$25M
raised in investment
How Cocoa Supply went from 14% to 95% AI visibility in five months
Cocoa Supply is a wholesale cacao supplier that had been in business 13 years with strong organic search traffic and genuine category expertise, and was still absent from almost every AI answer about bulk cacao. In five months of GEO work from May to September 2026 it went from appearing in 14 of 100 tracked buyer prompts to 95, from an average mention position of 13.6 to 2.1, and from zero first place rankings to 59. It is now the second most cited cacao brand in AI search, behind only Barry Callebaut.
Cocoa Supply is a wholesale cacao supplier with 13 years in the business. By every measure marketing teams have used for a decade, the company was in good shape. Organic search traffic was strong, the site ranked for the terms buyers typed, and the people running it knew cacao at a level most of their competitors could not match. That expertise was real and it was already written down.
The AI picture was the opposite. Across 100 buyer prompts we tracked monthly, Cocoa Supply appeared in 14 of them. When it was named at all, it landed at an average position of 13.6, which in practice means buried near the bottom of a long list. It held first place on none of them. A company with 13 years of category authority was, to an AI assistant, close to a stranger.
The buying behavior moved and the measurement did not follow it. A procurement lead sourcing bulk cacao used to run a search, open five tabs and compare. Now a growing share of them open ChatGPT or Gemini and ask who supplies bulk cacao, then treat the shortlist that comes back as the shortlist. There is no page two to scroll to and no second result to click. There is one paragraph, and either your name is in it or it is not.
Cocoa Supply was not in it. Competitors were. That gap never showed up in the analytics, because an assistant naming a rival to a buyer who never visits your site generates no traffic to miss and no ranking to lose. The company was losing the shortlist stage of the deal without a single number moving to tell it so.
An AI answer is assembled from sources the model already trusts, so the work is making Cocoa Supply one of those sources rather than optimizing a page and waiting. Four things ran at once.
We placed editorial coverage on high domain authority publications, because a claim about cacao sourcing carries very different weight to a model when a trusted third party publishes it than when a supplier says it on its own site. Alongside that we built out the wider authority footprint the engines read when they decide whether a brand is a real player in a category.
We then turned the client’s own cocoa knowledge into long form content written to be quoted. Not marketing copy about being a trusted partner, but specific, checkable answers to the questions buyers actually ask about origin, fermentation, volume and pricing, structured so a model can lift a passage and attribute it cleanly. We implemented schema markup so the facts on the site were machine readable rather than inferred, and we built a genuine presence on Reddit and Quora, which are sources these models lean on heavily when a question calls for practitioner experience rather than a vendor page.
Underneath all of it we tracked 100 buyer prompts every month across ChatGPT, Claude, Gemini and Google AI Overviews. Four engines, because they do not agree with each other and a brand can be winning in one while invisible in another.
The most useful number in this engagement is a zero. For four consecutive months, Claude cited Cocoa Supply exactly zero times. Not rarely. Never. Every other channel was showing movement and that one line stayed flat, month after month, while the work continued.
In month five it recorded 54 first place Claude citations.
This is the shape GEO actually has, and it is the reason the discipline gets abandoned early. Authority compounds quietly before it converts. The placements, the content and the third party signals accumulate in a model’s picture of a category for months without producing anything visible, and then cross a threshold and convert all at once. A client who stopped at month four would have concluded, with a reasonable looking spreadsheet in front of them, that the work did not do anything. They would have been wrong by four weeks.
Because the field got harder, not easier. Over the engagement the number of rival brands appearing anywhere in the tracked cacao prompts grew from 1,423 to 3,355. More than twice as many names competing for the same finite space in the same answers.
In that crowd the established leaders went backwards. Barry Callebaut lost 110 first place citations in a single month. Valrhona lost 59. Of the five most cited brands in the category, Cocoa Supply was the only one to gain share. A rising result in a shrinking field proves very little. This one happened while the field more than doubled and the brands above it were losing ground.
Source: BusySeed GEO tracking, 100 monitored buyer prompts across ChatGPT, Claude, Gemini and Google AI Overviews, measured monthly from May 2026 to September 2026. Competitive share figures from the same tracking set.
Client pull quote from Cocoa Supply goes here, pending their approval.
Book a free 15-minute strategy session with founder Omar Jenblat. He'll look at your funnel live and show you exactly where the revenue is hiding. No deck, no fluff, no obligation.
