How AI Search Is Recommending Personal Loans and Online Lenders: Monthly Trends

Executive Summary

Upstart took the category lead in October 2026, closing at 79.8% valid recommendation coverage, and the top two positions are now separated by just 0.4 points. SoFi sits second at 79.4%, up 10.9 points from 68.5% in September. This is the tightest the top of the field has been across the four tracked months and represents a leadership change from September, when SoFi held first place.

The strongest riser against the July baseline is Upstart, up 7.3 points from 72.5%, with its coverage swinging from 67.7% in September to 79.8% in October. SoFi follows at 5.3 points above its July baseline of 74.1%, having recovered from an August trough of 64.5%. The October month alone produced two significant month-over-month movements: SoFi up 10.9 points and Upstart up 12.1 points against September. Both are now above their baseline readings.

The sharpest decline across the series belongs to Happy Money, down 8.7 points from 17.0% in July to 8.3% in October, its third consecutive monthly decline and the only three-month losing streak in the field. Achieve Home Loans also records an 8.3-point fall from baseline, from 8.3% in July to 0.0% in October, reflecting the tracking transition to Achieve that began in September. Achieve itself posted 12.3% in October, down 3.3 points from 15.6% in September, a move within normal variation.

The category result is mixed rather than orderly: the leader changed, three brands are classified as significant risers over the series, and two are significant decliners, while Best Egg and Upgrade held their positions within normal month-to-month variation. October's qualified observation count is 569, down 21 from July's 590.

Each monthly run begins with 800 prompt-surface observations (644 unique questions in October, 654 in September, 628 in August, 583 in July) across the benchmark's defined AI surface universe. Of those, 800 mentioned a tracked brand or competitor; 687 were relevant and 113 were irrelevant in October, versus 700 relevant and 100 irrelevant in September, 738 relevant and 62 irrelevant in August, and 723 relevant and 77 irrelevant in July. The public metrics use the 569 qualified observations in October and the 590 qualified observations in July that survive both qualification stages.

Key Findings

AI Response Inconsistency Alerts

Questions This Section Answers

• Which lenders had conflicting product information across AI platforms in October?
• What kinds of product facts were most often contradicted between AI surfaces?
• Which source pages were cited behind the conflicting rate and eligibility claims?

Five high-severity factual inconsistencies were detected across four AI platforms in October, affecting two tracked lenders: SoFi and Upstart. In every case, the conflicting claims concern concrete product facts such as pricing ranges, eligibility thresholds, and loan maximums. AI platforms provided conflicting information about the same lender within the same question, meaning a loan seeker could receive different answers to the same prompt depending on which surface they used.

SoFi

Three high-severity pricing conflicts involve SoFi's stated APR range. When asked "Who has the best installment loans?", Gemini stated SoFi's APR range runs from 6.99% to 35.49%, while Copilot stated the range runs from 8.99% to 29.99%. Neither range can be simultaneously correct for the same product. Gemini's response cited MoneyLion, Money.com, and Bankrate pages including "Best Installment Loans of 2026: Compare Rates, Terms and Fees" and "Best Personal Loans of August 2026." Copilot's response cited MoneyLion alongside Finder and a BankingVibe page, "Best Personal Loans of 2026," which was flagged as the source carrying the 8.99% to 29.99% claim.

When asked "What are good installment loans?", ChatGPT stated SoFi's APR range runs from 6.49% to 35.49%, while Copilot stated a range of approximately 6% to 20%. The two answers disagree on the ceiling by more than 15 points. ChatGPT's response referenced NerdWallet's personal loans page, NerdWallet's average personal loan rates article, and the Consumer Financial Protection Bureau's explainer on the difference between interest rate and APR. Copilot's response cited WalletHub, MoneyLion, and NerdWallet's installment loans page. A NerdWallet awards page stating "SoFi Personal Loan Est. APR 6.49-35.49%" was flagged as the source supporting the ChatGPT-side claim.

A third conflict on the same question pairs Gemini's 6.99% to 35.49% range with Copilot's approximately 6% to 20% range. Gemini's response cited a single MoneyLion page, while Copilot again drew on WalletHub, MoneyLion, and NerdWallet. Read together, the three conflicts show the same pattern: AI platforms handling the same lender and the same product characteristic disagree on where SoFi's pricing starts and, more sharply, on where it ends.

Upstart

Two high-severity conflicts involve Upstart's product terms. When asked "What is the lowest credit score for a debt consolidation loan?", ChatGPT stated that Upstart lists a 620 minimum credit score, while Gemini stated that Upstart does not have a strict minimum credit score requirement. The two claims are directly opposed on an eligibility threshold that materially affects who can apply. ChatGPT's response cited Forbes Advisor's debt consolidation loans page and an Experian article on debt consolidation with bad credit. Gemini's response cited LendingTree's debt consolidation loans for bad credit page, the same Forbes Advisor page, and an Achieve page on credit score requirements for consolidating debt. The LendingTree page was flagged as the source carrying the no-formal-minimum claim, with the excerpt stating that Upstart and OneMain Financial "don't publish a formal minimum credit score requirement."

When asked "Which one is better, Upgrade or Upstart?", AI Overviews produced two conflicting answers on the same surface. One answer stated that Upstart offers larger loan amounts, up to $75,000 versus $50,000 for Upgrade. The other stated that both Upgrade and Upstart offer personal loans up to $75,000. These claims cannot both be true. Both responses drew on Credible and MoneyLion comparison pages, with the first also citing a Finder comparison and the second citing a MoneyLion page comparing Upstart with LendingClub. A Finder page stating that Upstart "offers higher loan amounts (up to $75,000 vs. Upgrade's $50,000)" was flagged as supporting the first claim, alongside WalletHub and CreditNinja excerpts, while a third flagged excerpt lists Upgrade's range as $1,000 to $50,000. This conflict matters commercially because it sits inside a direct head-to-head comparison prompt, exactly the question type the benchmark's qualified set does not currently capture.

Benchmark Context

Questions This Section Answers

• How much did the qualified observation count change between July and October?
• What does the qualification funnel show about relevance filtering over the period?

The report separates the raw collection universe from the qualified analysis set. Brand-level recommendation percentages are calculated within the qualified benchmark set.

The qualification funnel tightened between July and October. Unique questions rose from 583 to 644 while irrelevant prompts grew from 77 to 113, and the qualified observation count fell from 590 to 569. September's funnel recorded 654 unique questions and 700 relevant prompts.

Benchmark-Level Metrics

The recommendation-shaped answer share climbed sharply to 63.1% in October, and the valid recommendation shortlist share reached 89.3%. August and September sat in the low 40s for recommendation-shaped share, so October's jump marks a change in what AI systems produced, not just in how brands were ranked. The category leader by coverage changed hands from SoFi to Upstart.

AI Recommendation Trend

Questions This Section Answers

• Which brands gained or lost valid recommendation coverage between July and October?
• Which brands moved beyond normal variation, and which held their positions?
• What does the compressed 0.4-point gap between Upstart and SoFi mean for the category leader picture?

Two Brands Now Share the Top of the Field Within Half a Point, While the Rest of the Category Holds Its Positions

No single brand's move explains the category result on its own. SoFi and Upstart both cleared their significance thresholds against baseline, Happy Money and Achieve Home Loans both declined significantly, Achieve rose on the strength of its tracking transition, and Best Egg and Upgrade held within normal variation. The net effect is a top pair that has compressed to a 0.4-point gap while the bottom of the field continues to lose ground.

What Changed This Month

Questions This Section Answers

• Why did Upstart overtake SoFi for the category lead in October?
• How does SoFi's first-choice rank-one strength compare with its second-place coverage rank?
• What drove Happy Money's third consecutive monthly decline in coverage and presence?

Upstart

Upstart moved from second place in September to first in October, posting 79.8% valid recommendation coverage against 67.7% in September, a 12.1-point gain. Against the July baseline of 72.5%, the brand is up 7.3 points, a significant rise. Upstart's coverage path across the series runs 72.5%, 69.9%, 67.7%, then 79.8%, meaning October reversed a two-month slide in a single step.

The gain is broader than coverage alone. Upstart's recommended top-3 rate rose 15.8 points against baseline, from 28.8% to 44.6%, and its rank-one rate rose 5.2 points from 7.5% to 12.7%. Its raw mention presence rate reached 91.0%, the highest in the tracked set, with 518 of 569 qualified observations mentioning the brand.

The distinction worth noting is that the rank-one improvement, while large, leaves Upstart far behind SoFi's 44.6% rank-one rate: Upstart is placed first in 72 of 569 qualified observations, while SoFi is placed first in 254. Upstart's lead is therefore a coverage lead, not a first-choice lead.

Highest-priority diagnostic: Which prompt types and surfaces drove the coverage swing from 67.7% to 79.8%, and does the top-3 gain hold across all six AI surface families?

SoFi

SoFi holds second place at 79.4% valid recommendation coverage, up 10.9 points from 68.5% in September and up 5.3 points from the July baseline of 74.1%. The brand is now above its baseline reading for the first time in the series, having dipped to 64.5% in August before recovering. Its September-to-October move is significant, and its baseline-to-current move is significant as well.

Placement strength carried the month. SoFi's recommended top-3 rate rose from 50.3% in September to 69.8% in October, and against baseline it is up 11.0 points from 58.8%. Its rank-one rate climbed to 44.6%, up 10.7 points from 33.9% in July and the highest rank-one reading in the category. SoFi ranked first in 254 of 569 qualified observations and appeared in 506, with 49 neutral mentions and no negative mentions.

The distinction is that SoFi lost the coverage crown while strengthening its first-choice position: it leads every rank-one and top-3 comparison in the field even as Upstart edges it on overall coverage. The benchmark also recorded three high-severity pricing inconsistencies involving SoFi's APR range across ChatGPT, Copilot, and Gemini in October, which is a presence and accuracy question separate from its recommendation strength.

Highest-priority diagnostic: Which prompts still place SoFi outside the recommendation set, and which cited source pages drive the conflicting APR ranges across surfaces?

Happy Money

Happy Money declined 8.7 points from 17.0% valid recommendation coverage in July to 8.3% in October, the largest baseline-to-current drop in the field. The series runs 17.0%, 12.8%, 10.3%, then 8.3%, a three-month losing streak and the only one of its kind among tracked brands. Its September-to-October move of 2.0 points is within normal variation, so the decline is cumulative rather than sudden.

The retreat shows in presence as well as coverage. Happy Money's raw mention presence rate fell 9.2 points against baseline, from 18.3% to 9.1%, and its valid recommendation count of 47 out of 569 qualified observations is the smallest in the tracked set. Its recommended top-3 rate is 3.3% and its rank-one rate is 0.0%, down 0.2 points from a July baseline of 0.2%.

The small counts matter here: 47 valid recommendations and 52 present mentions mean individual percentage movements carry more weight than the same movements would for a brand with several hundred qualifying observations. The notable-gaps view shows Happy Money trailing Upstart by 71.5 points in October, a gap that has widened every month in the series.

Highest-priority diagnostic: Which surfaces or prompt types account for the sustained presence decline, and is the loss concentrated on specific AI platforms?

Achieve and Achieve Home Loans

Achieve recorded 12.3% valid recommendation coverage in October, down 3.3 points from 15.6% in September, a move within normal variation but still 12.3 points above the July baseline of 0.0%. That baseline figure reflects the brand's entry into the tracked roster in September; the related tracking name, Achieve Home Loans, moved from 8.3% in July to 4.6% in August, then to 0.0% in both September and October, an 8.3-point decline from baseline that is classified as significant.

Achieve's raw mention presence rate is 18.4% in October, and its recommended top-3 rate is 3.2%, up from 2.1% in September. Its rank-one rate is 0.7% with 4 first-place placements out of 569. Its valid recommendation count is 70, down from 91 in September, reflecting both the coverage dip and the smaller October qualified set.

The distinction is that two rows in this benchmark describe the same underlying brand under different names: Achieve's 12.3% and Achieve Home Loans' 0.0% are a measurement transition, not two separate competitors. The notable-gaps view also shows Achieve's margin over Happy Money swinging from -17.0 points in July to +4.0 points in October.

Highest-priority diagnostic: Which legacy prompts still surface the previous name, and which surfaces drove the presence gain relative to the July baseline?

Best Egg

Best Egg held third place at 29.5% valid recommendation coverage, up 1.4 points from the July baseline of 28.1% and down 0.6 points from 30.1% in September. Both moves fall within normal month-to-month variation, and the series runs 28.1%, 29.0%, 30.1%, then 29.5%.

The placement story is stronger than the coverage story. Best Egg's recommended top-3 rate rose 6.6 points against baseline, from 6.6% to 13.2%, while its rank-one rate held at 1.6%. Its raw mention presence rate sits at 37.8%, down 3.7 points from 41.5% in July but within normal variation. Its valid recommendation count is 168 out of 569 qualified observations.

The distinction is that Best Egg is being recommended in more top-three positions while being mentioned in fewer answers overall, meaning the improvement comes from stronger placement inside existing visibility rather than from broader exposure.

Highest-priority diagnostic: Which prompt types drove the top-3 improvement, and does the pattern hold across all six AI surface families?

Upgrade

Upgrade held at 71.7% valid recommendation coverage in October, down 0.2 points from the July baseline of 71.9% and up 6.6 points from 65.1% in September, a significant month-over-month recovery that still leaves the brand flat across the series. Its recommended top-3 rate rose 5.5 points against baseline, from 41.4% to 46.9%, while its rank-one rate dipped 1.2 points to 7.6%.

Upgrade's raw mention presence rate is 81.2%, down 3.9 points from 85.1% in July, within normal variation. Its valid recommendation count is 408, and its recommended top-10 rate is 64.9%. The notable feature of Upgrade's series is how completely it round-tripped: 71.9%, 66.6%, 65.1%, then 71.7%.

The distinction is that October's gains brought Upgrade back to where it started rather than forward, and its top-3 rate remains below its coverage rate, indicating the brand is frequently in consideration sets but less often in the top three.

Highest-priority diagnostic: Which prompts produced September's dip and October's recovery, and which competitor captured the top-3 placements during the dip?

Buyer-Intent Interpretation

Questions This Section Answers

• Which buyer-intent clusters were captured in the October qualified observations?
• What do the empty Pricing & Value and Multi-Brand Comparison classes mean for what the benchmark can measure?
• How do the October inconsistency findings relate to the question types the qualified set does not capture?

The October qualified observations all fell into the Brand Recommendation class, leaving the Pricing & Value and Multi-Brand Comparison classes empty in the qualified benchmark set. This is a meaningful limitation given what surfaced outside it: the October inconsistency data records conflicting APR ranges for SoFi across three surfaces and a conflicting head-to-head loan-maximum claim for Upstart and Upgrade inside an AI Overviews comparison answer. Those are exactly the questions the two empty classes are designed to measure, and they are currently being answered inconsistently in the raw collection without appearing in the qualified denominator.

The benchmark therefore measures which brand an AI recommends for a general loan-seeking prompt. It cannot yet quantify how AI systems characterize rates and fees, or which brand wins when a user names two lenders side by side, even though the inconsistency record shows both question types are live in the market.

Brand Opportunity Summary

Questions This Section Answers

• Which brands showed notable coverage gains or declines in October, and what deserves attention next?
• What does the summary identify as the highest-priority diagnostic for each tracked lender?

The benchmark identifies where attention is warranted; a company-level analysis is needed to explain why.

Evidence Behind the Benchmark

The aggregate metrics are built from prompt-level observations covering query, surface, recommendation outcome, rank, sentiment, and citations where those are exposed. Company-level analysis can go deeper into prompt, competitor, surface, and evidence patterns. Source presence is not automatically treated as proof of causation. In October, five high-severity factual inconsistencies were recorded across four AI platforms, and the flagged source pages behind them included lender comparison pages, rate aggregators, and personal finance publishers, which is the level of detail a company-level audit can examine systematically.

About This Benchmark

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

• AI Visibility Industry Market
• AI Visibility Industry Market Methodology
• AI Visibility Industry Market Metrics
• AI Visibility Industry Market Standards

Report-Specific Interpretation Notes

• Small-count movement: Happy Money (47 valid recommendations), Achieve (70), and Best Egg (168) carry smaller denominators than SoFi (452) and Upstart (454), so their percentage movements warrant more caution.
• Tracking-name transition: Achieve replaced Achieve Home Loans in the tracked roster in September; the two rows describe the same underlying brand under different names.
• Qualified denominator versus raw collection: All percentages use the 569 qualified observations in October and the 590 in July, not the 800 raw prompt surfaces.
• Directional analysis: Month-over-month movement identifies changes worth investigating; it does not by itself establish the cause of those changes.

Next Step

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

Beneath the aggregate percentage sit questions the benchmark alone cannot answer: which high-intent prompts each brand wins, which competitor takes the recommendation when a brand loses, what attributes AI systems associate with each option, and which external sources shape those answers. October's inconsistency record shows this is not hypothetical: AI platforms returned different APR ranges for the same lender and conflicting loan maximums inside a direct comparison question, and the benchmark's qualified set does not currently capture those question types.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy, turning the benchmark's category-level signals into actionable intelligence.

Request an AI visibility audit