Credit Complaints Rose 17.7x Since 2020. Phone Complaints Rose 1.06x.

Everyone reported the explosion. Nobody asked how it arrived.

The headline number is public and it’s been quoted everywhere. The CFPB’s own December 2025 report leads with “an almost 3,000% increase in NCRA complaints since January 1, 2020.”

True. And close to meaningless on its own.

Because here’s what happens when you split those complaints by how they were filed. Credit complaints filed on the web rose 17.7x between 2020 and 2025, from 271,010 to 4,799,271. Complaints filed by telephone over the identical period rose 1.06x, from 5,232 to 5,554.

Not 106%. 1.06 times. 322 more phone calls in five years.

The entire explosion is web. Every additional complaint arrived through a form.

What we measured

The CFPB records how each complaint reached it, in a field called Submitted via. We counted credit reporting complaints by that field, per year, and indexed to 2020 because 2020 is the CFPB’s own baseline in the report we’re extending. Not because it flatters the trend.

Year	Web	Phone	Postal	Web index	Phone index
2019	128,258	4,380	2,774	47	84
2020	271,010	5,232	2,819	100	100
2021	299,608	4,818	889	111	92
2022	589,773	9,421	2,679	218	180
2023	1,033,654	8,891	3,788	381	170
2024	2,352,770	6,354	5,724	868	121
2025	4,799,271	5,554	3,820	1,771	106

We show 2019 through 2025 in full. No window was picked to manufacture a shape: the series includes the year before the surge and every year since.

2026 is excluded from the indexed chart. The file ends 2026-07-16, so it’s about half a year. Web is already at 3,832,014 through mid-July, which is 14.1x the 2020 full year, but plotting half a year against full years would misstate the trend in one direction or the other depending on what the reader assumes.

The channel nobody checked

Phone is an accidental control group, and it’s the most useful thing in this dataset.

Filing by phone means a human being calls the CFPB and talks to a person. You can’t bulk-submit it. You can’t script it. It’s slow and it’s annoying, which is exactly what makes it a clean measurement of one thing: how many people are distressed enough to pick up the phone.

That number has not moved in six years. 5,232 in 2020. 5,554 in 2025. Postal mail, same story, flat.

So run the logic. If a wave of genuine new credit trouble had hit American files after 2020, it should show up in every channel. Distressed people call. They write. They didn’t.

What this rules in. This is strong evidence that the surge is a change in how complaints are submitted, not purely a change in the underlying volume of credit problems. It’s consistent with automation, bulk submission, and third-party filing at scale.

What it does not prove, and we’re not going to let this slide past: it does not prove any particular actor caused it, and it does not prove that no real increase in credit problems happened. Web filing also grew because filing online got easier, better known, and more heavily marketed. Both things can be true. We’re not picking one and dressing it up as a finding.

Sourced rather than speculated, here’s the regulator’s own read. The CFPB attributes complaint growth partly to “third parties using the complaint process, including credit repair organizations and those who offer credit advice online and on social media platforms (so called ‘finfluencers’ or ‘FinTok’)” and to “an increase in the number of complaints submitted in a single session” (CFPB, FCRA 611(e) report, December 2025). That’s their assessment. Our channel data is consistent with it. It isn’t proof of it.

The caveat that belongs on every complaint statistic, including ours

This is the part that costs us, so we’ll say it plainly.

Every other page we publish from this dataset counts complaints. 1,923,820 “not mine” complaints. 383,484 complaints in Georgia. 4,810,498 credit complaints in 2025. All real counts. All from this file.

And every one of them is a count of complaints filed, not of people harmed or errors confirmed. The channel split is the clearest possible evidence of why those came apart after 2020. If a number in this dataset went up 17.7x while the hardest-to-automate way of producing it went up 1.06x, then the number is measuring submission behavior at least as much as it’s measuring credit problems.

So if you’re quoting a CFPB complaint volume, quote this one next to it. Including ours.

What this means for your file

Honestly? It means be skeptical of anyone waving a complaint count at you, and that includes us.

There’s a practical read, though. A complaint is not a fix. Nearly five million credit complaints got filed in 2025 and the top one was “this isn’t mine,” which means a lot of people are using a federal complaint form as their first move.

The thing that actually changes what’s on your report is knowing what’s on it. All three, read line by line, tracked over time. Credit Booster AI handles that reading and tracking on a 7-day free trial. It won’t file anything for you on a form, and it can’t make a bureau agree with you.

Method

Source. CFPB Consumer Complaint Database, bulk CSV, https://files.consumerfinance.gov/ccdb/complaints.csv.zip. Pulled 2026-07-16, upstream Last-Modified: Thu, 16 Jul 2026 09:33:37 GMT. SHA256 9746821f7bfffb09947cab3e1d9acb3400edabb78388c8fbe70ff71f35aaaf48.

N. 16,934,825 complaints, 2011-12-01 to 2026-07-16. Every row read. Nothing sampled.

Channel. The CFPB’s Submitted via field, unmodified.

Filter. Credit reporting products only, counting both product labels the CFPB has used across the series. It renamed the product mid-series, and filtering on a single label silently drops years of data.

Index base. 2020, the CFPB’s own baseline in its December 2025 report.

Omitted from the chart. Fax and email. Fax reached 0 by 2023 and email is approximately 0 throughout. Both are in the raw aggregates.

Monthly figures. Complaints naming the three nationwide bureaus, across all products rather than credit-reporting products alone, went from 12,819 in January 2020 to 459,024 in January 2026 and 671,541 in June 2026. That is a wider filter than the rest of this page uses, and it is labeled here rather than blended into the yearly series. Those are single months, so the partial-2026 problem doesn’t apply to them.