For the majority of food & drink categories, the market-leading brand is not the most prominent in answers from the major AI engines, analysis by Tastewise has found.
In 28 of 40 categories analysed, the ‘most visible’ brand according to ChatGPT, Claude and Gemini was not the brand with the biggest market share in that category, the study shows.
In nine out of 40 categories, the actual market leader failed to appear in AI chatbots’ top 10 brands, and in 10 out of 40 categories the top brand offered up by AI was not in the market’s top five.
For example, in energy drinks, Celsius was found to lead AI-given recommendations, appearing in more than 77% of answers, ahead of Monster (67%) and Red Bull (60%). Meanwhile, Frank’s RedHot trumped Hellmann’s in AI responses to a question relating to condiments and sauces.
“The pattern across categories is consistent: market size and shelf presence do not predict AI visibility,” said Alon Chen, CEO and co-founder of Tastewise. “Brands are winning and losing the AI shelf on entirely different terms than the ones they’ve spent decades optimising for.”
“Your customers stopped Googling ‘best snack’ and started asking Claude. For food and beverage, AI answers are the new shelf, and most brands have no idea if they’re on it,” he added.
The analysis also questioned the AI engines with ‘better for you’ type questions – with the prompts including phrases like ‘healthiest’, ‘gluten-free’, ‘high-protein’, ‘clean label’ and ‘best protein bar for muscle’ – and compared them with ‘broad ask’ type questions like ‘best tasting’ and ‘best value’.
On the ‘better for you’ questions, the actual market leader loses 6.2 percentage points of visibility and the market’s top five as a group lose 6.9 points of their share of mentions.
In 30 out of the 40 categories analysed, the most visible brand changes with this type of question.
For example, Lindt displaces Tony Chocolonely as the most visible brand following a ‘better for you’ prompt, Halo Top displaces Häagen-Dazs, and Wendy’s trumps Burger King.
“The rule the data supports is not ‘challengers beat incumbents on specific questions’. It is: the brand that owns the attribute in the question wins the answer – and incumbents usually do not own the health attribute,” Tastewise said.
Many brands are investing in what’s sometimes called AI SEO, LLM optimisation, or generative engine optimisation, to ensure they are more visible and frequently recommended by AI engines.
LLMs tend to prioritise mentions in authoritative published content such as news websites and academic papers, and user-generated content such as product reviews and online forums like Reddit.
Unlike with social media and SEO, this makes it much harder for brands to pay their way into LLM search results.
OpenAI’s ChatGPT does support brand ads – which launched to UK users in June, despite CEO Sam Altman previously saying that he “hates ads” and considers “ads as like a last resort for us for a business model” and is “uniquely unsettling to me” – which appear below the LLM’s response to a user’s question.
Earlier this year, Anthropic – the company behind Claude – vowed that the platform would remain ad-free. “Our users won’t see ‘sponsored’ links adjacent to their conversations with Claude, nor will Claude’s responses be influenced by advertisers or include third-party product placements our users did not ask for,” it said.
Google avoids placing native ads inside standalone Gemini chats. However, it does embed ads directly into AI search features like AI Overviews.
Tastewise is an agentic intelligence system built for food and beverage. Its AI agents take brands from a product concept through to a shelf-ready plan, rather than simply flagging trends. It is used by Waitrose, PepsiCo, Kraft Heinz and Nestlé, and by more than 80% of the world’s leading food and beverage brands.
It has recently launched Famewise, a AI brand visibility tracker.
A third of consumers now prefer to start or support shopping in AI chat apps, according to a February Valtech survey. At the same time, trust in AI agents is, for now, low. March research from Billion Dollar Boy found only 35% of consumers believe AI provides quality product recommendations, while 63% are uncomfortable with AI using their browsing and purchase history to deliver recommendations.
No comments yet