The Doctor Was Real.

The Directory Wasn’t.

One psychiatrist, twenty-nine directory listings, a federal complaint alleges. An inspector general’s federal sample, a state attorney general’s phone survey, a $2.5 million settlement, and a new federal statute now describe the same class of failure. The doctors were never the accusation. The record of them was.

By William Douglas Baker

Founding Editor

In a search described in the complaint filed against EmblemHealth — “New York, NY” as the location, a 200-mile radius — more than 1,000 directory listings referred to only 50 actual providers, the plaintiffs allege. One entry became its own exhibit. In the directory, the complaint states, “the same psychiatrist is listed twenty-nine times.” Then the sentence that names the arithmetic: “Simply put, this one psychiatrist generates twenty-nine results, instead of one.” The complaint was filed December 30, 2025, in the U.S. District Court for the Southern District of New York by the American Psychiatric Association and the New York State Psychiatric Association, on their own behalf and for their members, joined by six plan members. Case No. 1:25-cv-10783.

The complaint gives the duplication a purpose: it inflates, “making it appear that EmblemHealth’s network of mental health providers is vastly larger than it actually is.” The scale allegation is specific: of approximately 154,700 search results across six New York cities, the complaint counts only 12,550 actual providers — roughly 90 percent of the results, it alleges, were duplicate entries. The industry has a name for the outcome — a ghost network, a directory that overstates available in-network care through inaccurate or unusable listings. And the complaint runs the harm in both directions. Patients dial through pages of entries that resolve to no appointment. The clinicians behind the entries, it alleges, are “injured due to a loss of control of their identities, harm to their reputations, and erosion of trust in the patient populations that they serve.” The doctors in this story are not accused of anything. They are the ones being multiplied.

The numbers say this is not one insurer’s filing cabinet. In October 2025, the federal inspector general for Health and Human Services examined encounter data from a ten-county sample of Medicare Advantage and Medicaid managed care plans and reported that, on average, “55 percent of behavioral health providers listed in plans’ networks did not provide a single service to enrollees in 2023” — a Medicare Advantage figure. In one Medicare Advantage plan, only three of 356 listed network providers supplied a service to that plan’s enrollees that year. In a separate survey of inactive providers from the sampled Medicare Advantage and Medicaid plans, OIG estimated from their responses that “72 percent of the inactive providers should not have been listed” as network providers at all. A provider with no recorded service is not the same as a provider with no open appointment; the report investigates the reasons separately. New York’s attorney general had already run the experiment by telephone: staff called 396 listed, in-network mental health providers across thirteen plans, and callers were offered appointments with only 56 of them. The report’s summary line: “86 percent of the listed, in-network mental health providers staff called were ghosts, as they were unreachable, not in-network, or not accepting new patients.” Success rates by plan “ranged from 0 percent to 35 percent.”

Between the credential and the appointment

Note what none of this required: a fake doctor. The directory failures examined here concern how providers are represented to patients seeking care. A boundary belongs on the table before the interpretation does: these sources document directory problems and barriers to care. They do not establish how an AI answer engine ingested, reconciled, or presented any of these entries, and this edition does not claim they do. The question the twenty-nine listings raise survives that boundary, because a directory is a published record, and published records get read — by people, and by whatever tools people consult. Readers of a record never meet the clinician. They meet the entries. What does any reader receive when one person arrives as twenty-nine?

The complaint names what a broken record does to a professional’s standing. “Reputational damage can happen with extraordinary speed in the era of online reviews,” it observes, in a passage about psychiatrists losing control of their own names. Dr. Robert Trestman — “a leading ghost networks expert for the association,” as ProPublica describes him — put the mechanism in one sentence: “What we do is based on trust,” he told the outlet. “So when our name appears in a listing that says you can get care, and then they call us, and we say, ‘Sorry, not taking new patients,’ it has a really negative impact.”

Ninety days, five business days, one monitor

The corrective instruments arrived sixteen days apart, and both impose duties concerning provider-directory accuracy. On February 3, 2026, the REAL Health Providers Act was signed into law as part of the year’s appropriations package. Beginning with plan year 2028, organizations offering the Medicare Advantage plans the provision covers must generally “verify the provider directory information of each provider listed in such directory” at least every 90 days (hospitals and certain other facilities may follow a less frequent schedule authorized by the Secretary, never less than annually), mark unverifiable information as possibly out of date (“may not be up to date”), and “remove a provider from such directory within 5 business days” of determining the provider has left the network. On February 19, Attorney General Letitia James announced the EmblemHealth settlement: outreach to every credentialed behavioral health participating provider “Every ninety (90) days (a ‘Verification Cycle’),” corrections to listing information within two business days of learning of a change, an independent monitor, and restitution for eligible members, with eligibility adjudicated by that monitor — plus the $2.5 million payment. One provision reaches past record-keeping altogether: before giving a member a referral, EmblemHealth must contact the provider and “confirm in real time that they accept new patients and can treat the Member.” The distinction between maintaining a record and confirming usable care is drawn inside the settlement itself. “Health insurers cannot mislead consumers with inaccurate provider directories while families are left without care,” James said. The two instruments establish directory-accuracy obligations with different scopes and timetables. Both address directory accuracy. The settlement also requires measures to improve access to care.

In the Agentics framework — an interpretation this publication brings to the record, not a finding of any instrument above — what these documents exhibit has a clinical name: Signal Fragmentation, a characteristic of Digital Derangement Syndrome™. This edition’s specimen comes from a regulated industry, which is precisely why it is legible: health care has inspectors general, attorneys general, and statutes that force its registries into the public record. These sources establish oversight requirements in the health-plan context. They do not establish what verification or correction obligations apply to professional directories in other industries. No instrument in this edition asks.

Answer Engine Authority™ treats consolidation as the working response to fragmentation: one identity, coherent entries, signals that reconcile instead of multiply. The lesson of the twenty-nine listings is not that directories fail. It is that being listed and being findable are different states — and a directory listing alone does not establish that a patient can obtain an in-network appointment.