AI names Ally Bank most in banking, 31 points clear

When buyers ask AI assistants questions like “What are the best options for checking and savings accounts in 2026?” or “Best affordable checking and savings accounts for everyday banking under $50?”, a small set of bankingget cited every time. Most don't. This report measures which.

Who AI cites most

Ally Bank is cited in 72 of 100 prompt-engine pairs (72%). 95% confidence interval 66-82.

Concentration

The top 3 brands (Ally Bank, Chase, Capital One) capture 48% of all citations in this segment. 12 of 20 tracked brands are cited in fewer than 1 in 10 prompt-engine pairs.

Where the field sits

Of 20 brands tested: 1 in default-choice, 6 in first-encounter, 13 not yet cited. Overall, AI cites a brand from this segment in 15% of buyer-intent prompt-engine pairs.

Engine asymmetry

Marcus by Goldman Sachs is cited in 78% of Claude prompts but only 8% on Perplexity, visibility is engine-specific, not universal.

Notable absence

TD Bank is not yet cited, 0 prompts across all 100 prompt-engine pairs. A recognizable brand AI is not yet surfacing.

Average Panel Score across the segment is 18.

Analyst note. Ally Bank leads with a Panel Score of 76, far ahead of Chase's Panel Score 45.

Ally Bank is the leader in this segment with a Panel Score of 76. It performs well across engines, especially on ChatGPT and Claude at 88 each. Chase follows at Panel Score 45. Many brands are not making an impact, with 13 out of 20 falling into the invisible category. This creates a situation where Ally Bank is the primary choice for consumers.

Risk. Discover Bank has a Panel Score of 73 on Claude but is nearly invisible on Grok at 8, showing engine concentration risk.

Ranked by Panel Score, a 0 to 100 score that weights four buying phases: Discovery, Filtered Discovery, Comparison and Evaluation. Citations count fully and mentions count at half weight; engines are weighted equally. Spread is the gap between a brand's best and worst engine: under 15pp is durable, 50pp and up is engine-dependent. Engine columns count the prompts (out of 20) where that engine cited the brand as a recommendation.

Default choice (75+)

AI's go-to recommendation in banking. The tier other brands are competing into.

Repeat use (50 to 74)

Cited often enough to feel reliably present across prompts and engines. One signal away from default.

First encounter (25 to 49)

Discovered and cited occasionally, but visibility is inconsistent. Real to AI, not yet trusted.

Not yet cited (0 to 24)

AI does not surface this brand for buyer-intent prompts in banking.

How often each engine cites a brand from this category as a recommendation, averaged across all 20 brands tested. Perplexity and Grok search the live web; ChatGPT and Claude lean more on trained knowledge; Gemini blends both.

Citation rate by buying phase, averaged across all brands tested.

Each cell is the brand's citation rate for that phase, shaded darker as the rate rises. A solid dark row is a durable brand: AI cites it across the whole funnel. Dark on the left and pale on the right is awareness without recommendation, the demand-leak pattern. Pale on the left and dark on the right is a brand AI weighs in Comparison and Evaluation but does not surface in Discovery.

Derived metrics computed from the same data, showing how this segment behaves on AI search. The State of AI Search compares them across segments.

Each is a falsifiable, dated prediction we'll grade against next month's data. The full set across all segments lives on the State of AI Search page.

Aggregated across every brand, prompt and engine combination tested. These are the names AI surfaces when buyers ask about banking.

Ranked by mention count and engine breadth. These are panel candidates: brands AI treats as part of the category even though we are not measuring them yet. Open a row to read the AI quote that surfaced it.

The Panel Score is calibrated to a generic shopping-assistant prompt. Buyers don't arrive generically. We re-ran the same 20 canonical prompts five more times, each with a different buyer-persona signal in the system prompt: budget-conscious, premium, working professional, first-time, values-driven.

We re-ran the same 20 canonical prompts under eight demographic system-prompt variants: gender (men, women) crossed with generation (Gen Z 18-29, Millennial 30-45, Gen X 46-61, Boomer 62+). Age ranges follow Pew Research boundaries. The system prompt names the age range only; the labels here are descriptive.

Each is a category-specific decision drawn from this month's data, with the mapou service that puts it into practice.

1. Concentration risk

   In banking, AI effectively recommends 8.5 of 20 tracked brands. The top brand captures 22% of citations. The next 2 capture another 12%. Visibility is concentrated but not winner-takes-all.

   How mapou helps: GEO & Citation Architecture restructures your entity data so you can break into the top set.

2. Engine convergence

   Unusually for our panel, banking engines mostly agree. Cross-engine correlation is 0.75 (1.0 = perfect). Improvements on one engine are likely to lift others. Lower-effort optimization than categories where engines split.

   How mapou helps: AI Visibility Audit confirms your position is consistent across engines before you commit budget.

3. Persona-volatile category

   Banking rankings shift meaningfully by buyer persona. Ally Bank is the baseline #1 brand, but loses the top spot under at least one buyer signal (budget, premium, professional, first-time, values-driven). Top-3 overlap with baseline is only 67% across personas. Your baseline visibility number is incomplete.

   How mapou helps: Persona-Tuned Panel Score computes the visibility number for your actual buyer mix.

4. Visibility tier mix

   In banking, 1 of 20 tracked brands clear Panel Score 75 (default-choice tier). 13 are below Panel Score 25 (not yet cited). The strategy differs at each tier. If you are below 25, you need foundational visibility infrastructure before tactical optimization.

   How mapou helps: AI Visibility Audit identifies your tier; GEO & Citation Architecture moves you up.

The 20 canonical prompts span the four Panel Score phases (Discovery, Filtered Discovery, Comparison, Evaluation). The set is fixed at methodology v1.0 and reused every monthly run, so Panel Score changes are paired comparisons, not noise.

The exact prompt templates and phase-weighting formula are part of mapou's proprietary methodology, shared with paying clients alongside custom benchmarks for their brand. See the framework.