GEO agency: what it actually does, what it costs, and how to pick one

A new kind of supplier showed up over the past year: the GEO agency. Some have spent months working on the problem seriously; others changed the title of the proposal, added a paragraph about ChatGPT to the same old report, and raised the price. From the outside they look much alike, and the difference only becomes visible three months in, when someone has to show results.

This guide is for the buyer's side. What work GEO actually involves, how it should be measured, which questions separate the people doing it from the people naming it, what it costs in 2026, and — the part no agency puts on its website — when you're better off hiring nobody yet.

Key takeaways

• +GEO doesn't replace SEO, it sits on top of it. If your content can't be crawled, indexed or properly understood, fix that base before paying for AI-search-specific work.
• +Google says it plainly: there are no separate requirements or special schema for its AI answers, and optimizing for generative search is still SEO. What a good GEO agency adds is method — and above all, measurement.
• +The question that filters hardest: “how will you measure whether this worked?” Serious measurement runs each question in several phrasings and several times per engine. A screenshot measures nothing.
• +Since August 31, 2026, Search Console shows AI Overviews and AI Mode impressions worldwide. For Google there's now first-party data; for ChatGPT, Claude and Perplexity you still need outside observation.
• +Nobody can guarantee showing up in ChatGPT. Anyone who guarantees it is selling something they don't control: the models have no ranking dashboard and no position API.
• +At loco22, a monthly retainer starts at USD 500 with the audit included. As a rough guide, a standalone audit runs about USD 800 to 2,000. Far below that is usually an automated report.
• +If you sell something nobody researches before buying, or Google still isn't indexing your site properly, GEO probably isn't your priority yet.

What a GEO agency actually does

GEO (Generative Engine Optimization) is the work of increasing the likelihood that a brand, a page or a piece of first-party information gets discovered, understood, mentioned or used as a source in answers generated by systems like ChatGPT, Gemini, Perplexity, Claude and Google's AI Mode. It isn't only about being named: it's also about your page being cited, your data being used, and what you do being described correctly.

It's worth knowing up front what Google says: there are no additional requirements or special optimizations to appear in AI Overviews and AI Mode. The page has to be indexed and eligible to show with a snippet, and from Google's point of view, optimizing for generative search is still SEO. That doesn't make a GEO agency useless, but it does change what you should ask for: not a secret ingredient, but method and measurement you can verify.

At loco22 we organize that work into three fronts: citable content, entity clarity, and measurement. They aren't official requirements from any engine; they're our framework for organizing a problem that's still changing fast.

1. Content that's easy to cite

There's no mandatory format for getting an AI to cite a page, and Google makes clear you don't need to write in a special way for its AI features. Even so, research on generative engines finds that pages with clear structure, definitions and concrete figures contribute more to answers, and that matches our experience: clear answers up top, figures with their unit and date, explicit comparisons, sources, and claims that stand on their own are easier to retrieve and use as evidence. Writing that's excellent to read straight through can be hard to cite if its best idea is spread across five paragraphs.

2. A clear, consistent entity

The more consistent the public information about a company, the easier it is for search engines and AI systems to correctly connect its name, activity, location, people, products and sources. Structured data, official profiles, the same name, address and phone everywhere, identifiable authors and external references all help reduce ambiguity, although no single element guarantees a mention or a citation. It's the dullest part of the job and one of the most rewarding: a phone number from one country and an address in another plant a doubt that's hard to clear later.

What schema does (and doesn't do)

• +What it does: explicitly describes organizations, people, products, articles, services and the relationships between them. Google still recommends it as part of SEO, and it's worth getting right.
• +What it doesn't do: guarantee you'll appear in ChatGPT, Gemini or AI Overviews. Google says explicitly that structured data isn't required for generative search and that there's no special schema for it.

3. Repeated measurement across engines

Without this, the other two fronts are an act of faith. The problem is that AI answers vary: the same question can return different sources depending on the run, the model, the country or the day. A critical review of 45 GEO studies published in July 2026 found substantial run-to-run variability and recommends measuring with repetitions, paraphrases and controls. That's why asking once and saving a screenshot measures nothing. It's repetitive work, which is exactly why it's the first thing missing from cheap proposals.

How a GEO agency should measure (and what to demand)

This is what's worth getting in writing before you sign, whoever the supplier is:

• +A set of real questions from your field, built with you, each in several phrasings: the same intent said three or four different ways.
• +Several runs per question and per engine — ChatGPT, Gemini, Perplexity, Claude — logging the model, the date and the country. One result is an anecdote; ten are a rate.
• +A baseline taken before anything is touched, so there's something to compare against in three months.
• +Search Console for Google: since August 31, 2026, its generative AI reports show AI Overviews and AI Mode impressions worldwide, by page, country and date (no clicks yet). The other engines have no equivalent dashboard.
• +AI referral traffic in your analytics: ChatGPT, for example, tags the visits it sends with utm_source=chatgpt.com.

And what doesn't count as measurement: a “position 3 in ChatGPT”, an isolated screenshot, a single run per question, or the number of articles published that month.

Seven questions worth asking before you sign

• +“How will you measure whether this worked?” If the answer doesn't include a question set with several phrasings, repeated runs and a baseline, there's nothing to compare against in three months.
• +“How much of this is technical SEO?” The honest answer is “quite a lot, especially early on”. If they say none, they don't understand how engines reach content: Google needs to index you to show you in its AI answers, and ChatGPT and Claude use their own crawlers.
• +“Show me a case where you measured brand citations, not rankings.” It's the hardest test and the one fewest people can pass today, because the discipline is new. An “I don't have one yet, but here's how I'd measure it” is a better sign than an invented case study.
• +“What will you measure with the platforms' own data, and what with outside tools?” For Google, Search Console's generative AI reports now exist; for ChatGPT, Claude and Perplexity, they don't. An up-to-date agency keeps the two apart and doesn't present a sample as official data.
• +“What happens when the model changes?” It happens often and nobody controls it. Someone with a plan will say so without drama; someone without one will act like it doesn't happen.
• +“What do you need from my side?” A GEO project with no access to someone who knows the business produces content that's correct and generic. If they ask for nothing, they're planning to write without you.
• +“What do I keep if we stop working together?” The schema, the content, the question set and the measurement history should live on your site under your name. If the work lives inside an agency tool, you rented instead of building.

Nobody can guarantee ChatGPT will name you. There's no ranking dashboard and no position API: there's a probability that moves with every model update. Anyone guaranteeing a result there either doesn't know that, or is counting on you not knowing it.

Signs you're being sold SEO under a new name

None of these proves anything on its own, but three together is reason enough to get a second proposal.

• +The monthly deliverable is a Google rankings report with an AI section that measures nothing verifiable.
• +The plan boils down to “optimized content”: no mention of entity clarity, external sources or how it will be measured.
• +They talk about “ranking a keyword in ChatGPT” as if there were a stable 1-to-10 ranking. It doesn't work that way: one intent can trigger several searches — Google calls it query fan-out — and different sources and answers.
• +They show you a ChatGPT screenshot naming you as proof of results. It's one run among many possible ones: without repetitions you can't tell whether it was the rule or the exception.
• +They promise results in 30 days. Technical fixes do land fast; a stable presence across several engines does not.
• +They sell llms.txt as if it were the strategy. It's a text file you make in an afternoon: it can help as a complement, but Google says explicitly it isn't needed to appear in its AI features.
• +They never asked what you sell or who to. Without that you can't define the question set, which is the heart of the measurement.

What it costs in 2026

The market is young and prices are scattered: there are very cheap automated diagnostics and retainers of several thousand dollars a month. To give you a verifiable starting point, we'll begin with ours: at loco22, the monthly SEO + GEO retainer starts at USD 500, with the initial audit and the 90-day plan included in the first month, billed in dollars.

As a rough guide — our own estimate, not a survey or a market-standard rate — a standalone audit (diagnosis and prioritized plan, no execution) usually runs USD 800 to 2,000 depending on site size, and projects spanning several markets or languages rarely come in under USD 2,000 a month.

Far below that, what you're usually buying is a tool-generated report with little human work on top. It isn't a scam — the report exists — but the value in GEO lies precisely in the manual part: reading what the engines actually answer, repeating the questions to separate the rule from chance, working out why they cite someone else, and fixing the content accordingly.

When GEO probably isn't your priority yet

There are three situations where it's better to invest in something else first, and it's worth hearing that from someone who sells GEO. First: Google still isn't indexing your site properly. Fixing that is the priority, because it limits your visibility in Google Search, AI Overviews and AI Mode, and it often reveals technical problems that affect other systems too — even though ChatGPT and Claude have their own crawlers.

Second: you sell something nobody researches before buying, like a takeaway or a corner shop. There, the map and local reputation pay off more, and they also feed Google's AI answers: its documentation asks you to keep your Business Profile up to date. Third: you won't produce first-party information, update what you already have, or build presence beyond your own site. Without any of that the room for improvement shrinks a lot: the work stays technical and runs out within a few months.

What a good first quarter looks like

As a yardstick: in the first two weeks there should be an audit with a prioritized plan and a baseline taken with the protocol above — you need to know where you started. Weeks three to six go on the technical base: crawling and indexing, speed, structured data, a consistent entity and access for AI crawlers. From month two, content built to be cited plus the external mentions work. And every month, without exception, the question set measured again with the same repetitions, plus the AI impressions in Search Console.

If by the end of the third month nothing has moved on any of the metrics — mention and citation rates, impressions in Google's AI features, external mentions — it isn't necessarily badly executed, but it does need a concrete explanation of why, and what changes from here.

At loco22 we do exactly this work, and we're happy for you to put the seven questions above to us — we'd rather that than a sales call. If you want to see where you stand before deciding anything, the initial audit includes your baseline in ChatGPT, Perplexity and Gemini plus a prioritized 90-day plan. Tell us what you sell and who to, which is where this starts.