Your Credentialing System Thinks It's Current. Our Data Says Otherwise.

The Claim Arrives Clean. The Provider Isn't There.

A billing coordinator at a Texas clinic submits a claim in late January. The practice has been paneled with this payer for three years. Everything looks right on the credentialing side: the last attestation was recent, the file is complete, and the provider's NPI resolves without a problem. The claim comes back denied. Reason: the provider is not listed as in-network at the location on the claim.

She checks the payer directory. The provider is listed, but at their old address, the office they left fourteen months ago. Nobody flagged it. Nobody was watching.

This is not an unusual story. It's the default story, and the numbers behind it are larger than most credentialing teams realize.

What We Found Watching 1.8 Million Records

As of September 5, 2026, Argoseer is holding 1,122,451 distinct disagreements between payer directories and provider records across the practices we monitor. To be precise about what that means: a distinct disagreement counts each conflict once, on its logical key. Those same disagreements occupy 1,787,834 stored rows, because a finding gets stored once per practice row the NPI hangs off. The second number reflects storage, not additional problems. (Source: Argoseer pipeline, 2026-09-05.)

The breakdown is worth sitting with. The single largest category: a listed detail differs from the provider record, 878,568 disagreements. Second: listed at the wrong location, 670,403. Third: the practice accepts the plan but the directory does not show it, 141,931 disagreements. That third category carries direct compliance exposure under the No Surprises Act (Consolidated Appropriations Act, 2021), which requires health plans to apply updates within two business days of receiving new information and to verify all directory data at least every 90 days. A provider whose plan acceptance isn't reflected in the directory isn't just an inconvenience. It is a federal compliance failure waiting for a regulator to find it. (See American Optometric Association / CMS provider directory memo, 2022.)

These numbers are our numbers. The external literature corroborates the pattern. A 2023 JAMA Network Open study examined ~450,000 physician listings across five major national payers and found that only 19.4% had consistent address AND specialty information across directories, per TechTarget's coverage of the study. CMS's own national review, cited by Ideon in their 2026 compliance guide, found 48.74% of Medicare Advantage provider locations contained at least one inaccuracy. The dominant failure types CMS identified map directly to our top two categories: detail errors and wrong-location listings.

Confidence, Honestly Stated

Here's where I want to be careful about what our numbers actually mean.

Take the "provider missing from the payer directory" category. We have 909,700 single indicative signals pointing to providers who appear to be missing. We also have 7,974 cases where two independent detectors agreed, which is our higher confidence tier. (Source: Argoseer pipeline, 2026-09-05.) The confirmed number is the one I'd put in front of a compliance officer. The raw signal count is the size of the universe we're still sorting through.

The gap between those two numbers isn't evidence of sloppiness. It's evidence of how hard this problem actually is. A single signal that a provider is missing from a directory might reflect a processing lag, a directory pull that ran before an update propagated, or a genuine gap. Two detectors agreeing is a different thing. Both are useful, and they're useful for different purposes: the raw count tells you where to look, the confirmed count tells you what to act on.

This distinction matters because the industry has a habit of publishing large, alarming numbers that don't explain their denominator. I'd rather show the gap than pretend it isn't there.

The Billing Coordinator's Problem, In Full

Back to that Texas clinic. Let's trace what actually happened.

The provider completed a CAQH re-attestation in October 2025. CAQH's cycle, per analysis by credyapp.com citing CAQH standards, requires re-attestation every 120 days, with more than 2.5 million active profiles feeding over 1,000 payers. The update was logged. The provider's primary practice location changed in November 2025. The practice notified the payer. The payer logged the update.

Then one payer's directory didn't reflect it. Maybe the update hit a processing queue. Maybe the directory pull ran on a stale cache. By December the directory showed the old address. By January, the claim denied.

The No Surprises Act required the payer to apply that update within two business days of receiving it. Whether it did or didn't is genuinely hard to know from the practice's side. What the practice does know is that the denied claim cost staff hours to work, and the revenue took 47 days to recover.

A 2024 study in Health Affairs Scholar found that 44.8% of previously flagged provider listings (2,316 of 5,170 surveyed) still showed at least one inaccuracy at follow-up. A separate study published in American Journal of Managed Care (PMC) followed 1,802 inaccurate listings for 403 to 574 days (mean 541 days) and found that 40.3% were still listed inaccurately; only 13.3% had been corrected. Eighteen months. That's how long some of these errors sit.

The Texas billing coordinator didn't have a process problem. She had a data freshness problem that no credentialing workflow built around 120-day attestation cycles was designed to catch.

The Regulatory Clock Is Running Faster Than the Attestation Cycle

Four things happened in roughly the same window, and the timing is not coincidental.

The No Surprises Act set a 90-day verification cadence and a two-business-day update window for commercial plans starting January 2022. CMS's rule for Medicare Advantage plans (effective January 1, 2026) requires attestation to directory accuracy annually and updates within 30 days of a known change, per Neolytix. NCQA's 2026 standards, analyzed by BluePeak Advisors, require directory accuracy assessments every six months and updates within 30 days, and the standards documentation explicitly describes the 2026 requirements as demanding "operational and technical redesign, not just policy updates."

Verifiable's August 2026 analysis put it plainly: four overlapping regulatory developments are "converging on the same infrastructure gap: the systems and processes most health plans use to manage provider data were not built to do what these changes now require."

None of this indicts any specific credentialing platform. Argoseer is additive to those stacks, not a replacement for them. Your credentialing system tracks what you filed. We watch whether it's still true after you filed it. Those are different problems.

The AI Disclosure Problem Nobody Is Watching

This part surprised me, honestly.

In a scan of Texas practice websites conducted February 27 through March 6, 2026, our detectors found 495 practice websites carrying AI diagnostic tools with no practitioner-review statement, 14 named chatbot vendors with no AI disclosure at all, and 12 disclosures present but hidden with CSS so they would never actually render in a browser. (Source: Argoseer pipeline, 2026-03-06. These are positive detections: the tool was found on the page.)

The ONC HTI-1 Final Rule (89 FR 1192, 45 CFR § 170.315(b)(11)), effective January 1, 2025, requires 31 source-attribute disclosures for any "Predictive DSI" tool used in certified health-IT contexts, covering intended use, training data, validation, and known risks, per Live Compliance's August 2026 analysis. The rule governs certified health-IT developers, not practice websites directly, which means a straight mapping of our website-disclosure findings to this specific statutory obligation is complicated. I won't pretend otherwise.

But the 495 tools with no practitioner-review statement is a real number on its own. Whatever the exact regulatory hook, a practice that deploys an AI diagnostic tool without disclosing it to patients is carrying reputational and liability exposure that most credentialing checklists don't even have a field for.

Why Manual Audits Can't Close This Gap

The industry spends an estimated $4 billion annually on provider data accuracy, per Ideon's 2026 network management analysis. Error rates remain persistently high. The Atlas Systems 2026 data accuracy report found that 58% of health plan members have encountered a directory error at least once, and 80% of those members say it made them trust their plan less.

The math on manual auditing is simple and unfavorable. If 81% of directory entries have errors, and the average fix takes staff time measured in hours, and errors persist for a mean of 541 days without automated correction, the backlog is not something a quarterly audit cycle addresses. The 2025 NAMSS conference, per QGenda's reporting, centered on "technology as a force multiplier," with the profession explicitly naming AI-assisted tools integrated with payer databases as the operational direction.

HealthStream's 2026 credentialing trends report noted that organizations using credentialing platforms with built-in AI features report higher satisfaction with credentialing quality than those still evaluating. But adoption remains uneven, which means most practices are not yet capturing that benefit.

The disagreement I'd name plainly: Atlas Systems cites a 20% inaccuracy rate as one industry-wide figure, while the JAMA Network Open study found 81% of entries had some form of error across five major national payers. Our own data, 1,122,451 distinct disagreements across 1.8 million monitored records, sits between those two figures in implied rate but is measuring something different: disagreements between a specific provider record and what a directory actually shows, not a population-level audit. What would settle the comparison is a shared denominator, which the industry doesn't have yet.

Who Is Watching the Data Between Attestations?

The structural problem is this: CAQH updates every 120 days. The No Surprises Act expects directories current within 2 business days of a change. The provider data that feeds a payer directory is not a point-in-time fact. It changes, and the changes don't announce themselves.

An address update filed in November might not propagate to every payer directory until February. A specialty change submitted through one credentialing pathway might not reach a secondary payer for three re-attestation cycles. A new plan acceptance that the practice knows about might not appear in the public directory for weeks. None of these are processing errors in isolation. Together, at scale, they add up to 1,122,451 distinct disagreements that someone has to find.

The question isn't whether to automate this. The question is what you're doing with the output when the tool surfaces a finding. That workflow, what happens after the flag, is where most of the real work sits. Argoseer can show you what the data says. Resolving it still requires the credentialing coordinator, the payer contact, the updated attestation. The tool doesn't replace that. It just means you're not discovering the problem from a denied claim.

What I don't know is how many of those 1,122,451 disagreements are currently sitting in someone's credentialing file as verified, current, and accurate. My honest guess is most of them.

The real question for any practice manager reading this: when your credentialing system last said everything was fine, how long ago was the data it was checking? If you want to see what Argoseer is flagging for practices like yours, the monitor page is a reasonable place to start.