If your agency does SEO, content or PR, you've probably heard the question already: 'What does ChatGPT actually say about us?' More and more of your clients' customers ask an AI for advice instead of googling. They don't get a list of ten links; they get one answer with two or three names. GEO marketing is about how your client becomes one of those names.
What GEO marketing is
GEO stands for generative engine optimization: everything you do to make generative AI models (ChatGPT, Perplexity, Gemini, Claude) mention a business, describe it correctly and recommend it when someone asks a relevant question. Where SEO is about positions in a list, GEO is about presence in an answer.
In practice GEO consists of four parts, all of which an agency delivers or explicitly outsources:
- Tracking: establishing in what share of the relevant AI answers the client is mentioned, per AI model, and next to which competitors.
- Understanding: finding out why. Which sources does the AI cite, which pages does it read, what does it literally say, and in what tone?
- Improving: executing actions that raise the odds of a mention: listings on the sources AI trusts, content for questions that have no answer yet, technical access for AI crawlers, structured data.
- Proving: tracking again after every action and showing the effect in numbers the client understands.
That last part is what separates GEO from 'AI advice'. Without repeated tracking it's an opinion; with tracking it's a service a client is happy to pay for every month.
How AI models choose who to name
An AI model doesn't invent businesses. When someone asks 'which heat pump installer in Boston is good?', the model retrieves (or recalls from training) pages that answer that question: comparison sites, industry directories, trade media, forums, review platforms and the websites of businesses themselves. The answer is a summary of what those sources say. If your client is on the sources the model trusts, it gets mentioned. If not, it doesn't, however good its website is.
Three things stand out as soon as you start tracking this per client:
- Every AI model chooses differently. Perplexity leans heavily on current web sources, ChatGPT more on what it already 'knows' plus web search results, Gemini on the Google ecosystem. A client can be in 40% of answers on one model and 5% on another.
- A handful of domains deliver most of the citations. In most markets, six to ten sources come back in almost every answer. Missing from those sources is the most common cause of invisibility.
- Answers vary from one run to the next. The same question gives a slightly different answer tomorrow. That's why you track repeatedly and talk about a win rate, not 'in' or 'out'.
How GeoReady tracks this exactly is in the method; the underlying mechanics in how ChatGPT chooses which businesses to recommend.
The difference from SEO
GEO is not a replacement for SEO and not a marketing trick to sell SEO twice. It's a second KPI with partly different levers. The overlap is large: a fast, well-structured site with clear business information helps both. The differences come down to three things.
The outcome. SEO delivers a position in a list of ten; GEO delivers presence in a selection of two to five. The second is binary and harder: you're in or you're not.
The levers. In SEO, backlinks and on-page optimization dominate. In GEO it's about listings on the specific sources AI cites (which often don't have high authority in SEO terms), about completeness and concreteness of information, about consistency of business details across the web, and about technical access for AI crawlers such as GPTBot, PerplexityBot and ClaudeBot.
The measurement. SEO rankings are stable and publicly measurable. AI answers vary and differ per model; you need a fixed prompt set, tracked on fixed days across several models, to get a reliable picture. The full comparison is in GEO versus SEO.
What an agency needs to be able to do
The good news: you can already do most of it. GEO work looks a lot like what an SEO or PR agency does every day, just with a different goal and a different yardstick.
- Approaching sources. Getting a listing on a comparison site, an industry directory or a trade publication is PR work. The difference is that you now know exactly which sources matter, because you see which ones the AI cites.
- Creating content for questions. AI models ask themselves sub-questions (fan-outs) to build an answer. If your client has no page answering such a sub-question, that's a concrete content gap. That's content work with a clear brief.
- Technical work. Opening robots.txt for AI crawlers, adding an llms.txt, adding structured data (Organization with sameAs, FAQPage, LocalBusiness), matching page titles to the questions. That's technical SEO with a few new parts.
- Tracking and reporting. This is the part most agencies don't build themselves. You need a fixed prompt set, several AI models via their APIs, repeated runs and a breakdown of every answer. For that you use a platform; GeoReady is built for it, with a free agency portal where you see all your clients side by side.
Why now
The head start you can take now won't be there in a year. Right now, most businesses and agencies don't track their AI visibility at all. Whoever first occupies the right sources and answers the right questions in a market becomes the business the AI 'knows'. AI models are conservative: once a business is mentioned, it tends to stay mentioned, because new sources repeat the existing answers.
For an agency there's a commercial reason on top. Your existing clients already have the question; you don't need to open a new market. One extra service, sold to clients who already trust you, with a fixed monthly fee and a report that shows every month what changed. How to build and price that service is in pricing GEO as a service and in the 30-day plan.
Questions agencies ask
Isn't GEO just SEO with a new name? No. The overlap is real, but the yardstick (presence in an answer instead of a ranking), the main lever (listings on cited sources) and the measurement (repeated, across several AI models) are different. A client can rank #2 on Google and appear in zero AI answers.
Can I track it myself by checking ChatGPT? For a first impression, yes; as a service, no. One check is a snapshot of one model at one moment. You need a fixed set of questions asked on fixed days to several models, with the answer stored word for word. Read how GeoReady tracks.
How fast do you see results? Source actions (a listing on a frequently cited comparison site) often work within weeks. Content actions within one to three months. Technical actions are a precondition, not an accelerator. GeoReady measures the effect of every completed action thirty days later, in points.
What does it cost my agency? The agency portal is free; per client you pay a plan from €95 per month (Pro). What you charge your client is up to you. See GEO for agencies.
GEO (generative engine optimization) makes AI models mention and recommend a business. For agencies it's a natural extension of SEO and content, with a measurable outcome per client.