Ingredients: which ones does AI recommend?

A supplier’s reputation alone will not secure a place in ChatGPT’s answers. An exploratory study by Pitman Comms, the communications and visibility consultancy, reveals differences depending on the expertise sought and raises questions about how companies communicate their capabilities.

A supplier recognised for its active ingredients may disappear from the recommendations as soon as the enquiry turns to biotechnology. This discrepancy, which says nothing about its scientific capabilities, is a recurring theme in the first Visibility Index Report, published on 2 November 2026 by Pitman Comms, founded by Briton Simon Pitman.

The study used six sourcing questions focused on the UK market, each submitted to ChatGPT three times on 18 October. Each one requested five suppliers without naming any companies. The five categories covered were innovative actives, functional ingredients, natural or botanical ingredients, sensory solutions and biotechnology. A sixth question concerned distributors offering technical support.

Uneven recognition of expertise

Croda Beauty appears in all fifteen ingredient shortlists. Several specialists are also named in every response within their respective fields: Givaudan Active Beauty and Provital for actives; Elementis, Lubrizol and Seppic for sensory, texture and rheology solutions. In biotechnology, only Croda Beauty and Vytrus Biotech feature in all three replies, out of nine suppliers named overall.

Distributors also feature prominently in the answers on functional ingredients, accounting for 12 of the 15 recommendations, including repeat appearances. For the manufacturers they represent, this raises a question of attribution: are the raw material’s origin and its developer’s expertise clearly identified?

Making expertise clear

For marketing teams, much of the report’s value lies in its proposed approach. Beyond checking whether a company is named, the task is to examine which capabilities AI associates with it. Does a recently launched ingredient appear in answers about its intended application? Are its benefits accurately described? Are technical data and application information readily available, including through distributors?

These checks fall within the remit of GEO (Generative Engine Optimisation), which encompasses all the methods designed to ensure the visibility of content in the responses provided by generative artificial intelligence systems and conversational engines. They can help guide revisions to a product data sheet or inform an expert article, provided the information gap has first been established.

The study remains a snapshot: its findings are neither a supplier ranking nor a measure of purchasing behaviour. They do, however, serve as a reminder that expertise recognised within the industry may not be fully reflected in an AI tool’s answers.