GEO case study: how to get a coffee brand recommended by ChatGPT

The challenge: a well-ranked site that AI ignores

On paper, everything looked fine. The site is technically sound (80/100), well structured (74/100), and Google understands it. Around a hundred product pages and a hundred blog posts: a solid foundation.

But buying habits are changing. To choose a coffee, more and more shoppers ask an AI for advice instead of comparing ten Google results. So the starting question was simple: when a customer asks an AI for a recommendation, is the brand part of the answer?

That is what GEO (Generative Engine Optimization) is about: being cited in an AI's answer, not just ranked on Google.

The diagnosis: 3 findings

Finding 1: the brand is missing from recommendations

REVVDA asked 9 GPT and Gemini models 3,415 buying questions: "best coffee", "where to buy", "comparison"… The result: 837 brands cited, and never the client's. (How to find which competitors AI recommends instead of you.)

Comparison queries concentrate the citations. That is where the gap is widest.

Finding 2: it is not an image problem

When the brand is named, not a single answer is negative. AI models know it well (93/100) and praise its smoothness and low acidity.

An AI can know a brand very well… and still never recommend it.

The real issue is discoverability: nothing on the web nudges AI models to cite it. This paradox is not an isolated case: we saw it again with a premium sleepwear brand that AI models know at 97/100.

Finding 3: 4 content blockers

1. 01

   No facts (1/15). Product pages describe feelings, not data: no precise origin, no figures.

2. 02

   No quotable sentences (7/20). Not one short, assertive sentence such as "[Brand] roasts 100% Arabica beans for a coffee that is never bitter."

3. 03

   No structured data (5/25). No Article schema on the blog, no breadcrumbs.

4. 04

   Thin pages. 83 pages have fewer than 300 words and lack industry vocabulary (Arabica, roast profile, tasting notes).

On top of that come classic SEO errors: 20 missing or duplicate H1s, 174 title tags to rewrite, 51 orphan pages.

The solution: a two-step action plan

The audit produced 29 actions, ranked by impact and effort. Rather than launching everything at once, the plan moves in two steps.

Step 1: fix the basics every engine reads

Step 2: make the brand quotable by AI

• ●A clear definition on the homepage: "Who is [Brand]?", in two factual sentences.
• ●Facts on every product page: origin, roast level, flavor profile, brewing tips. (How to turn thin product descriptions into fact-rich ones.)
• ●Comparison content, where citations are won.
• ●Marked-up FAQs (FAQPage) on key pages.

Every action becomes a ticket, with the before, the after and the rationale. On a product page, one vague sentence about how the beans look becomes a full paragraph: origins, roast, flavor notes and customer benefit.

Go further: How to turn an SEO/GEO audit into a prioritized roadmap.

Key takeaways

1. 01

   Measure SEO and GEO separately. A site can score 75/100 on one and 0% on the other.

2. 02

   Write facts, not slogans. Figures, origins, precise tips: that is what AI models pick up.

3. 03

   Target comparisons. That is where competitors capture up to 34.7% of citations.

FAQ