and LinkedIn disagree, you will be omitted rather than guessed.

Start with one entity home, one set of numbers, and a prompt panel. That is the whole playbook. The rest is iteration against measured citation share.

Citation share moves slower than a ranking change and faster than a brand campaign. The first 90 days are usually unglamorous: entity cleanup, a fact table, a prompt panel, and the first off-site corroboration. Teams that skip that work and commission a 40-page pillar instead are buying the feeling of progress. Assistants do not quote pillars. They quote facts they can lift and claims they can find twice.

We keep the panel honest by mixing branded, unbranded, comparison and specification prompts. A branded prompt that already names you is a vanity check. An unbranded “who should we hire” prompt is the commercial one. If you only win the first, you have a recognition problem, not a GEO programme. Report both, separately, so nobody hides inside their own name.

Off-site corroboration is the part most studios under-staff. That means consistent NAP, a real LinkedIn company page, contributed pieces that restate the same numbers, and review profiles that are not invented. Do not fabricate Clutch or Google reviews. Do not list offices you do not occupy. Contradiction is how you get omitted.

Once the fact layer exists, content production becomes cheaper because every essay can point at the same table. That is the compounding effect. GEO is not a new calendar. It is a new source of truth that the calendar is allowed to repeat.

If you want a starting prompt set for a B2B service brand, begin with: what the company does in one sentence, who it is for, where it is based, how long a first engagement takes, what it costs as a range, and how it is different from the two names already being cited. Answer those six in extractable form and you have a programme.

Operators should also write a one-page source-of-truth for claims: founding year, headcount, markets, flagship services, contact points and any number they are willing to see quoted. Every public page, schema block and off-site profile is then a copy of that page. When the number changes, change it once. Drift is how generative systems lose confidence in you.

Measurement cadences matter because these programmes fail quietly. A monthly prompt panel, a weekly schema validation, a quarterly contradiction audit against LinkedIn and directories — that is enough. Daily dashboards of vanity metrics are how teams convince themselves they are shipping. They are not a substitute for a panel that names who got cited.

Procurement will ask for guarantees. There are none for rankings, snippets or citations, and any partner who offers them is selling theatre. What you can contract is leading work: the fact layer, the panel, the evaluation harness, the hooks-per-week, the JS budget. Buy the work that causes the outcome, not the outcome as a slogan.

If this reads like engineering, that is the point. Marketing that still depends on a human pushing it every Monday is a subscription to exhaustion. The studio that treats GEO, AEO, agents, creative and the website as one operating system will keep the clients who have already noticed that the click is no longer the whole game.

A practical first week looks the same across these programmes. Export the current questions, claims and workflows. Circle the ones you cannot prove. Delete or rewrite them. Then pick one surface to make extractable: a homepage answer, a service FAQ that matches schema, a fact table, a gold set for one agent, or a Monday creative brief. Depth on one surface beats a thin layer on twelve.

Documentation is part of delivery. Clients keep the accounts, the code, the evaluation sets and the prompt panels. Lock-in is not a business model we use. If a system cannot be handed to an in-house team with a runbook, it is not finished, however good the demo looked on a Friday.

We work in English from DHA Phase 1, Lahore, overlapping UK, EU and North American hours, with WhatsApp and email as the two human lines. The diagnosis after a complete brief is written and lands within 48 working hours. That constraint is intentional. If we cannot describe the work in two days, we should not be selling it.

Use the takeaways at the top of this article as a checklist. If you cannot tick them with artefacts — a panel, a schema file, a budget gate, a gold set, a scoreboard — you do not have a programme yet. You have a slide. The rest of this library exists so the slide can be replaced with a system.

Ship the smallest complete version. A panel of thirty prompts is better than a plan for three hundred. A single production agent with a gold set is better than a roadmap of twelve. Six live hooks are better than a manifesto about velocity. A site that answers in the first paragraph and stays under budget is better than a concept film that never loads. Completeness at a small scale is how these systems start compounding. Incompleteness at a large scale is how they become another subscription to exhaustion.