Draft ACDIS/AHIMA brief: AI-generated queries must meet the same compliance rules as human queries

An AI- or CAPD-drafted query that reaches a provider would have to clear the same compliance rules as one a coder writes by hand. That is the thrust of a new Section XI in the draft 2026 update to the ACDIS/AHIMA Guidelines for Achieving a Compliant Query Practice, which states that technologically generated or automated queries, including the ones tools call prompts, nudges, advisories, or alerts, are "subject to the same compliance standards as manually constructed queries." The document is a draft: the public comment period closed June 12, 2026, and the organizations expect a final version in summer 2026. It is not in effect yet, so treat what follows as where the standard is heading, not where it already is.

What the draft actually says about AI-generated queries

The core rule is short: if a communication meets the brief's definition of a query, it follows every compliance element in the brief, no matter what generated it or what the platform labels it. The draft defines a query as a communication that presents a provider with a specific diagnosis or documentation option for a specific patient encounter.

Passive decision support that only surfaces reference information, without steering the provider toward a specific answer for a specific patient, is not a query. When in doubt, the draft says, treat it as one.

Section XI names the tools it has in mind: computer-assisted physician documentation (CAPD) running live during an encounter, computer-assisted coding (CAC) flagging documentation gaps after the fact, large language model and generative AI platforms that draft or deliver query language, and EHR-integrated advisories and alerts. All of them fall inside the same standard.

Why this lands on a risk-adjustment coder's desk

Query compliance has historically read as an inpatient CDI concern. This draft widens that. It says the guidance "applies equally across inpatient, outpatient, ambulatory, and professional fee environments," and its definition of a query professional explicitly includes "Hierarchical Condition Category (HCC) coders operating in physician group or risk adjustment settings, and any vendor or technology platform acting in a query-generating capacity on behalf of a covered organization."

That pulls physician-group and risk-adjustment coders, and the query tools they use, squarely inside the same rulebook. A leading query that inflates a RAF score is the exact pattern RADV reviewers and the OIG look for, so a tool that generates one is a compliance exposure, not a shortcut.

The five checks to run on an AI-drafted query

Section XI holds a technology-generated query to the same required elements as any compliant query, and spells them out for technology output. Translated into a pre-flight check a coder or CDI reviewer can run before that query reaches a provider:

A human still owns every query the tool sends

The draft is direct that automation does not move accountability. It states that even when technology assists, humans remain responsible for compliance, and that "organizational process owners retain accountability for every query delivered to a provider, including those generated through automated or AI-assisted processes."

It gives two ways to meet that: review technology-generated queries before they go out, or, if the workflow sends them without pre-delivery review, run structured audits of the tool's query output on a regular basis. It also warns not to assume a vendor-supplied or EHR-integrated tool produces compliant queries by default, and notes that a tool can produce leading or incomplete queries when its configuration or training data introduces bias toward reimbursement over clinical accuracy.

What changed from the 2022 brief

The 2026 draft supersedes the 2022 version. Beyond the new technology section, ACDIS and AHIMA flag substantive updates in three other areas coders should read alongside it.

Where this fits with the rest of the AI-in-coding picture

None of this asks a coder to reject AI query tools. It draws the line where the coder still has to stand: a tool can surface a documentation gap and draft the language, but a human query professional decides whether the query is compliant and answers for it if it is not. That is the same place the GAO landed on AI coding accuracy and where clean risk-adjustment documentation has always lived. The draft is a preview of the bar. The wording can still change before the final version, but the direction is set: a query is a query no matter what drafts it, and a human still answers for it.