How AI Search Is Recommending Savings Account: Monthly Trends

Executive Summary

SoFi leads AI search recommendation coverage in Savings Account for September 2026, appearing as a valid recommendation in 76.4% of qualified observations, up 1.8 points from August 2026's 74.6%. The gap to second-place Capital One widened to 3.8 points, as Capital One's coverage slipped 2.1 points to 72.6% over the same period. This is a stable leadership position, with no change at the top between months.

The largest upward mover this month was Axos Bank, whose valid recommendation coverage rose 3.3 points from 54.5% in August 2026 to 57.8% in September 2026. This movement was not flagged as significant in the benchmark's classification, but it pushed Axos Bank's rank-one rate up 0.8 points to 5.9% and its presence rate up 1.3 points to 62.8%.

The sharpest decliner was Marcus by Goldman Sachs, which fell 6.6 points from 54.8% in August 2026 to 48.2% in September 2026. This was a significant decline, one of two in the category this month. EverBank also posted a significant decline, down 4.5 points to 17.7%.

This month's category-level story is one of measured movement rather than wholesale change. Most brands moved within expected ranges, but the significant declines for Marcus by Goldman Sachs and EverBank, alongside Axos Bank's climb, represent the most notable shifts in the mid-tier this month. The benchmark's category summary confirms movement this month, with no significant risers and two significant decliners.

Each monthly run begins with 800 prompt-surface observations (586 unique questions in August 2026 and 606 in September 2026) across the benchmark's defined AI/search surface universe. Of those, all 800 in each month mentioned a tracked brand or competitor; 732 were relevant and 68 were irrelevant in both months. The public metrics use the 712 qualified observations in August 2026 and 716 in September 2026 that survive both qualification stages.

Key Findings

Benchmark Context

The Savings Account benchmark separates the raw collection universe from the qualified analysis set. In August 2026, 800 prompt-surface observations produced 712 qualified observations; in September 2026, 800 prompt-surface observations produced 716 qualified observations. Brand-level recommendation percentages are calculated within each month's qualified set.

Benchmark-Level Metrics

The metrics below summarize the qualified benchmark at the aggregate level for both months.

The recommendation-shaped answer share and valid shortlist share both declined between August and September 2026, even as the qualified observation count rose slightly. The data does not establish why the answer-format mix shifted; it is a pattern worth monitoring alongside individual brand movements.

AI Recommendation Trend

SoFi Holds the Lead While the Competitive Middle Repositions

SoFi remains the coverage leader for a second consecutive month, with Capital One holding second and Ally Bank holding third. The more notable activity this month occurred outside the top three, where Marcus by Goldman Sachs and EverBank each posted significant declines while Axos Bank continued to climb.

The most notable shift below the top three was the widening gap between Axos Bank and Marcus by Goldman Sachs, from 0.3 points in August 2026 to 9.6 points in September 2026, alongside Marcus by Goldman Sachs and EverBank's significant declines this month.

What Changed This Month

SoFi: Steady Leadership Built on Stronger Placement

SoFi's valid recommendation coverage rose from 74.6% in August 2026 to 76.4% in September 2026, an increase of 1.8 points. The brand remains the category leader, with its raw mention presence up 4.5 points to 87.2%, meaning SoFi is appearing in more AI answers overall.

SoFi's placement quality also moved. Its top-3 recommendation rate rose 8.1 points to 40.4%, and its rank-one rate climbed 3.7 points to 20.7% (148 rank-one recommendations in September 2026, compared with 121 in August 2026). This means SoFi is not just present more often, it is being recommended more prominently.

The distinction is between presence and recommendation strength. SoFi's presence rate (87.2%) now exceeds its coverage rate (76.4%) by more than 10 points, indicating that while it appears in most answers, it is not always the named recommendation. The top-3 rate crossing 40% is the strongest signal, a threshold no other brand in this category reached in September 2026.

Highest-priority diagnostic: Which prompt types are driving SoFi's top-3 gains, and are they concentrated in specific AI surfaces or spread across all six?

Axos Bank: Steady Climb Reshapes the Mid-Tier

Axos Bank's valid recommendation coverage rose from 54.5% in August 2026 to 57.8% in September 2026, a 3.3-point increase. This movement was not flagged as significant, but it places Axos Bank clearly in the top half of the category, now 9.6 points ahead of Marcus by Goldman Sachs, a gap that widened from just 0.3 points in August 2026.

Placement metrics followed the same path. Axos Bank's top-3 rate rose 1.9 points to 20.4%, and its rank-one rate rose 0.8 points to 5.9% (42 rank-one recommendations in September 2026, up from 36 in August 2026). Its presence rate rose 1.3 points to 62.8%, alongside the gains in placement.

The commercial consequence is that Axos Bank has moved from a mid-pack player to a clear fourth position, now within 10.9 points of Ally Bank in third. The brand's net sentiment rose to 97.3% in September 2026, alongside rising top-3 and rank-one rates.

Highest-priority diagnostic: Which specific savings account sub-topics or question phrasings are associated with Axos Bank's improved placement, and which surfaces are contributing most?

Marcus by Goldman Sachs: Significant Decline Breaks the Upper Tier

Marcus by Goldman Sachs experienced the sharpest drop in September 2026, with valid recommendation coverage falling 6.6 points from 54.8% in August 2026 to 48.2%. This was a significant decline, ending the brand's position in the top half of the category and dropping it to 5th place.

The decline was accompanied by weakening placement metrics. Top-3 rate fell 3.2 points to 13.8%, rank-one rate fell 2.9 points to 2.4% (down from 5.3% in August 2026, with rank-one count down from 38 to 17), and presence rate declined 3.4 points to 55.9%. The brand's net sentiment also slipped, from 96.9% to 94.3%.

The gap analysis shows Marcus by Goldman Sachs losing ground against multiple competitors at once. The gap to SoFi widened from 19.8 points to 28.2 points, the gap to Axos Bank inverted from a 0.3-point lead to a 9.6-point deficit, and CIT Bank moved from 9.6 points behind to 1.0 point behind. This is a broad-based repositioning rather than a single-competitor effect.

Highest-priority diagnostic: Which competitor is appearing in place of Marcus by Goldman Sachs, and what evidence sources or product attributes are associated with the shift?

EverBank: Significant Decline in an Already Low Base

EverBank's valid recommendation coverage fell 4.5 points from 22.2% in August 2026 to 17.7% in September 2026, a significant decline. With only 127 valid recommendations in September 2026 (down from 158 in August 2026), the brand remains in the lower tier and is now approaching the range occupied by Synchrony Bank and Barclays.

The decline was primarily a presence story. Raw mention presence fell 4.2 points to 20.8%, while top-3 and rank-one rates each rose slightly. This means EverBank's challenge is visibility, not recommendation quality when it does appear: it is being mentioned and recommended less often overall.

The distinction matters for prioritization. EverBank's top-3 rate of 4.9% and rank-one rate of 1.1% in September 2026 are comparable to or better than some brands with higher coverage. The erosion is in the breadth of AI answers where EverBank appears at all, a different problem from weak placement.

Highest-priority diagnostic: Which AI surfaces reduced EverBank's mention frequency most, and what content or source patterns could restore presence without sacrificing placement quality?

Buyer-Intent Interpretation

In September 2026, all 716 qualified observations fell into the Brand Recommendation cluster, the same pattern as August 2026. The benchmark's defined universe for this vertical is currently capturing explicit recommendation requests rather than pricing or comparison queries.

The practical consequence is that the public benchmark measures who AI systems name first and most often when shoppers ask for a recommendation, but it cannot yet answer which brand wins on rate transparency, fee comparison, or direct head-to-head showdowns. Those commercial questions remain open and would require a dedicated study design targeting pricing and comparison language.

Brand Opportunity Summary

The benchmark identifies where attention is warranted across all ten tracked brands; a company-level analysis is needed to explain why these patterns are emerging.

Evidence Behind the Benchmark

The aggregate metrics are built from prompt-level observations capturing the query, the AI surface, the recommendation outcome, the rank when a brand is named, the sentiment expressed, and the citations where exposed. Company-level analysis can go deeper into prompt wording, competitor substitution patterns, surface-by-surface performance, and the external evidence sources shaping each answer. Source presence in an AI answer is not automatically treated as proof of causation; it is a signal of which evidence the model weighed.

Report-Specific Interpretation Notes

• Movement is measured against the qualified benchmark set, not the raw collection universe. A change in coverage can reflect either a change in AI behavior or a change in the mix of questions that qualified.
• Small counts matter. Brands such as Barclays, with 30 valid recommendations, can show percentage movement that looks meaningful but rests on a small base of observations. Interpret with caution.
• Movement shown here identifies where attention is warranted, not the cause of that movement.
• Presence and recommendation are distinct signals. A brand can be mentioned often (high presence) yet named as the top pick rarely (low rank-one rate), which points to very different strategic problems.

About This Benchmark

This report is part of the LLM Authority Index AI Market Discovery research program.

• AI Industry Market Discovery
• AI Industry Market Discovery Methodology
• AI Industry Market Discovery Metrics
• AI Industry Market Discovery Standards

Next Step

The Public Benchmark Shows Where a Brand Is Winning or Losing. A Company-Level Audit Shows Why.

Beneath the aggregate coverage percentages sit the questions that matter most: which high-intent savings account prompts is a brand winning, which competitor appears in its place when it loses, what product attributes do AI systems associate with each option, and which external sources are shaping those answers. The September 2026 data shows significant movement for two brands and a steady climb in the mid-tier, but the public benchmark cannot explain the prompts, surfaces, or evidence patterns behind those shifts.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy. It answers not just what changed, but why it changed and what to do next.

Request an AI visibility audit