What does this research mean for the field?

AI answer engines display substantial variance in brand recommendation, with Claude showing a 4x higher average Share of Voice compared to ChatGPT for identical queries about mid-market DTC brands. Novelty: ClaimNovelty.NOVEL_FINDING. Consensus alignment: ConsensusAlignment.NEUTRAL.

What question did this study set out to answer?

This paper explores AI brand recognition using a structured audit across multiple DTC brands and categories.

June 30, 2026Open Access

AI Recommendation Is Not Fame: Structural Findings from a 37-Brand, Multi-Engine Audit of Mid-Market DTC Brands

Key Points

This paper explores AI brand recognition using a structured audit across multiple DTC brands and categories.
Applied a multi-factor brand-recognition audit to a dataset of 37 DTC brands across five categories.
Used a five-factor measurement instrument scored on Share of Voice, Visibility Vitals, sentiment, and citation rank.
Focused on the Food & Nutrition subset where identical protocols allowed for controlled queries across multiple engines.
Claude's average Share of Voice is 39, ChatGPT's is 9.5, Gemini's is 31.5, and Perplexity's is 19.5 for the same brands and queries.
The 4x gap between Claude and ChatGPT represents a significant finding but its cause remains unclear.

Abstract

Third paper in a three-part series on AI brand recognition: The Mention Density Model (DOI: 10.5281/zenodo.20379032) — why recognition happens A Multi-Factor Brand-Recognition Audit for AI Answer Engines (DOI: 10.5281/zenodo.20538106) — how to measure it This paper — what the data reveals Two companion papers established a mechanism and a measurement instrument for studying AI brand recognition: The Mention Density Model proposed that citation rate is governed by a brand's mention density across training and retrieval corpora; A Multi-Factor Brand-Recognition Audit (MFBRA) supplied a five-factor measurement instrument scored on Share of Voice, five Visibility Vitals, sentiment, and citation rank. This paper applies that instrument to a first-party dataset of 37 audited DTC brands across five categories and reports what the data looks like — not what it proves. This paper is explicitly an exploratory audit, not a test of the mention density model. The dataset has a structural limitation that prevents any such test: audit depth varies across categories, query sets differ by category, and each brand was audited once with no repeated draws. Cross-category comparisons are not valid, and the data cannot separate stable signal from run-to-run variance. Within those constraints, one result stands out. In the Food & Nutrition subset — the only part of the dataset where all 10 brands ran identical protocols (10 queries x 4 engines, no errors) — Claude's average Share of Voice is 39, compared to ChatGPT's 9.5, Gemini's 31.5, and Perplexity's 19.5. The 4x gap between Claude and ChatGPT for the same brands on the same queries is the strongest quantified result in this dataset. Its cause is unresolved: it could reflect genuine per-brand preference differences, or answer breadth — Claude listing more brands per response than ChatGPT. The contribution of this paper is the dataset itself, the disclosed methodology, and the open questions it surfaces for a properly controlled follow-up study.

Read Full Paperexternally

Ask AI

Helpful

Bookmark

View Full Paper