How templates took over America's financial complaint database — an analysis of 18.2 million CFPB complaints and 3.85 million consumer narratives, 2011–2026, with an original open dataset.
On 14 January 2025, the Consumer Financial Protection Bureau received 46 complaints filed under one particular issue code: "Other transaction problem." On 15 January it received almost 4,000. On 17 January, more than 13,000.
Almost all of them were about two companies: Block, Inc., which runs Cash App, and Early Warning Services, LLC, the company behind Zelle. By the end of January, the CFPB had received 42,521 complaints under that single issue. In a typical month of 2024, it received between 200 and 300.
Somewhere in that flood was a complaint from a consumer in New York, filed on 18 January. It was short and written in the consumer's own words: money had been taken from the account several times; each time, the bank and Zelle said it was the consumer's fault and could not help; in the end, the account was closed "with no warning or no reasoning." Early Warning Services closed the complaint with an explanation. Read the full case →
That complaint stood out because it was rare. When we read through the narratives filed against Block that week, 61.6% were word-for-word copies of another narrative. Against Early Warning Services, 55.9% were. The single most repeated Cash App complaint appeared 9,865 times, identical down to the last character.
Consumer Reports has reported that the surge followed social-media videos falsely telling viewers that anyone who filed a CFPB complaint would be paid from the agency's settlements with Zelle and Cash App. [1] The CFPB had in fact ordered Block on 16 January 2025 to pay up to $175 million over fraud and customer-service failures at Cash App — but the redress was for consumers identified under the order, not for anyone who filed a complaint. [2]
That week is not the story. It is the doorway into it. When we measured the same thing across every one of the 3.85 million narratives the CFPB published between 2015 and August 2026, we found that copied text is not an exception confined to one viral week. In the parts of the database that have grown fastest, it has become the norm.
The CFPB's Consumer Complaint Database is one of the largest public records of consumer experiences with financial companies anywhere. Consumers file complaints about credit cards, bank accounts, mortgages, credit reports, debt collectors and more; the CFPB sends them to the company, and the company responds. Since 2015, consumers could also choose to publish a written account — the "narrative" — with personal details removed.
Three things are essential to read the rest of this investigation correctly:
One more change shapes everything that follows. On 14 August 2026, the CFPB stopped publishing narratives in its public database, saying they are unverified, one-sided and not representative. [3] Previously published narratives were moved to an archive in the agency's FOIA Reading Room. This investigation is based on that archive, joined to the full database by complaint ID.
From December 2011 to 3 October 2026, the public database recorded 18,167,713 complaints. For most of its history it grew steadily: 168,272 complaints in 2015, 277,239 in 2019. Then it accelerated: 800,245 in 2022, 1.29 million in 2023, 2.73 million in 2024 and 5.44 million in 2025.
Almost all of that growth has a single source. In 2015, credit reporting — complaints about the information credit bureaus hold on consumers — made up 20.4% of all complaints. By 2025 it was 88.4%, and in 2026 so far 91.1%. The three nationwide credit bureaus — TransUnion, Equifax and Experian — account for 78% of every complaint in the database's history.
Because credit reporting is now so large, any figure calculated on the database as a whole mostly describes credit reporting. Throughout this investigation, we report it separately.
To measure copied text, we checked every narrative at three levels: identical text; identical text once capitalisation, numbers, punctuation and the CFPB's redaction marks are removed; and near-identical text with small edits, using a standard similarity technique (MinHash). We then separated two very different situations that both produce repeated text:
We count a text as a shared template only when it appears in at least two different locations (ZIP code, or state when the ZIP is masked). Two different people in the same ZIP code would count as one location, so these figures are a lower bound.
In 2015, 1.0% of narratives were shared templates. In 2019, 8.9%. In 2022, 27.4%. In 2024, 49.3%, and in 2025, 50.7%. Add the narratives that repeat a single person's own text, and only 29% of the narratives published in 2024 and 2025 were written once, by one person, for one complaint.
The pattern is concentrated. In 2025, 60.0% of credit reporting narratives were shared templates, and only 14% of credit reporting narratives in 2024–2025 were original. In debt collection, 27.9% were shared templates, and 40% matched another narrative at all. In mortgages, under 1% did.
The largest single template we found appeared in 34,309 narratives in 2024, filed from 2,588 different locations, almost all against the three credit bureaus. It opens by invoking the Fair Credit Reporting Act and listing accounts that, it says, violate the consumer's "federally protected" rights. Other widely repeated texts cite sections of federal law at length, and some rely on legal-sounding theories such as "estoppel by silence" — the claim that a company's failure to answer a letter legally settles the dispute in the consumer's favour.
Sometimes the copying is visible. One complaint filed against Equifax in April 2025 is a formal letter "against [ Credit Bureau Name ]" — the template's placeholder, never filled in. Another, filed against TransUnion in October 2025, is addressed from beginning to end to a different company. During the January 2025 surge, many narratives that were not exact copies turned out to be the same text retyped by hand, with words dropped and sentences broken off — suggesting it was copied out rather than pasted.
None of this means the complaints are false. A person with a genuine problem can use a template, and template wording often describes real rights. What we measure is standardisation — how much of the record is written by the people filing it.
Templates are sold and shared because people believe they get results. The data suggests they do — but not the results most people need.
We compared original complaints and shared templates about exactly the same issue, filed between 2022 and 2025. When the goal is money, templates almost never get it:
| Issue | Monetary relief — original | Monetary relief — shared template |
|---|---|---|
| Bank accounts: problem caused by funds being low | 18.6% | 0.3% |
| Bank accounts: managing an account | 13.5% | 0.2% |
| Money transfers: other transaction problem | 4.6% | 0.0% |
But when the goal is a correction — removing or changing an item on a credit report, or stopping collection of a disputed debt — templated complaints end in non-monetary relief more often:
| Issue | Non-monetary relief — original | Non-monetary relief — shared template |
|---|---|---|
| Credit reporting: incorrect information on your report | 35% | 51% |
| Debt collection: attempts to collect debt not owed | 17% | 39% |
| Debt collection: written notification about debt | 15% | 46% |
That is consistent with what credit repair templates are designed to do: dispute entries on a report. It is a correlation, not proof that the template causes the outcome — the companies involved and the people who use templates may differ in other ways. But the gap holds issue by issue, which makes a simple explanation less likely.
If you want your money back, describe your own problem in your own words, with dates, amounts and documents.The practical lesson of the data
The complaints that recovered money in our sample of cases did exactly that. A Pennsylvania consumer documented, step by step, how a $3,400 fraud refund promised by Bank of America never arrived. Read the case → An Illinois consumer explained how a check U.S. Bank itself issued to pay a mortgage was stolen, altered and cashed twice. Read the case → A Delaware consumer laid out the dates of every deposit required for a $600 account-opening bonus Citizens Bank had not paid. Read the case → All three were closed with monetary relief.
Set credit reporting aside and the rest of the database tells a different story. From 2015 to 2020, complaints about everything else were stable, at roughly 135,000 to 160,000 a year. In 2025 they reached about 633,000 — around four times the 2020 level.
Debt collection is the biggest driver. Complaints rose from 68,000 in 2023 to 156,000 in 2024 and 283,000 in 2025. The fastest-growing issue was "attempts to collect debt not owed," which rose from 36,000 to 116,000 complaints in two years. Over the same period, mortgages — 28% of non-credit-reporting complaints in the database's early years — fell to around 5%.
A typical case reads like this: a Texas consumer explained that they closed a credit card in 2018 after losing it, yet seven years later Encore Capital Group was still pursuing the account, and it was still damaging their credit. The complaint was closed with an explanation. Read the case →
One feature of the data needs further investigation: in 2024 and 2025, the three credit bureaus appear among the companies with the most debt collection complaints — about 70,000 in 2025 — even though they are not debt collectors. These look like complaints about collection accounts shown on credit reports, filed under the debt collection category. We flag it here and in the dataset rather than reclassify them ourselves.
Across the whole database, the share of complaints closed with monetary relief fell from about 6% in 2015 to 0.4% in 2026. That number is real, but misleading: credit bureaus almost never pay money, and they now receive 91% of complaints. Most of the fall is a change in the mix of complaints.
Remove credit reporting, though, and a genuine decline remains. Monetary relief held steady at around 7.5–8% of non-credit-reporting complaints from 2012 to 2023, then fell to 6.2% in 2024 and 4.0% in 2025. Excluding the January 2025 surge — in which almost every complaint was closed without relief — the 2025 figure is 4.2%. Within individual products the trend is the same: bank accounts fell from 19.7% (2015) to about 13%; credit and prepaid cards from 21.7% to 15.7%; student loans from 5.3% to 0.2%.
Because company responses are self-reported, the data cannot tell us whether consumers are actually receiving less, or whether companies are recording outcomes differently. It shows what companies told the CFPB.
Comparing companies fairly is harder than it looks. A bank that handles many credit card disputes will look different from one that mostly handles checking accounts, and a company hit by a wave of templates will look worse than it may be. So we compared each company with the industry within the same product, using only complaints that are not known templates, filed from 25 August 2023 onwards.
Even then, the differences are large. For bank account complaints, where the industry closed 13.2% with monetary relief, Bank of America recorded 38.3%, Citibank 24.0%, Wells Fargo 13.2% and Capital One 3.1%. For credit cards (industry: 15.0%), Bank of America recorded 35.1%, Citibank 29.5%, American Express 24.6% and Capital One 12.2%. For money transfers (industry: 6.3%), Coinbase recorded 24.9%, Bank of America 20.7%, PayPal 15.1%, JPMorgan Chase 3.7% — and Block, across more than 28,000 complaints that were not templates, recorded none.
Two cautions apply. These are the outcomes companies reported: a company that refunds a customer but records the complaint as "closed with explanation" will appear to give nothing. And roles differ: Early Warning Services runs the Zelle network, but refunds are usually made by the customer's own bank, so its near-zero rate is expected.
Raw complaint counts mostly reflect size. To compare banks by how often their customers complain, we divided each bank's bank-account complaints by the number of deposit accounts of $250,000 or less it reported to the FDIC — the closest public measure of retail customer accounts, although it also includes some small-business accounts. [4]
Among the largest banks in 2025, Citibank received 22.9 bank-account complaints per 100,000 accounts; Capital One 17.7; Wells Fargo 13.0; Bank of America 8.3; JPMorgan Chase 7.4; and U.S. Bank 4.3.
The way you measure matters. Dividing all complaints by deposits made Capital One look more than five times worse than its peers — but most of its complaints are about credit cards, which generate complaints without generating deposits. Measured per customer account, it ranks fourth. We publish all three measures in the dataset so readers can see the difference for themselves.
Every year since the database began, companies have responded to more than 99.5% of complaints within the CFPB's timeliness window. That figure measures speed, not resolution. A complaint can be answered on time and still be closed without any relief — as almost every complaint in the January 2025 surge was.
The headline figure also hides gaps. Between 2024 and 2026, 7,524 student loan complaints were filed against "Servicer under contract with Federal Student Aid" — a label available to borrowers who don't know which company services their federal loan. Every one of those complaints is recorded as untimely. The label is not a company, and the data does not show what happened to those complaints afterwards; but in the public record, they have no response at all.
Narratives were never the whole database: only about 21% of complaints have one. In 2025, 19% of credit reporting complaints had a published narrative, compared with 77% of money transfer complaints. In 2026 the share fell sharply: from 4.1% of complaints in January to 1.5% in July, before the CFPB's announcement. (Narratives are published with a delay and the archive ends on 14 August 2026, so the most recent weeks are incomplete: the single narrative for August's 643,950 complaints reflects that cut-off, not a trend.)
The CFPB's reasons for stopping — that narratives are unverified, one-sided and unrepresentative — are partly borne out by this investigation: a large share of recent narratives were not individual accounts at all. But narratives were also the only place where the details lived. Without them, the January 2025 surge would have looked like 42,000 new victims of Zelle and Cash App, not a wave of copied text. Template detection like ours is only possible with the text. Researchers can still use the archive; future complaints will not have that record.
These cases are drawn from narratives consumers wrote in their own words, linked to the CFPB's public record by complaint ID. Each is an allegation by the consumer; we report how the company says it closed the complaint, not whether the allegation was true. Full details are on Kibbo's Real Cases section.
Filed during the January 2025 surge, this short complaint describes repeated losses, being told it was the consumer's fault, and an account closed without explanation.
Bank of America agreed that withdrawals were fraud and promised a refund check. The consumer documented repeated branch visits as the check failed to arrive.
A mortgage payment check issued by U.S. Bank was stolen, altered and deposited three times; two deposits cleared. After two years, the consumer filed a third attempt.
A servicemember asked Ford Motor Credit to apply a state servicemembers' protection law to a loan's interest rate and was refused by its legal department.
The consumer closed a lost card in 2018 and says collection calls and credit damage continued for seven years.
Police reports filed and creditors notified, yet fraudulent accounts remained on the consumer's credit report — an original complaint in a category dominated by templates.
The consumer met every deposit condition of a mailed offer, was told repeatedly the bonus was coming, and documented each step.
Data. The full CFPB Consumer Complaint Database (18,167,713 complaints received 1 December 2011 to 3 October 2026) and the CFPB Narratives Archive (3,851,415 narratives, 19 March 2015 to 14 August 2026), both downloaded on 3 October 2026 and joined by complaint ID (99.9% match in a random test).
Product families. The 21 product names used by the CFPB were mapped to 11 families that are comparable across the 2017 and 2023 category changes. Issues were compared only within the same category period.
Template detection. Exact duplicates; normalised duplicates (lowercase, without redaction marks, numbers or punctuation); and near-duplicates using MinHash with locality-sensitive hashing (word 5-shingles, 128 permutations, Jaccard similarity ≥ 0.8; narratives under 30 words excluded from this level). Near-duplicate clusters were built within each year, and within each half of 2025. A hand check of 100 near-duplicate pairs confirmed near-identical text; a check of 50 pairs just below the threshold found most were also the same template, so our figures are conservative.
Shared templates. A text counts as a shared template when it appears in at least two distinct locations (published ZIP code, or state where the ZIP is missing or masked). Texts from a single location are classified as one person filing repeatedly. A hand check of 30 shared-template pairs found 3 where the ZIP codes shared their first three digits and might belong to the same person.
Company comparisons use complaints filed from 25 August 2023 onwards that are not known templates, compared with the all-industry figure for the same product family. Bank rates use FDIC Call Report data: deposit accounts of $250,000 or less (field DEPSMB) at each year end.
Monetary relief rates exclude complaints still in progress. 2026 is incomplete.
Every decision, check and correction is recorded in the project's research log, published with the dataset. Errors found after publication will be logged on Kibbo's Corrections Log.
All 18.2 million complaints with harmonised product families, template classification for every narrative, company business types, complaints per 100,000 residents by state, and bank complaint rates — plus a data dictionary, the full code and SHA-256 checksums. Free to use under CC BY 4.0.
Download the full dataset (Zenodo, DOI) →Summary tables, README and data dictionary →
See all datasets in Kibbo's Consumer Data Library →
Kibbo Research →