AIVO Brand Alpha: Putting a Dollar Value on How AI Allocates Demand
AI assistants now decide which brands a buyer hears about at the moment of choice. AIVO Brand Alpha puts a signed dollar figure on whether that decision is working for a brand or against it.
When small-business owners asked AI assistants to recommend a bank without naming one, the 30 US banks in our study were the final recommendation in just 15% of ChatGPT and Gemini answers. Across all 2,160 conversations, a single fintech was recommended outright more often than all 30 banks combined.
Those banks hold the relationships. AI is increasingly deciding who wins the new ones. Brand valuation has had no way to measure that shift in money. Today we publish one: AIVO Brand Alpha, set out in a new AIVO Standard working paper (DOI 10.5281/zenodo.23119452).
The blind spot in brand valuation
The established brand valuation frameworks measure a brand's hold on people. Interbrand, Kantar BrandZ and Brand Finance combine financial performance with measures of awareness, predisposition, loyalty and pricing power, and they do that well.
What none of them observes is the intermediary that now sits between the buyer and the choice. A buyer who asks an AI assistant which bank, insurer or moisturiser to choose receives a shortlist and usually a single recommendation. Among consumers who use them, generative AI assistants already rank as the second most influential purchase touchpoint, according to BCG's 2026 consumer research.
AI assistants have become a new allocator of demand. A brand the assistant consistently recommends captures that demand. A brand it leaves out loses it, whatever its awareness or its share of the market.
What Brand Alpha measures
Brand Alpha is the annual dollar difference between the AI-influenced revenue a brand captures and the revenue its market position predicts it would capture if AI assistants treated it neutrally.
\text{Brand Alpha} = \text{TAM} \times \text{APIR} \times (\text{OWR} - \text{MS})TAM is the category's annual revenue. APIR is the share of purchase decisions in the category that AI materially influences. OWR, the Organic Win Rate, is the share of unprompted buying conversations in which the assistant makes the brand its final recommendation. MS is the brand's market share.
A positive figure means the AI channel works harder for the brand than its size suggests. A negative figure means the brand's market position is not being carried through to the moment the AI makes its choice.
Take a hypothetical brand with 8% of its market that wins 3% of AI recommendations, in a category where AI influences $500m of spending a year. It falls five points short of parity, a Brand Alpha of minus $25m a year. A challenger with 2% of the same market and 6% of recommendations would show plus $20m.
Brand Alpha builds on two methods AIVO has already published. LLM Equity Valuation (LEV) estimates the AI-influenced revenue a brand captures, which is largely a function of size. Brand Alpha asks the question a board can act on: is AI helping or hurting us relative to our market position? Revenue at Risk is reported alongside it, measuring revenue exposed when a buyer asks for the brand by name and the assistant recommends a competitor.
Why market share is the baseline
Any measure of over- or under-performance needs an expectation to measure against. Brand Alpha uses market share: the share of recommendations a brand would receive if the AI expressed no preference of its own.
We chose it because it is the least arbitrary option available. Market share records what buyers actually bought. It is observed in purchases, so it is not contaminated by the AI behaviour being measured. It covers every brand in a category on one definition, and it is already reported for most categories. Equal shares, survey preference and the AI's own consideration set were each considered, and each fails at least one of those tests.
The baseline makes no claim that AI ought to reproduce market share. An assistant may favour a challenger because its product better fits what the buyer asked for. Brand Alpha measures the commercial consequence of the AI's choices for each brand and leaves the question of whether those choices are right to the brand and its customers.
What the first data shows
The first category is US small-business banking. On 18 September 2026 we ran 2,160 four-turn decision conversations on ChatGPT, Gemini and Perplexity, covering 30 banks across four situations: a startup opening its first account, a business switching provider, a credit-led relationship and international payments. American Banker reported the study on 24 September.
The inputs Brand Alpha needs from the AI side are already measured. In conversations where the owner named no bank, the 30 banks combined were the final recommendation in 15% of ChatGPT and Gemini answers and 18% of Perplexity answers. Only eight of the 30 won any unprompted recommendation at all. Across the full study, Mercury was recommended outright 594 times against 452 for all 30 banks together.
Banks still win the credit conversation. Outside it, the AI sends business owners elsewhere, and on the evidence so far most large banks will show a negative Brand Alpha. How negative, in dollars, is what the first table will publish.
What Brand Alpha is not
Brand Alpha is positional. It describes how AI assistants allocate demand today, and it makes no forecast of revenue, market share or returns. It is also not a brand valuation in the ISO 10668 sense: it values one channel's allocation of demand relative to a baseline, and it complements the established frameworks rather than competing with them.
Its central assumption is that AI-influenced purchases follow the AI's final recommendation share. That assumption is stated openly in the paper and has not yet been tested against purchase data. The paper sets out three tests to do so, beginning with a survey of small-business owners that records what the assistant recommended and what they actually chose, and it commits to revising Brand Alpha if the results do not support it.
Because Brand Alpha is in dollars, it scales with the size of a category. It should be compared between brands in the same category and for one brand over time. Comparisons across categories should use the AI Share Index, the ratio of AI recommendation share to market share, which the paper reports beneath every Brand Alpha figure.
What comes next
The first AIVO Brand Alpha table will cover US small-business banking, with a dollar value, an AI Share Index and Revenue at Risk for each bank. Its dollar figures will be published once category revenue and small-business market share are sourced to the standards the paper sets out, with the AI purchase influence rate shown as a range until it has been measured. Tables for insurance and US consumer goods will follow.
Banks, insurers and consumer brands that want to see their position before the tables are published can contact AIVO directly.
Reference
Sheals, P., & de Rosen, T. (2026). AIVO Brand Alpha: A Framework for Valuing Brands in the AI Recommendation Layer (Version 1.0) [Working paper]. AIVO Standard. Zenodo. https://doi.org/10.5281/zenodo.23119452
The working paper is DOI-anchored and openly licensed under CC BY 4.0. It has not been peer reviewed. It builds on LLM Equity Valuation (LEV): A Framework for Quantifying AI Purchase Recommendation Equity in Brand Acquisitions, Working Paper WP-2026-02, Version 2 (https://doi.org/10.5281/zenodo.22908818), and the AI Win-Rate Methodology v2.0 (https://doi.org/10.5281/zenodo.22663407).
Comments ()