GEO for agencies: method, real cases and an action plan for a profitable offer

GEO is taking more and more room in client conversations. Many agencies already sell "AI search optimization" — because clients ask about it, because the topic is trending, or because it naturally extends an existing SEO offer. The problem: in practice, many teams apply classic SEO reasoning to a lever that doesn't work the same way. The real challenge isn't just appearing in ChatGPT or Gemini answers: it's knowing when GEO makes sense, which queries to target, which technical building blocks to lay first, how to measure — and how to turn it into a repeatable process.

GEO in practice

GEO means working on a brand's presence in LLM-generated results. It's not about "optimizing a few pages for AI", but about understanding which sources answers actually draw on, on which types of queries the models genuinely rely on web content, and steering that presence in a way that serves the business. In practice, GEO gets really interesting on high-intent queries: comparisons, "best X" lists, requests close to a purchase or a decision. On purely informational queries, the effort is harder to justify and the business impact less direct.

Why many agencies still get it wrong

The first trap: taking the usual SEO process and bolting it onto GEO. The result: fuzzy client objectives, no real KPIs, under-equipped teams, and monitoring tools that measure visibility without producing an actionable plan. Before selling GEO, three things must be clarified: the client's maturity (if the technical foundations are weak, GEO comes too early), the type of queries to target (without business framing, GEO becomes abstract visibility) and internal execution capacity (purely SEO teams, with no GEO method, run out of steam fast).

First: a technical GEO audit

No serious GEO without solid foundations. A GEO audit looks at six things: structured data (schema alone isn't enough, but it's part of the foundations); editorial structure (readable, segmented, easily "chunkable" content with clear, synthetic blocks); HTML hierarchy (headings, metas, page organization); internal linking (which circulates signals and clarifies relationships between contents); credibility signals (E-E-A-T: page clarity, editorial consistency, perceived reliability of the brand); and content freshness, reported as a significant signal in several field observations.

The right focus: transactional queries

"Best B2B CRM", "best mechanical breakdown insurance", "software X comparison", "best burger restaurant in Angers"… That's the best starting point for structuring an offer: easier to sell to the client, clear business intent, comparison content that's easier to produce — and LLMs often lean on comparative or synthesis sources for these queries. The brand positions itself as a solution, not just as a source of information.

Measuring: attribution is fuzzy, be pragmatic

A user discovers a brand through ChatGPT, comes back via Google, clicks an organic or sponsored result, converts later: the original source is lost to classic attribution. The most concrete recommendation: add a declarative "How did you hear about us?" question in the funnel, before the quote or form is sent. It's not perfect, but it captures information analytics will never see — to be cross-checked with GA4 and other signals.

Three concrete strategies to activate

1. Internal comparison tools: embed comparative content on the site that positions the brand within its competitive landscape. Real example: a comparison tool on a mechanical breakdown insurance site, with a market comparison and pages pitting the brand against other players — designed to be picked up in generated answers on comparative queries, not just for SEO.

2. Placements in third-party comparisons: identify the comparison content that already carries weight in LLM answers, then negotiate an insertion or a better position. It's a more advanced version of link building — observed orders of magnitude range from €50 to €300 per month depending on the site. If the page lives off affiliation, you have to offer more value than it generates.

3. Satellite sites: launch dedicated external comparison sites or content, creating additional sources likely to be cited in answers. More aggressive, to be handled with judgment, but effective in well-chosen niches.

Not all LLMs behave the same

ChatGPT digs for deep sources in the SERP, not necessarily the top 3; Gemini and AI Overviews stay closer to the most visible results. Three possible postures: aim for the top of the SERP, multiply relevant sources even deep ones, or combine both depending on the target platform.

Two real cases, two approaches

Roquebrune-Cap-Martin: prompts defined by theme, a visibility audit, citations classified by type of presence (clickable link, citation without link, indirect mention, absence), then a detailed action plan with post-audit KPIs — a structured approach that turns GEO analysis into actionable deliverables. Weekendou: a brand not yet strong in SEO on "creative gifts" — rather than waiting for a full SEO catch-up, direct GEO work through a high volume of articles positioning the brand as an expert. Proof that a targeted GEO investment can be justified even when SEO isn't perfect.

Start with a pilot client

The right approach is not to sell a standardized "GEO offer" to every client. Start with a pilot client: test the method, produce concrete deliverables, structure the process, level up the team — then transpose. And technical execution makes all the margin: being able to quickly build comparison tools, dedicated pages, audit tools and automations (with Claude Code, for instance) is often more profitable than staying stuck in heavy stacks. GEO is not a rebranding of SEO: it's a method — client qualification, technical foundations, transactional focus, source influence, pragmatic measurement, execution.