Part of our Generative Engine Optimization complete guide.
In late 2023, maybe 3 or 4 agencies called themselves "GEO" globally. By 2026, there are hundreds. Most are SEO agencies that rebranded "AI" on their landing page without changing their method. A minority have built real expertise.
Without clear criteria, impossible to tell one from the other before you've signed. Here's the grid I use to qualify a GEO agency, and the cases where going through an agency isn't the right call.
The 5 criteria that separate a real GEO agency from marketing packaging
Before looking at price, five questions eliminate 80% of the market.
Can the agency name citation factors without fuzzy jargon? Topical authority, fan-out queries, chunking, brand signals, brand-topic-proof associations. If those terms don't come out naturally in a 30-minute discovery call, or come out unexplained, you're facing an agency that rebranded its offering without reworking its method.
Does it cover at least 5 LLMs in tracking? ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews at a minimum. An agency that only tracks ChatGPT misses 60 to 70% of real AI traffic in 2026. That's a red flag for insufficient maturity.
Does it use a professional monitoring tool? Mentionable, Otterly, Peec AI, Profound, Semrush AI Toolkit. If the agency measures via manual screenshots or homemade Excel sheets, its ability to deliver reproducible results is weak.
Does it show case studies with documented before/after metrics? Citation rate per LLM, monthly evolution, Share of Voice. Vague cases ("we grew the client's visibility") with no hard numbers are worthless. Metrics must be named, dated, attributable.
Does it explain its method or hide behind a black box? A serious agency explains how it works and pushes you to understand. An agency refusing to go into detail ("we have our secret methods") creates intentional dependency. Avoid.
The 5 red flags that should make you walk away
Promise of results in 30 days. Physically impossible. LLMs recrawl over weeks, integrate changes with a delay, and most citation factors build over several months. Any promise of results in under 90 days is commercial bluff.
Pricing under 1 000 euros per month with broad promises. A serious GEO engagement costs between 1 500 and 5 000 euros per month. Below that, either it's classic SEO rebranded, or the agency is underselling to close fast and will rush execution.
Total absence of case studies, or generic ones. "Our clients doubled their AI visibility" with no client name, no metric, no date. That's a signal the agency has nothing to show or doesn't dare to show it.
Exclusive focus on ChatGPT. In 2026, Perplexity, Gemini, and Google AI Overviews combined weigh more than ChatGPT in many niches. A single-platform agency misses the market.
Refusal to document the method. If you can't ask "what did you work on this month and why?", you're paying blind. A good agency keeps a detailed monthly report with actions taken, results, and decisions for the next month.
When hiring a GEO agency doesn't make sense
Most content on "how to choose a GEO agency" forgets that an agency isn't the right pick for everyone. Three profiles are better off not hiring.
The solopreneur. With a limited marketing budget, the cost of a serious agency (1 500 to 5 000 euros per month) is often prohibitive. A structured training (free or up to 1 500 euros one-shot) combined with a tracking tool (29 to 89 euros per month) produces comparable results on a scope of 10 to 20 prompts.
The small business with in-house marketing. If you have even one marketing person able to spend 5 hours per week on the topic, training + tooling almost always beats an agency on cost-efficiency. You keep the expertise in-house.
The market test. If you're not certain your audience massively uses LLMs to search for recommendations in your category, hiring an agency means investing 30 000 euros per year before validating the channel. Better to validate with a light plan (DIY or training) before scaling with an agency.
When a GEO agency becomes the right choice
Three profiles actually benefit from an agency.
The mid-size company that has already validated its customers use LLMs and has no internal resource to own the topic full-time. The large account managing multiple brands or markets and needing to industrialize. The company that must catch up on a severe lag against already well-positioned competitors, with a budget to compress the 12 to 18 months DIY demands.
In these three cases, a serious agency pays for itself. You still need to pick the right one with the grid above.
The compromise that often works best
Between the full-agency model and pure DIY, a common compromise delivers solid results: follow a structured GEO training to establish the analytical lens, then delegate heavy execution to an agency while staying able to challenge its choices, and keep tracking and steering in-house via a monitoring tool.
This model combines the best of both: you don't pay 5 000 euros to learn what a 500-euro training teaches, and you don't spend 20 hours per week executing what an agency would do in 5 hours at its optimal cadence.
