Home / Blog / Guide

Guide

What a £1,800/mo GEO agency actually sells you (and what to do instead)

By Zeb Choudhry · Visus · 24 August 2026 · 8 min read

A post did the rounds recently: a small business owner was quoted $1,800 a month plus a $1,400 setup fee for a "generative engine optimisation" package - four articles a month, some schema, a dashboard. Then they discovered the articles were being drafted with ChatGPT. The reaction was, understandably, fury.

It is worth pulling that reaction apart, because "an agency used AI to write the articles" is not automatically a scandal - plenty of good writers use AI as a first draft. The real problem is the gap between what was charged and what was delivered, dressed up in language designed to make a simple thing sound like alchemy. So here is an honest breakdown of what a GEO retainer actually contains, which bits are worth money, and what you can do yourself for free before you sign anything.

What is actually in a typical GEO retainer

Strip away the branding and most AI-SEO or GEO packages are built from the same four components. None of them are magic. Here is what each one really is, and what it is worth.

Add it up and the picture is clear: the audit and schema are mostly one-off, the content is the recurring cost, and the monitoring is worth what its honesty is worth. That does not make every agency a cowboy. It just means you should know which line items you are actually paying a premium for.

The reputable-versus-cowboy tell

The single fastest way to sort an honest GEO firm from a hype merchant is to listen to how they talk about outcomes and timelines. This is not subtle once you know what to listen for.

A reputable firm will give you honest timelines and no guarantees. They will say something like: long-tail, low-competition questions can start moving in around 60 to 90 days; competitive terms where everyone is fighting for the same citation realistically take 3 to 6 months, sometimes longer. They will talk in terms of signals, readiness and probability, because that is the truth of how AI answers work - the models are non-deterministic and change on content that never moved.

A cowboy guarantees the thing nobody can guarantee. Any promise of "guaranteed AI recommendations", "we'll get you cited in ChatGPT", or a specific number ("ranked #1 in AI answers within 30 days") is a red flag. Nobody controls what an AI model recommends. You can improve the readiness signals that make citation more likely - crawlability, entity clarity, corroboration - and then you wait and measure. Anyone selling certainty is selling something they cannot deliver. That is the tell.

What you can do in-house first, for free

Before you pay anyone a retainer, spend an afternoon on the fundamentals. Most of what an entry-level GEO package delivers in month one is here, and none of it needs an agency:

• Fix crawlability - make sure your robots.txt and settings are not blocking AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended). This is often a single line and it is the most common self-inflicted wound.
• Sharpen entity clarity - state plainly who you are, what you do, where you operate, and who you serve. Models need to understand your business before they can recommend it.
• Make your content answer-ready - add a real FAQ that answers the questions buyers actually type, in plain language, with the answer right after the question.
• Put your pricing where it can be read - visible pricing, or at least a clear pricing structure in text, is a strong signal. Hidden pricing gives an AI nothing to cite.
• Get your reviews on the page as text - testimonials rendered as readable text (not trapped in a widget or an image) are corroboration a model can actually use.
• Add basic schema - Organisation, FAQ and Product/Service markup. Copy-paste templates exist for all of them.

Do these six things and you have closed most of the gap that a first-month audit would have flagged - at no cost. If you want to see exactly which of them you are missing, the AI visibility checker gives you a directional readiness score and a prioritised list in plain English. And if you want to measure whether any of it is working over time, our guide on how to measure AI visibility without fooling yourself walks through the frozen-question-set method the honest agencies use on their own sites.

When an agency is genuinely worth it

None of this means "never hire anyone". There are real situations where a good generative engine optimisation agency earns its fee:

• Ongoing execution at scale - if you need a steady stream of genuinely researched, expert content every month and cannot produce it in-house, that is a real service with real cost.
• No in-house time - the fundamentals above are simple, but "simple" is not "free of your time". If your hours are worth more elsewhere, paying someone competent to execute is a rational trade.
• Complex or technical sites - large catalogues, multiple locations, or messy migrations where the schema and crawl work genuinely is recurring.

The difference is that you should hire from a position of knowledge. If you have already run a baseline and fixed the obvious gaps, you can hand an agency a precise brief and judge their work against it - instead of paying a retainer to rediscover things you could have found yourself. That is also the honest case for structured GEO services: they are worth most when they execute a plan you understand, not when they sell you a mystery.

The person who got quoted $1,800 a month was not angry that AI was involved. They were angry that they could not tell what they were paying for. The fix is not cynicism - it is a baseline. Know your starting point, know which line items are one-off versus recurring, and know that anyone guaranteeing AI recommendations is guessing. Then the decision to hire, or not, is yours to make with your eyes open.