98941 CPT Code: How AI Downcoding Is Quietly Reducing Chiropractic Reimbursement

You see a long-term patient with cervical, thoracic, and lumbar complaints. You assess each region, identify subluxations in all three, perform CMT in each, and document the patient's response. By the CPT code definition, that's a 98941 — three spinal regions adjusted, moderate complexity, properly billed.

Two weeks later, your EOB shows the claim was paid. You move on. But when you compare line items at the end of the month, you see the 98941 was paid at the 98940 rate — about $15-25 less than billed, depending on the payer. No denial. No documentation request. Just a quiet reduction.

This is AI downcoding, and in 2026 it became one of the largest silent revenue leaks in independent chiropractic. Multiply that $15-25 across hundreds of CMT visits per month, and the math gets serious fast.

The 98940 / 98941 / 98942 family

Chiropractic CMT coding is unusual in healthcare in that it's region-counted. The five spinal regions for CMT coding purposes are cervical, thoracic, lumbar, sacral, and pelvic. Three CPT codes correspond to how many regions were adjusted:

The downcoding patterns AI tools target are predictable: 98942 → 98941, and 98941 → 98940. Both involve the system reducing the region count by one tier — which is the easiest justification an AI auditor can produce because it's hard to disprove without explicit region-by-region documentation.

Why chiropractic is hit harder than most specialties

Three structural factors make chiropractic an unusually frequent AI downcoding target.

The "documented region" standard is fuzzy. Unlike E/M coding, where 99214 vs 99213 is decided by medical decision-making criteria with relatively clear thresholds, chiropractic region counting depends entirely on what's written in the SOAP note. AI tools read the note and count distinct region-specific findings. If the note says "adjusted C, T, and L" without separate clinical findings for each, the AI may treat it as one or two regions rather than three.

Medicare's coverage rules created the audit template. Medicare covers chiropractic only for manual manipulation to correct subluxation, and requires documentation of the specific subluxation (with PART exam findings: Pain, Asymmetry/misalignment, Range of motion abnormality, Tissue/tone changes). Commercial AI tools picked up this standard and now apply it to non-Medicare claims, even where the underlying contract doesn't require PART documentation explicitly.

High visit frequency increases pattern detection sensitivity. Chiropractic patients often come in 2-3 times per week during a treatment plan. AI tools flag unusual frequency patterns, and a practice billing 98941 on most visits gets flagged faster than one billing a mix.

How to detect AI downcoding in your own practice

The data lives in your ERAs. Compare what you billed to what was paid, line by line.

The 30-minute audit

1. Pull your last 90 days of ERAs for one payer at a time. Most chiropractic-friendly EHRs (ChiroTouch, Genesis, ChiroSpring) export this as a CSV.
2. Filter to CMT lines (98940, 98941, 98942).
3. Compare the "billed CPT" to the "paid CPT" field. Count any row where they differ.
4. Calculate your downcoding rate per payer: percentage of 98941s billed that were paid as 98940. Above 10% is a pattern. Above 25% is systematic.

Run this for your top 3-5 commercial payers and your Medicare Advantage plans separately. You'll often find one or two payers driving most of the loss. That gives you targets — both for batch appeals and for renegotiation.

How to document 98941 to survive AI review

The fastest documentation upgrade most chiropractic practices can make: structure your SOAP note to identify each region treated as a separate line, with a distinct clinical finding, a specific adjustment, and a response to treatment. Don't write a single grouped statement.

Instead of:

"Adjusted C, T, L. Patient tolerated well."

Write:

"Cervical: C3-C4 right rotation restriction with PART findings, diversified technique applied, post-adjust ROM improved 15°. Thoracic: T6 right rotation with palpable hypertonicity in upper trapezius, prone drop technique, hypertonicity reduced. Lumbar: L4-L5 left lateral flexion restriction with paraspinal asymmetry, side-posture diversified, restoration of segmental motion noted."

This isn't longer. It's just structured to survive AI parsing. Most AI claim reviewers do exactly enough natural language processing to identify region-specific clinical findings — give them three, and 98941 sticks.

How to appeal a downcoded 98941

Most payers have a 60-180 day appeal window. Three things make appeals win:

• Documentation that maps explicitly to each region. If the note already has region-by-region findings, the appeal is straightforward. If not, the appeal is much weaker — which is why the documentation upgrade above matters more than appeal mechanics.
• Reference to the payer's published medical policy. Most major payers publish a chiropractic CMT policy. Quote it. Make them prove their AI followed it.
• Pattern documentation in the cover letter. Don't appeal as if this is a one-off. State the systematic count: "We've identified [N] downcoded 98941 claims to [Payer] over the past [period], totaling $[amount]."

Single-claim appeals are rarely worth the time on their own. Batch appeals — submitting 20-50 downcoded claims at once with a systematic-pattern cover letter — carry meaningful leverage because they signal regulatory complaint risk to the payer.

What's coming in 2026

The legislative response is moving but unevenly. Several state bills proposed in 2026 would require payers to disclose AI involvement in claim adjudication, and some would create a private right of action for systematic downcoding. None have become law yet. What's more likely in the next 6-12 months:

• State insurance commissioner investigations after enough provider complaints accumulate. Maryland and California have hinted at this.
• Class action litigation from physician groups and billing companies. Multiple cases were filed in 2025-2026 against major payers.
• Quiet payer policy revisions. AI thresholds get retuned without announcement when pressure builds. The only reliable detection is monitoring your reimbursement patterns over time.

None of this helps a chiropractic practice getting downcoded today. What helps today is detection, region-by-region documentation, and batch appeals on systematic patterns.

The flip side worth knowing about

AI downcoding by payers is the mirror image of AI upcoding by health systems — where hospitals use AI tools to push claims toward higher-paying codes. Same technology, opposite direction of force. Independent chiropractic practices sit between the two pressures: they don't have the AI tools that large systems use to push codes up, but they're absorbing the AI tools commercial payers use to push codes down. The only available defense is regular monitoring of your own reimbursement data — claim-by-claim, payer-by-payer.

The practices that catch this in 2026 are the ones treating ERA data as something to read, not just file.