How AI Search Is Recommending Bad Credit Loans: Monthly Trends

Executive Summary

Upstart extended its coverage lead in October 2026, but the story of the month is the second tier. Upstart's valid recommendation coverage reached 88.9% in October 2026, up 13.9 points from the July 2026 baseline of 75.0% and 13.5 points above its September 2026 level of 75.4%. That is the largest single-month move in the four-month series. The gap to the next brand, Avant at 78.0%, stood at 10.9 points.

Avant posted the strongest full-series gain among the chasing group, rising 9.5 points from 68.5% in July 2026 to 78.0% in October 2026, with 12.0 points of that arriving in the latest month alone. OneMain Financial rose 10.5 points across the same window to 71.0%, and Upgrade rose 9.8 points to 70.2%. All three are classified as significant risers against the baseline. The practical consequence is a tighter cluster of brands now competing for placement behind a leader that also grew.

The sharpest decline came from Universal Credit, down 5.4 points from 19.3% in July 2026 to 13.9% in October 2026, a two-month downward streak. Achieve fell 4.2 points over the series to 6.1%, a three-month decline that is the longest sustained drop in the category.

The category was not stable this month. Four brands recorded significant baseline increases and two recorded significant baseline declines, with the largest prior-month movement belonging to Upstart at 13.5 points up. The notable gaps widened again: Universal Credit to Upstart moved from 55.7 points in July 2026 to 75.0 points in October 2026, and OneMain Financial to Universal Credit widened every month from 41.2 points to 57.1 points.

Each monthly run begins with 800 prompt-surface observations (596 unique questions in July 2026 and 612 in October 2026) across the benchmark's defined AI/search surface universe. Of those, 800 mentioned a tracked brand or competitor in July 2026 and 800 in October 2026; 762 were relevant and 38 irrelevant in July 2026, while 780 were relevant and 20 irrelevant in October 2026. The public metrics use the 689 qualified observations in July 2026 and the 692 in October 2026 that survive both qualification stages. August 2026 produced 708 qualified observations and September 2026 produced 707.

Key Findings

AI Response Inconsistency Alerts

Questions This Section Answers

• Where did AI platforms give conflicting answers about bad credit loan eligibility?
• What kinds of lender attributes did AI platforms describe inconsistently?

Nine critical or high-severity factual inconsistencies were detected across seven AI platforms, concentrated in four companies. These are cases where the same platform family gave incompatible answers about the same lender attribute, most often eligibility thresholds and loan amount ranges.

Upstart

Upstart carried five of the nine inconsistencies, four of them in eligibility. On maximum loan amount, when asked "Is there another loan company like Upstart?", Copilot stated that Upstart's loan range is $1,000 to $50,000, citing LendEDU, Finder, and Swipe Solutions, while Gemini stated the maximum is $75,000, citing CreditNinja and Financer. Both figures appeared on source pages flagged in the data, one carrying a "$1K-$50K Loan Amounts" excerpt and another stating loans run "from $1,000 up to $75,000."

A second amount conflict arose on the question "Where can I loan money asap?" Gemini described Upstart as offering "fast online personal loans in the $1,000-$35,000 range" while Google AI Overviews described "short-term relief loans between $200 and $2,500." Both answers drew on upstart.com source pages alongside other lender sites.

Upstart's minimum credit score drew three separate conflicts. On "Which personal loan is the easiest to get?", Gemini said Upstart does not rely solely on a hard credit minimum and uses alternative data such as education and employment history, citing WalletHub and MoneyLion, while Copilot stated a minimum credit score of approximately 300, citing CNBC, LendEDU, and MoneyLion. On "What credit score is needed for a medical loan?", Gemini stated Upstart sets no minimum credit score requirement at all, while Copilot stated a minimum of 560, citing WalletHub, Finance Loan Advisor, and LendEDU. On "What is the best direct lender for bad credit?", Copilot stated a credit score range of 580 or above, citing Capital Lending News, Business Insider, and Debt.com, while ChatGPT reported a minimum as low as 300, citing Forbes Advisor, Credible, and the Wall Street Journal. Three of those source pages were flagged, including one carrying a "MIN CREDIT SCORE 300" excerpt and another stating lenders "accept credit scores as low as 580."

Upgrade

Upgrade carried two inconsistencies, both on minimum credit score. When asked "What loan can I get with a fair credit score?", Copilot stated a minimum credit score of 620, citing NerdWallet, U.S. News, and Bankrate, while Google AI Mode stated 580, citing Citi and Reprise Financial. Flagged sources on the second side describe fair credit as falling between 580 and 669. When asked "Can I get a $5000 personal loan with a 600 credit score?", Copilot stated a minimum score of 600, citing Teller, WalletHub, and Global Loan Advisor, while Google AI Mode stated Upgrade accepts scores starting at 580, citing Bankrate, Experian, and Investopedia. One flagged source listed both figures in the same excerpt, showing "Upgrade 22%-35% 580+ Accessible."

Avant

Avant carried one inconsistency, on minimum credit score. When asked "What is the easiest company to get a personal loan?", Gemini stated Avant's minimum is around 580, citing LendingTree, MoneyLion, and avant.com, while Google AI Overviews stated it is as low as 550, citing CNBC and Forbes. A flagged CNBC source contained both numbers in the same excerpt, describing approval for borrowers "with scores as low as 550" while also noting "Avant has a minimum credit score requirement of 580."

OneMain Financial

OneMain Financial carried one inconsistency, on maximum loan amount. When asked "big loans for bad credit", Google AI Overviews stated "secured personal loans up to $30,000" while Google AI Mode stated a maximum loan amount of $20,000. Both answers drew on overlapping source sets including Axos Bank, CNBC, and Bankrate.

Benchmark Context

Questions This Section Answers

• How many prompt observations survive qualification into the public benchmark set?
• What benchmark-level metrics changed between July 2026 and October 2026?

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

Benchmark-Level Metrics

The following metrics summarize how the qualified dataset performed at the category level across the reporting window.

The intermediate months matter here. Recommendation-shaped answers fell to 54.5% in August 2026 and rose to 57.7% in September 2026 before the October 2026 jump to 73.0%. The valid shortlist share peaked at 81.1% in August 2026, dipped to 75.1% in September 2026, and then reached 92.2% in October 2026. The category is answering with structured recommendation lists far more often than it was in July 2026.

AI Recommendation Trend

Questions This Section Answers

• Who leads in AI recommendation coverage for bad credit loans, and by how much?
• Which brands posted significant gains or declines against the July 2026 baseline?

A Wider Leader and a Rising Second Tier

Upstart leads and extended its lead in October 2026, but the more commercially consequential shift is that Avant, OneMain Financial, and Upgrade all posted significant baseline gains, narrowing the space between themselves while the gap to the leader grew. The field below the top four remains far behind.

Several brands moved beyond normal month-to-month variation in this window: Upstart, Avant, OneMain Financial, and Upgrade on the upside, and Achieve and Universal Credit on the downside. The category-level result was not one large shift but the combination of four significant upward movements and two significant declines.

What Changed This Month

Questions This Section Answers

• Did Upstart's coverage gain come from more first-place answers or more top-three placements?
• Why did Universal Credit's coverage fall even though its placement rate held steady?
• How did Avant and Upgrade differ in where their placement gains occurred?

Upstart: The Leader Widened

Upstart's valid recommendation coverage rose from 75.0% in July 2026 to 88.9% in October 2026, a gain of 13.9 points, with 13.5 points of it arriving between September 2026 and October 2026. That prior-month jump is the largest single-month move recorded by any brand in the series.

Placement hardened alongside coverage. Upstart's top-three rate rose from 56.6% in July 2026 to 76.2% in October 2026, a 19.6-point increase, while its rank-one rate moved from 42.4% to 44.6%. Raw mention presence stayed near ceiling, moving from 93.8% to 96.1%. Its valid recommendation count rose from 517 in July 2026 to 615 in October 2026.

The distinction worth noting is that Upstart's growth came primarily from moving into more top-three positions rather than from claiming more first-place answers, and its presence rate was already close to saturation. That combination limits how much further coverage can expand without displacing others.

Highest-priority diagnostic: Identify which specific prompt sets and AI surface families account for the 13.5-point prior-month gain, and whether the increase is concentrated in one surface family.

Avant: Sharpest Second-Tier Gain

Avant's coverage rose from 68.5% in July 2026 to 78.0% in October 2026, up 9.5 points, with the largest share of that arriving as a 12.0-point prior-month jump from 66.0% in September 2026.

Placement improved more than presence. Avant's top-three rate rose from 47.2% in July 2026 to 61.0% in October 2026, and its rank-one rate rose from 7.5% to 15.8%. Its raw mention presence rate actually slipped slightly from 86.2% to 84.7%. Valid recommendation count rose from 472 to 540.

The distinction is that Avant is being surfaced at similar rates but placed considerably higher within the answers it appears in. That pattern points to answer composition rather than answer inclusion.

Highest-priority diagnostic: Determine whether Avant's rank-one increase is concentrated in the same prompts where it previously appeared in lower positions, or in newly entered prompts.

OneMain Financial: Coverage Up, Rank-One Flat

OneMain Financial's coverage rose from 60.5% in July 2026 to 71.0% in October 2026, up 10.5 points, including a 9.6-point prior-month gain from 61.4% in September 2026.

Its top-three rate rose from 32.7% to 46.4%, a 13.7-point increase. Its rank-one rate moved from 1.5% to 3.0%, which did not clear the significance threshold. Raw mention presence was essentially flat at 75.8% to 76.4%, and valid recommendation count rose from 417 to 491.

The distinction is that OneMain Financial converted presence into mid-list recommendations without comparable growth in first-place recommendations. The bank appears frequently and is recommended often, but rarely as the single leading answer.

Highest-priority diagnostic: Examine which competitor occupies rank one in the prompts where OneMain Financial reaches the top three but not the top spot.

Upgrade: Rank-One Rate More Than Doubled

Upgrade's coverage rose from 60.4% in July 2026 to 70.2% in October 2026, up 9.8 points, including a 5.8-point prior-month gain.

Placement gains were concentrated at the top of the answer rather than in broader inclusion. Upgrade's top-three rate rose from 28.5% in July 2026 to 43.5% in October 2026, and its rank-one rate more than doubled, from 6.5% to 13.6%. Raw mention presence held essentially flat, moving from 77.1% to 76.5%. Valid recommendation count rose from 416 to 486.

The distinction is that Upgrade's growth is concentrated in first-place recommendations rather than in appearing more often, a different pattern from OneMain Financial's mid-list gains.

Highest-priority diagnostic: Determine which prompts drive Upgrade's 13.6% rank-one rate and whether that concentration sits on a small number of surfaces or is spread across the prompt set.

Universal Credit: Significant Two-Month Decline

Universal Credit's coverage fell from 19.3% in July 2026 to 13.9% in October 2026, down 5.4 points, with a 2.5-point prior-month decline from 16.4% in September 2026. Its decline is now two consecutive months.

Presence collapsed more than coverage. Raw mention presence fell from 24.7% in July 2026 to 15.5% in October 2026, an 9.2-point drop, while the top-three rate fell only from 5.2% to 4.3%. Valid recommendation count fell from 133 to 96, and rank-one recommendations fell from 5 to 2.

The distinction is that the brand is being mentioned far less often while continuing to place similarly when mentioned. That is a visibility problem at the retrieval stage, not a ranking problem within answers.

Highest-priority diagnostic: Identify which prompt categories Universal Credit stopped appearing in between July 2026 and October 2026, and which brands absorbed that presence.

Achieve: Three-Month Decline

Achieve's coverage fell from 10.3% in July 2026 to 6.1% in October 2026, down 4.2 points, across three consecutive declining months. Its prior-month decline from September 2026 was 1.0 point, which did not meet the significance threshold on its own.

Presence fell in parallel: raw mention presence dropped from 14.2% in July 2026 to 10.0% in October 2026, and the top-three rate fell from 2.2% to 1.0%. Valid recommendation count fell from 71 to 42. Rank-one recommendations were flat at 0.4%.

The distinction is that Achieve's decline combines a smaller presence base with a weaker placement rate, which compounds. The brand's absolute counts are moderate rather than tiny, so the direction is meaningful even though the base is small relative to the category leaders.

Highest-priority diagnostic: Determine whether Achieve's three-month decline reflects a consistent surface-level pattern or a scatter of small prompt losses.

Prosper: Quiet Series Gainer

Prosper's coverage rose from 13.4% in July 2026 to 17.1% in October 2026, up 3.7 points, but did not clear the significance threshold. Its prior-month gain was 1.3 points. Raw mention presence was flat at 21.2% to 21.4%, top-three rate moved from 1.7% to 1.9%, and valid recommendation count rose from 92 to 118.

The distinction is that Prosper's improvement sits just below the threshold that would mark it as significant, and it comes with no meaningful change in presence or rank-one standing. Its rank-one rate remained at 0.3%.

Highest-priority diagnostic: Track whether Prosper's slow upward drift continues into the next period or plateaus below significance.

Buyer-Intent Interpretation

Questions This Section Answers

• Which buyer-intent clusters actually appeared in the qualified prompt set?
• What can the benchmark not yet answer about pricing or head-to-head comparisons?

In October 2026, all 692 qualified observations fell into the Brand Recommendation cluster, as they did in July 2026, August 2026, and September 2026. The Pricing & Value and Multi-Brand Comparison clusters remained empty across the full four-month window. The public benchmark therefore measures which brand AI systems recommend for bad credit loans, but it cannot yet answer commercial questions about price positioning, value comparisons, or head-to-head differentiation. Brands competing on rate or fee advantages will need company-level analysis to understand how AI systems discuss those attributes.

Brand Opportunity Summary

Questions This Section Answers

• Which diagnostic question matters most for each brand in October 2026?
• Which brands' coverage figures rest on small absolute counts and should be read with caution?

The benchmark identifies where attention is warranted based on coverage and movement; a company-level analysis is needed to explain why these patterns are occurring.

Evidence Behind the Benchmark

The aggregate metrics are built from prompt-level observations, including the query, surface, recommendation outcome, rank, sentiment, and citations where exposed. Company-level analysis can go deeper into prompt, competitor, surface, and evidence patterns. Source presence is not automatically treated as proof of causation.

About This Benchmark

This report is part of CiteWorks Studio's AI Visibility Industry Market 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: Freedom Debt Relief (13 valid recommendations in October 2026) and National Debt Relief (26 valid recommendations in October 2026) have coverage figures built on small absolute counts; their movements should be read with caution.
• Qualified denominator: All brand-level percentages are calculated against the 692 qualified observations in October 2026, not the 800 raw prompt surfaces collected.
• 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 coverage percentages sit the questions that matter for competitive strategy: which high-intent prompts a brand is winning, which competitor takes the recommendation when a brand loses, what attributes AI systems associate with each option, and which external sources shape those answers. The public benchmark tracks the outcome; it does not reveal the mechanism.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy. The public percentage cannot identify the prompts, competitors, or sources causing the result. This turns the benchmark's directional signals into actionable intelligence for a single brand's position in AI-driven discovery.

Request an AI visibility audit