How AI assistants recommend banks to small businesses, and which banks they ignore.
When US small businesses ask AI assistants where to bank, fintechs win more than three recommendations in four. Of 18 major US banks measured, 12 are in severe AI deficit, and five were never recommended once in 1,080 conversations.
The Brand Alpha score runs from 0 to 100, where 50 means a bank is recommended exactly in line with its market share. Dollar figures show the estimated annual gain or loss for every $100m of a bank's new small-business banking revenue.
* See the notes under the ranking table.
* Provisional result, shown as a lighter bar. Score = 50 x AI Share Index, capped at 100.
AIVO US small-business banking study, 18 Sep 2026 · 1,080 conversations · market shares: Coalition Greenwich 2026 and FFIEC CRA 2024
The five banks at zero were not recommended in any of the 1,080 conversations.
U.S. Bank and Chase lead; 16 of 18 banks sit below parity, most of them far below.
| Rank | Bank | Brand Alpha score (50 = parity) | AI Share Index | Market share | AI share | Rating | Status | Annual impact per $100m revenue: mid (low to high) |
|---|---|---|---|---|---|---|---|---|
| 1 | U.S. Bank † | 72.5 | 1.45 | 2.6% | 3.80% | Provisional | +$3.7m(+$3.0m to +$4.3m) | |
| 2 | Chase | 64.0 | 1.28 | 12.1% | 15.46% | Provisional | +$2.3m(+$1.9m to +$2.7m) | |
| 3 | Huntington Bank † | 44.5 | 0.89 | 1.1% | 0.99% | Provisional | −$0.9m(−$0.8m to −$1.1m) | |
| 4 | M&T Bank † | 23.5 | 0.47 | 0.5% | 0.23% | Provisional | −$4.7m(−$3.9m to −$5.5m) | |
| 5 | PNC | 11.0 | 0.22 | 4.9% | 1.10% | Confirmed | −$7.1m(−$5.9m to −$8.4m) | |
| 6 | Truist | 4.5 | 0.09 | 3.9% | 0.36% | Confirmed | −$8.5m(−$7.0m to −$9.9m) | |
| 7 | Regions Bank † | 3.5 | 0.07 | 0.8% | 0.05% | Confirmed | −$8.7m(−$7.2m to −$10.2m) | |
| 8= | Bank of America | 2.0 | 0.04 | 6.2% | 0.27% | Confirmed | −$9.0m(−$7.4m to −$10.5m) | |
| 8= | Fifth Third Bank † | 2.0 | 0.04 | 0.8% | 0.03% | Confirmed | −$9.0m(−$7.4m to −$10.6m) | |
| 10= | Wells Fargo | 1.5 | 0.03 | 5.8% | 0.16% | Confirmed | −$9.1m(−$7.5m to −$10.7m) | |
| 10= | Zions Bank † | 1.5 | 0.03 | 0.4% | 0.01% | Provisional | −$9.1m(−$7.5m to −$10.7m) | |
| 12= | Citi † | 0.5 | 0.01 | 1.2% | 0.01% | Confirmed | −$9.3m(−$7.7m to −$11.0m) | |
| 12= | Synovus † | 0.5 | 0.01 | 0.4% | <0.01% | Confirmed | −$9.3m(−$7.7m to −$11.0m) | |
| 14= | Capital One † | 0.0 | 0.00 | 3.8% | 0.00% | Confirmed | −$9.4m(−$7.8m to −$11.1m) | |
| 14= | First Horizon Bank † | 0.0 | 0.00 | 0.5% | 0.00% | Confirmed | −$9.4m(−$7.8m to −$11.1m) | |
| 14= | Old National Bank † | 0.0 | 0.00 | 0.4% | 0.00% | Confirmed | −$9.4m(−$7.8m to −$11.1m) | |
| 14= | SouthState Bank † | 0.0 | 0.00 | 0.5% | 0.00% | Confirmed | −$9.4m(−$7.8m to −$11.1m) | |
| 14= | Silicon Valley Bank (First Citizens) † | 0.0 | 0.00 | 2.0% | 0.00% | Confirmed | −$9.4m(−$7.8m to −$11.1m) |
How to read it. AI share is each bank's share of the recommendations AI made. The AI Share Index divides it by market share (1.00 = fair share), and the Brand Alpha score is 50 × the index, capped at 100.
Most banks sit below fair share because the assistants usually answer with a fintech. Mercury alone receives more than a third of all recommendations, more than twice Chase's share.
All fintech and other non-bank providers combined: 77% of recommendations. All banks: 23%.
AIVO US small-business banking study, 18 Sep 2026 · 1,080 conversations on ChatGPT, Gemini and Perplexity
Chase is the only bank that competes with the fintechs at scale. Mercury, Wise, Bluevine, Relay, Brex and Airwallex were outside the 30 providers rated, so they have no Brand Alpha score in this edition.
ChatGPT, the assistant business buyers use most, recommends only Chase and U.S. Bank among the banks measured. Gemini spreads recommendations to regional banks, often answering credit questions with "a regional bank such as PNC or Truist".
| Bank | ChatGPT | Gemini | Perplexity | Combined |
|---|---|---|---|---|
| U.S. Bank | 1.87 | 0.80 | 0.43 | 1.45 |
| Chase | 1.81 | 0.26 | 0.96 | 1.28 |
| Huntington Bank | 0.00 | 2.44 | 2.25 | 0.89 |
| M&T Bank | 0.00 | 1.51 | 0.00 | 0.47 |
| PNC | 0.00 | 0.71 | 0.00 | 0.22 |
| Truist | 0.00 | 0.30 | 0.00 | 0.09 |
| Regions Bank | 0.00 | 0.22 | 0.00 | 0.07 |
| Bank of America | 0.00 | 0.09 | 0.31 | 0.04 |
| Fifth Third Bank | 0.00 | 0.14 | 0.00 | 0.04 |
| Zions Bank | 0.00 | 0.10 | 0.00 | 0.03 |
| Wells Fargo | 0.00 | 0.08 | 0.05 | 0.03 |
| Synovus | 0.00 | 0.03 | 0.00 | 0.01 |
| Citi | 0.00 | 0.03 | 0.00 | 0.01 |
AI Share Index on each assistant (1.00 = fair share). Capital One, First Horizon, Old National, SouthState and Silicon Valley Bank score 0.00 on all three. The combined figure weights assistants by business-buyer usage: ChatGPT 58.7%, Gemini 36.4%, Perplexity 4.9%.
Twelve of the 30 providers tested are not rated. Ten banks have market shares below 0.39%, the smallest AI share this sample can detect, so they cannot be judged above or below fair share. Two providers have no public market share.
| Provider | Market share (lending) | AI share |
|---|---|---|
| KeyBank | 0.29% | 0.35% |
| Citizens Bank | 0.24% | 0.13% |
| Webster Bank | 0.21% | 0.00% |
| Valley Bank | 0.20% | 0.00% |
| Live Oak Bank | 0.19% | 0.00% |
| BOK Financial | 0.17% | 0.00% |
| East West Bank | 0.10% | 0.00% |
| Associated Bank | 0.07% | 0.00% |
| Popular Bank | 0.03% | 0.00% |
| Axos Bank | <0.01% | 0.02% |
| Revolut | No public share | 0.37% |
| Square | No public share | 0.00% |
These banks can be rated with a larger sample or a regional study in their home markets, where their share is large enough to measure.
The index applies AIVO Brand Alpha v2.0 (doi:10.5281/zenodo.23183825) to AIVO's US small-business banking study of 18 September 2026.
Conversations. 1,080 unbranded four-turn small-business banking conversations, in which the user never names a bank, 360 on each of ChatGPT, Gemini and Perplexity. They cover four equally weighted situations: starting a business, switching banks, seeking credit, and banking internationally.
The full study ran 2,160 conversations. The other 1,080 began with the user naming a specific bank; those test whether naming a bank changes the outcome and are not used in this index.
Scoring the answer. The assistant's final answer is graded: a sole recommendation scores 1.00, a place in a list 0.50 shared, an undecided answer 0.25 shared, and absence 0. Assistants steered 90% of conversations to named providers.
Weighting. Assistants are weighted by business-buyer usage in Q4 2026: ChatGPT 58.7%, Gemini 36.4%, Perplexity 4.9%.
Market share. Coalition Greenwich 2026 small-business banking share is used for Chase, Bank of America, Wells Fargo, PNC and Truist. For every other bank, share comes from US regulators' 2024 small-business lending data: each bank's portion of $257.8bn in small-business loans originated by 690 reporting banks.
Benchmark stress test. For the five banks covered by both sources, the lending data rank them almost identically but put three at about half their Greenwich share. Every result was therefore rerun with lending shares scaled up by the median gap (1.88×). A verdict that changes direction under this test is marked provisional.
Dollar impact. Each bank's AI Share Index is converted to an annual figure per $100m of new small-business banking revenue. The model assumes 30% of decisions involve AI and a 37% recommendation effect. Low, mid and high cases apply factors of 0.70, 0.85 and 1.00. New revenue means revenue from new and switching relationships, because existing banking revenue is sticky.
AIVO Brand Alpha measures how often AI assistants recommend a brand compared with its share of the market, and what the gap is worth. It complements established brand valuations, and its full method is published openly for review and critique. Brand Alpha measures the emerging layer of competition, where AI assistants decide which brands buyers consider before they ever visit a website.
Sources: AIVO US small-business banking study (18 September 2026); Coalition Greenwich 2026 small-business banking share; FFIEC CRA 2024 disclosure file D1-1 (small-business loan originations); AIVO Brand Alpha v2.0 methodology.