AI in Healthcare Compliance: What It Actually Does for SNFs Right Now

What "AI in Healthcare Compliance" Actually Covers TodayCopy link

Search that phrase — or "AI for healthcare compliance," "healthcare AI compliance," any variant — and you’ll get a lot of "AI will transform healthcare compliance" think-pieces. Here’s what it actually means on the ground, in tools facilities are using right now:

• Policy and regulation tracking — surfacing a rule change (a new QSO memo, a bill signed into law) the day it happens, instead of waiting for the next in-service to catch up.
• Survey and audit prep — flagging gaps against F-tags and State Operations Manual chapters before a surveyor does.
• Documentation integrity checks — catching missing signatures, stale care plans, or inconsistent MDS coding before they become a citation.
• Decision support (not decision-making) — surfacing relevant policy or precedent for a human to weigh, never issuing the final call itself.

That last distinction is the one regulators keep coming back to, and it’s the throughline for the rest of this post.

Surveys Just Got Less Predictable — and LongerCopy link

If your facility’s survey playbook still assumes a predictable weekday window, it’s out of date. On January 30, 2026, CMS issued QSO-26-03-NH, revising State Operations Manual Chapters 5 (complaints) and 7 (survey and enforcement) — with a follow-up clarification memo in April. Survey teams must now stay onsite at least five consecutive hours on Day 1, and at least 10% of standard surveys must start off-hours — early morning, evening, or weekend — with at least half of those landing on a Saturday or Sunday.

Translation: "we know when surveyors usually show up" is no longer a safe assumption. A compliance program that only gets reviewed before an expected survey window is already behind.

Forget Headcount — Paperwork Is What Gets CitedCopy link

The 2024 federal minimum-staffing mandate — 3.48 nurse hours per resident day, a 24/7 onsite RN — is dead. Two federal courts vacated it in 2025 (Texas, April 7; Iowa, June), Congress separately blocked HHS from enforcing it until 2034, and CMS made the repeal official via an interim final rule effective February 2, 2026.

That closes off a storyline that dominated LTC compliance conversation for two years — but it doesn’t close off the actual exposure. F-tag citations, QAPI currency, and survey findings are driven by process and paperwork, not headcount alone. Facilities with staffing above the old proposed minimum still rack up repeat citations when policies lag or documentation doesn’t hold up. The staffing-mandate fight is over; the paperwork fight never stopped.

The MDS Validation Audit ClockCopy link

Since early 2026, CMS has been running data-validation audits on MDS assessments tied to the Quality Reporting Program and Value-Based Purchasing — up to 1,500 SNFs a year, roughly 1 in 10 of the country’s ~15,300 certified nursing homes. Selected facilities get notified through CMS’s iQIES portal and have 45 days to produce complete, defensible resident documentation. Miss it, and the facility eats a 2% cut to its FY2027 annual payment update.

Forty-five days to produce a clean record, for a random 10% of facilities, with real money on the line — that’s not an abstract "AI could help with compliance" pitch. It’s a specific, dated reason a facility’s documentation needs to already be in order, not scrambled together after a notice arrives.

Where Regulators Draw the Line: AI Can Inform, Never DecideCopy link

This is the part that matters most, and it’s not just a philosophy Relic Care’s Compliance Assistant has chosen to build around — it’s the line California and federal regulators are independently drawing too.

California’s SB 1120, effective January 1, 2025, bars health plans from letting AI or algorithmic tools deny, delay, or modify care based on medical necessity — that determination has to come from a licensed physician or qualified clinical professional, reviewing the patient’s own record. Worth being precise here: SB 1120 governs health plan and insurer utilization review, not a SNF’s own compliance software directly. But it’s the clearest statutory statement in California of exactly the principle a facility’s compliance program should already be operating on.

The federal government is converging on the same line from a different angle. In its first major Medicare Advantage compliance guidance update since 1999, OIG flagged sole reliance on AI — an algorithm that doesn’t account for a member’s individual circumstances — as a compliance risk, echoing a 2024 CMS warning against the same pattern. That guidance is Medicare Advantage-specific, not written for nursing facilities. But the posture it demands — audit your AI-assisted decisions, keep a human accountable for the final call — is exactly the posture a SNF compliance program should already hold, regardless of which regulator happens to be looking at which line of business.

Two different regulators, two different contexts, one converging answer: AI can surface an answer to check against your own judgment. It can’t replace that judgment.

When AI Drafts Something a Resident’s Family Will ReadCopy link

If your facility uses generative AI to help draft a written, chat, audio, or video communication about a resident’s clinical information, California’s AB 3030 already applies to you. Effective January 1, 2025, it requires a disclaimer — plus instructions for reaching a human provider — on any AI-generated patient communication about clinical information. The law covers "health facilities," a term that already includes skilled nursing facilities under Health & Safety Code Section 1250.

There’s a built-in exception worth knowing: the disclosure requirement doesn’t apply if a licensed or certified human provider reads and reviews the AI-generated communication before it goes out. Review it first, and the disclaimer requirement disappears — which is one more concrete reason "AI drafts it, a human reviews it" is the operating model to build around, not just a nice-to-have.

Know Your Tool’s Actual ObligationsCopy link

Not every AI-adjacent rule applies to every AI tool, and getting that wrong cuts both ways — overclaiming exposure wastes compliance staff’s time, underclaiming it misses a real obligation. ONC’s HTI-1 rule is the first federal rule requiring AI transparency disclosures — source data, development approach, risk-management practices — but that obligation runs to developers of certified health IT modules, not to every standalone compliance or documentation tool a facility might use.

That’s not a loophole to relax into. It’s a floor other regulators are clearly building above — the direction of travel is toward more disclosure and more accountability for AI-assisted decisions, not less. A compliance program is better served building toward that transparency now than waiting for a rule to force it.

Healthcare AI Compliance Isn’t Only About Note-WritingCopy link

Most "AI in healthcare compliance" content — including some of what’s already out there on clinical documentation — focuses on charting accuracy: is the note complete, does it support the billing code, does it hold up if a surveyor pulls the chart. That matters. But it’s not the only place compliance programs get caught out.

A recurring, distinct pattern in facility survey histories: citations that have nothing to do with note-writing quality at all — care planning gaps, RN-on-duty coverage documentation, infection-control program currency, QAPI program gaps — that keep recurring even when the clinical charting itself is fine. It’s a program-and-policy-currency problem, not a documentation-accuracy problem: the policy on the shelf hasn’t caught up to the rule that changed, or the program review that was supposed to happen quarterly didn’t.

That’s the gap a compliance-currency tool is built to close — not "does this note read well," but "does the facility’s program actually reflect the rule as it stands today." None of the general compliance-AI vendors in this space go deep on that SNF-specific, program-level distinction; most either stay at the level of "AI raises compliance considerations" in the abstract, or focus on a different buyer (enterprise GRC, data security, clinical training) entirely. A facility administrator or DON needs something built for the SNF-specific version of this problem, not a horizontal tool retrofitted to it.

A note on scope: This post describes AI-in-healthcare-compliance regulation and CMS guidance as of the dates cited above; it isn’t legal advice, and it doesn’t cover every state or every AI use case. If a specific law or guidance document applies to your facility’s situation, confirm the current text and talk to your own counsel or compliance officer before acting on it.