GEO: a practical guide for agencies

Your clients’ customers no longer search only on Google. They ask ChatGPT “which accountant should I choose in Lille?”, ask Perplexity “which gym is open on Sundays in Namur?”, and read the answer generated at the top of Google’s results. GEO (Generative Engine Optimization) is about making sure a business is present, accurately described and recommended in those answers.

For an agency, it is both a new service to sell and a powerful prospecting argument. This guide takes stock.

GEO and SEO: what really changes

Classic SEO aims for a position in a list of links. GEO aims for a mention in a written answer. Three differences follow:

1. There is no “first page”. An answer usually names between 2 and 5 businesses. Either you are in it, or you do not exist for that user.
2. The answer varies. The same question asked twice can produce two different lists. Visibility becomes a probability: “cited in 40% of answers”, not “3rd position”.
3. Sources matter as much as the website. AI assistants rely on pages they consider trustworthy: directories, comparison sites, local media, reviews, forums. A business missing from these sources has little chance of being recommended, even with an excellent website.

SEO remains the foundation: generative engines rely heavily on well-indexed pages. But it is no longer enough.

What AI assistants seem to value

No engine publishes its recipe, and you should be wary of anyone who claims to know it. By observing what answers cite, however, some constants emerge:

• Explicit, verifiable facts: area covered, opening hours, indicative prices, lead times, specialities, labels.
• Wording that answers the question asked: a “Boiler repair in Roubaix” page answers “who repairs boilers in Roubaix?” better than an “Our services” page.
• Consistency between sources: the same information on the website, the directories and the business listings.
• Freshness: outdated or conflicting information loses the citation.
• Third-party reputation: being mentioned by others (media, blogs, comparison sites) carries a lot of weight.

Measure before acting

A serious GEO service starts with an honest measurement:

1. Define the questions customers actually ask, by offer, area, audience and intent (discovery, comparison, decision, brand).
2. Ask them several times to each engine, because an isolated answer proves nothing. We explain why in our article on confidence intervals.
3. Extract the businesses mentioned, the sources cited and the tone.
4. Compare: who is recommended instead of your client, and on the basis of which pages?

That is exactly what Recoia automates.

Building a profitable GEO offer

For an agency, GEO sells well because it is demonstrable: you can show the prospect what the AI answers today. An offer structure that works:

• Initial audit (one-off): visibility per engine, competitors recommended, gaps, action plan.
• Remediation (project): answer pages, structured data, directory listings, factual FAQs.
• Monthly tracking (recurring): continuous measurement, alerts on significant drops, monthly report.

Monthly tracking is the key to recurring revenue: AI visibility moves with every model update and every new competitor.

What not to promise

Nobody can guarantee that a business will be cited by ChatGPT. Promises of that kind damage trust in the whole profession. Promise a method instead: measure, explain, fix, measure again — with figures that come with their margin of error.

In short

• GEO aims for a mention in a generated answer, not a position.
• AI visibility is a probability: it is measured over several runs.
• Third-party sources matter as much as the website.
• It is a demonstrable service, and therefore easy to sell — as long as you stay honest about what you can promise.