Bottom line: If your ophthalmology practice received a prior authorization denial you suspect was algorithm-driven, don’t start by writing a general appeal letter. Start by (1) identifying the exact denial reason and whether the payer disclosed a reviewer’s name and specialty, (2) requesting peer-to-peer review within the payer’s short window (often 5–10 business days), and (3) building a written appeal that answers the specific criterion cited — not the diagnosis in general. A template for that letter is below.

This article is general educational information for physicians and practice staff, not legal, financial, or insurance advice. Prior authorization rules, appeal deadlines, and payer clinical criteria vary by plan, state, and payer, and change frequently. Verify current requirements directly with the payer and, where relevant, your legal counsel before making appeal or billing decisions.

Prior authorization has always been a heavy administrative load for ophthalmology and retina practices, largely because of high-volume, high-cost anti-VEGF injections. What changed in 2025–2026 is that payers now routinely use automated systems somewhere in the review pipeline — and physicians increasingly can’t tell, from the denial letter alone, whether a human clinician meaningfully reviewed the case or whether an algorithm did the work with a rubber-stamp signature attached. This guide explains what’s actually true about AI and prior authorization in 2026, how to tell what kind of denial you’re dealing with, and gives you a specialty-specific appeal letter template built around the denial patterns that actually occur in ophthalmology.

What Counts as an “AI Prior Authorization Denial” in Ophthalmology

Not every PA denial touched by software is an “AI denial” in the sense most physicians mean. It helps to separate three roles AI can play:

1. Intake and triage — reading the submitted documentation, flagging missing elements (for example, a required OCT or fluorescein angiography result before an anti-VEGF injection), and prompting the submitter before a denial ever issues. Payer-reported data suggests this use has measurably reduced missing-information denials in some programs.
2. Risk scoring — flagging which requests are likely to be denied or need additional review, used to route cases to human reviewers.
3. Determination drafting or issuance — generating the actual approve/deny decision and denial-reason language, sometimes with minimal or templated human sign-off.

The controversy — and the reason state legislatures and CMS have gotten involved — centers on the third category. Congressional and litigation-disclosed evidence has described payer systems capable of processing dozens of denials per minute with only seconds of nominal human review per case, which critics argue does not constitute meaningful clinical judgment. In response, industry groups have publicly committed that algorithms alone are not used to deny requests involving medical necessity or clinical judgment, and several states now require a licensed clinician — in some cases a same-specialty clinician — to review any adverse determination before it is finalized. For federal Medicare programs specifically, current guidance describes a “human-in-the-loop” requirement: AI may assist, but a qualified clinician must make or ratify the final coverage decision.

The practical problem for your practice: none of this is easy to verify from the outside. A denial letter that arrives within hours of submission, uses templated language that doesn’t reference your specific chart findings, and does not name a reviewing clinician is the closest thing to an observable signal — and it’s exactly what your appeal strategy should target.

The 2026 Rules Ophthalmology Practices Need to Know

CMS-0057-F — and the drug-PA exclusion that matters most for anti-VEGF

The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F), finalized in January 2024, reached its main operational compliance date on January 1, 2026. For “impacted payers” — Medicare Advantage organizations, state Medicaid and CHIP fee-for-service programs, Medicaid managed care plans, and ACA marketplace (QHP) issuers on the federal exchange — the rule now requires:

• A decision within 72 hours for expedited/urgent requests and 7 calendar days for standard requests (down from a prior 14-day standard for many plans). Note: this specific turnaround requirement does not yet apply to QHP issuers on the federal exchange — it’s the one payer category CMS gave a later timeline to as part of the 2026 rollout, so confirm plan type before citing this deadline.
• A specific, member/provider-facing denial reason for any non-drug item or service, regardless of whether the denial is communicated by portal, fax, mail, or phone.
• Public reporting of PA approval, denial, appeal, and overturn metrics, with the first reporting deadline of March 31, 2026.

Here is the detail that most competing guides miss and that matters enormously to ophthalmology: these process requirements explicitly do not apply to prior authorization decisions for drugs. Anti-VEGF agents — Eylea, Lucentis, Vabysmo, Beovu, and Avastin/bevacizumab — are physician-administered drugs billed under the medical benefit. That means a retina practice cannot assume the 72-hour/7-day clock or the specific-denial-reason mandate automatically covers its single highest-volume PA category. Whether faster turnaround and reason-transparency apply to a given anti-VEGF denial depends on that specific payer’s own policy — not on CMS-0057-F.

This gap is not permanent. On April 10, 2026, CMS released a follow-up proposed rule — CMS-0062-P — that would extend interoperability, transparency, and decision-timeframe requirements to drug prior authorizations covered under both the medical and pharmacy benefit. As of this writing (mid-2026) it remains a proposed rule, not final, with most provisions slated to take effect October 1, 2027 if finalized. Until it’s finalized, treat the drug-PA exclusion in CMS-0057-F as still fully in force, but check CMS’s site periodically, since this is the regulatory area most likely to change the guidance in this article.

Before citing CMS-0057-F in an appeal for a drug-based denial, confirm the payer’s own published PA policy actually incorporates the rule’s protections; some payers apply them voluntarily to drugs, others do not.

The WISeR model — and why ophthalmology isn’t (yet) in scope

Separately, CMS’s Innovation Center launched the Wasteful and Inappropriate Service Reduction (WISeR) model on January 1, 2026 — a six-year pilot testing AI-assisted prior authorization in Traditional Medicare (which has historically had little to no PA) for a defined list of procedures, in six states: New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington. CMS states that no request under WISeR is denied before review by an appropriately licensed clinician, and that vendor performance is measured in part on accuracy and timeliness, not on denial volume alone. That said, physicians should know the underlying payment mechanics: WISeR vendors are compensated on a contingency basis — a share of the Medicare savings generated when a service is not affirmed or paid — rather than a flat administrative fee. Provider groups, including the American Hospital Association, have publicly argued this savings-linked structure functions as a denial incentive regardless of CMS’s framing; CMS disputes that characterization. Ophthalmology-specific services are not currently on the WISeR target list, and the American Academy of Ophthalmology has formally pushed back — jointly with other surgical societies — against expanding the model into ophthalmic and ambulatory-surgical-center procedures. Ophthalmologists in the six pilot states should still confirm current WISeR scope before assuming a given procedure is unaffected, since the target list can change.

State laws on AI claims review — and the ERISA exception

A growing number of states — California (SB 1120), Texas (SB 815), Arizona (HB 2175), and Maryland (HB 820) among the earliest as of 2026 — have enacted laws requiring physician or clinician review of adverse AI-generated coverage determinations, and some require reviewers to hold the same or a similar specialty as the treating physician. These statutes give you concrete language to cite in an appeal when a denial appears templated and unreviewed. The important limitation: ERISA self-funded employer plans are generally exempt from state insurance regulation, including these state AI laws. Before citing a state statute, check the denial letter or plan documents to confirm the patient’s plan is fully insured and state-regulated, not a self-funded ERISA plan — otherwise the citation will be disregarded and it weakens the credibility of the rest of the appeal.

Oversight layer	What it requires	Applies to anti-VEGF/drug PA?	Applies to ERISA self-funded plans?
CMS-0057-F (federal, 2026)	72-hr/7-day decisions (QHPs excepted for now), specific denial reason, public metrics reporting	No — drug PA is excluded from these process mandates	N/A (governs Medicare Advantage, Medicaid, ACA marketplace, not commercial ERISA plans)
CMS-0062-P (proposed, Apr 2026)	Would extend PA interoperability/transparency rules to drugs	Not yet — proposed only, targeted effective date Oct 1, 2027 if finalized	N/A
WISeR model (federal pilot)	AI-assisted PA with mandatory clinician review, 6 states, defined procedure list; vendors paid on savings from non-affirmed services	No — ophthalmology not currently listed	N/A — Traditional Medicare only
State AI-review laws (e.g., CA, TX, AZ, MD)	Physician/clinician (sometimes same-specialty) review of adverse AI determinations	Varies by state statute — check specific text	Generally no — most state insurance laws don’t reach ERISA self-funded plans
Payer’s own clinical policy	Sets actual step-therapy, documentation, and turnaround requirements for a given drug/procedure	Yes — this is usually the controlling document for anti-VEGF PA	Yes

Why Ophthalmology PA Requests Actually Get Denied

Published data on anti-VEGF prior authorization is unusually specific, and it tells a different story than “AI is denying everything.” In one multi-practice study of over 2,000 qualifying anti-VEGF PA requests, the large majority were ultimately approved. The real burden wasn’t denial — it was delay: a substantial share of requests were not approved same-day, a majority caused a delay of 24 hours or more, and a meaningful minority took over a month. Where denials did occur, they concentrated in a few identifiable categories:

• Step therapy (“fail-first”) requirements. The largest share of denials in the published literature traced to a payer’s requirement to document a failed or contraindicated trial of bevacizumab (Avastin) before approving a higher-cost agent like aflibercept (Eylea) or ranibizumab (Lucentis). This is the single most common, and most appealable, ophthalmology PA denial reason.
• Diagnosis-code or modifier mismatches. Anti-VEGF claims require the drug code (e.g., J0178 for aflibercept, billed per 1 mg unit; J2778 for ranibizumab, billed per 0.1 mg unit) to be paired correctly with CPT 67028 and a laterality modifier (RT, LT, or 50), and linked to a covered ICD-10-CM diagnosis (e.g., H35.32x for exudative/wet AMD, diabetic retinopathy codes with macular edema for DME, H34.81x/H34.83x for CRVO/BRVO). A denial coded as “non-covered diagnosis” is frequently a linkage error, not a true coverage dispute.
• Non-covered or off-label indications. Because bevacizumab is used off-label in ophthalmology, payable diagnoses for it are payer-policy-dependent and not uniform; a denial here often needs literature citation and community-standard-of-care documentation rather than a procedural argument.
• “No authorization on file” or lapsed authorization. These are largely unrecoverable through appeal because most payers do not allow retroactive prior authorization — prevention (verifying authorization status before every injection cycle) is the only real fix.

Note: exact percentages for the study cited above vary by publication and practice sample — confirm current figures against the source study before quoting hard numbers publicly.

Is This Denial Algorithmic? Five Signals to Check

You usually cannot get definitive proof that an algorithm — rather than a human — issued a given denial, but these patterns are the closest available signal, and they’re what your appeal should lean on:

1. Turnaround speed inconsistent with complexity. A same-day or sub-24-hour denial on a clinically nuanced case (e.g., a treatment-resistant DME patient being switched agents) is a flag.
2. Templated language that mirrors your diagnosis code, not your chart. If the denial reason reads like a form letter and never references your specific clinical findings, treatment history, or imaging results, it suggests limited individualized review.
3. No named reviewer, or a reviewer outside ophthalmology/retina. You are generally entitled to request the name, credentials, and specialty of whoever reviewed the case. Physician survey data has repeatedly found that most physicians who go through peer-to-peer review doubt the payer’s reviewer had appropriate specialty qualifications — so this check is worth making routinely, not just when you suspect AI involvement.
4. Denial reason doesn’t match your documentation at all — for example, a step-therapy denial issued when a documented bevacizumab trial and failure are already in the chart you submitted.
5. Reauthorization denied for a drug the patient has tolerated well for months, with no clinical change noted — reauthorizations are usually near-automatic approvals in the literature, so an unexplained denial here is worth escalating quickly.

None of these signals alone proves algorithmic denial without meaningful review — but each is a legitimate basis to explicitly request clinician-level, same-specialty peer review in your appeal, which is the single most effective lever available regardless of whether AI was actually involved.

The Appeal Process, Step by Step

1. Triage the denial reason before doing anything else. Read the letter and classify the denial: administrative (eligibility, missing modifier, coding error) versus clinical (medical necessity, step therapy, non-covered diagnosis). Administrative denials usually need a corrected resubmission, not a full appeal — treating them the same way wastes your appeal window on an easy fix.
2. Record the deadline immediately. Appeal windows are not uniform. Federal rules set a minimum 60-day appeal window for Medicare Advantage, Medicaid managed care, and ACA marketplace standard reviews; commercial plan windows can range from roughly 30 to 180 days depending on the insurer and state. The deadline printed on your specific denial letter controls — confirm it before drafting anything.
3. Request peer-to-peer review. This is a direct call between the treating ophthalmologist and the payer’s medical director, and it is generally the fastest path to reversal because it converts an impersonal determination into a physician-to-physician clinical conversation. Most payers require the request within 5–10 business days of the denial — often a shorter window than the formal appeal deadline, so don’t wait. Explicitly ask whether the original reviewer was board-certified in ophthalmology or retina, and request a same-specialty reviewer for the call if not.
4. Build the formal written appeal to the specific criterion cited. Don’t write a general “this patient needs treatment” letter. Pull the payer’s own clinical policy bulletin or step-therapy criteria for the drug/procedure in question (you’re entitled to request the specific criteria applied to your case), and address each element directly with your documentation.
5. Submit, track, and escalate. Log the submission method, date, and confirmation number. If the payer misses its own required turnaround time under an applicable rule, some states treat the authorization as automatically approved (“deemed approved”) — document any timeline violation and be prepared to proceed under that status where your state and plan type allow it.

The Appeal Letter Template

A strong physician appeal letter for an algorithmic or suspected-algorithmic ophthalmology PA denial does five things: cites the specific denial reason verbatim, addresses that reason point-by-point with clinical evidence, explicitly requests same-specialty clinician review, cites applicable regulatory protections only where you’ve confirmed they apply, and states a clear, time-bound ask.

[Practice Letterhead]
[Date]

[Payer Name]
Attn: Prior Authorization Appeals
[Address from denial letter]

RE: Appeal of Prior Authorization Denial
Patient: [Name] | Member ID: [ID] | DOB: [DOB]
Date of Denial: [Date] | Denial/Reference Number: [Number]
Requested Service: [e.g., Intravitreal aflibercept injection, CPT 67028 / J0178]
Diagnosis: [ICD-10 code and description, e.g., H35.3211 – Exudative AMD, right eye, with active choroidal neovascularization]

Dear Medical Director,

I am writing to formally appeal the denial of prior authorization for [service/drug] for the
above patient, denied on [date] for the stated reason: "[quote the exact denial language]."

Clinical summary: [2–4 sentences: diagnosis, disease severity/imaging findings, prior
treatment history, and why the requested service is clinically indicated now.]

Response to the stated denial reason:
[If step therapy: document the specific bevacizumab trial dates, dosing, and documented
lack of response or contraindication, with reference to chart notes/imaging.]
[If "not medically necessary": cite the specific clinical finding(s) that meet the payer's
own published criteria for this drug/procedure, and reference the criteria document by name
if available.]
[If diagnosis/coding mismatch: confirm the correct ICD-10, CPT, and modifier combination and
note the discrepancy in the original submission, if applicable.]

Request for clinician-level review: Given the clinical complexity of this case, I request
that this appeal be reviewed by a physician board-certified in ophthalmology or retina, and
I ask that the reviewer's name and credentials be provided in the written determination.
[If applicable and confirmed to apply to this plan type: If this determination was made
using an automated or algorithmic decision-support system, I request confirmation of the
specific criteria applied and the credentials of the clinician who made the final
determination, consistent with [cite specific, confirmed-applicable state statute or plan
document provision — do not cite unless verified for this plan type].]

Time-sensitivity: [If applicable] Delay in treatment carries a documented risk of
irreversible vision loss in this diagnosis; I am requesting expedited review consistent with
the urgency of this case.

I am available for a peer-to-peer discussion at [phone number] and can supply additional
records including [OCT, fluorescein angiography, treatment log, etc.] upon request.

Sincerely,
[Physician Name, Credentials]
[NPI] | [Practice Name] | [Direct Phone/Fax]

Enclosures: [Clinical notes, imaging reports, prior treatment log, payer's own clinical
policy citation if referenced above]

How to adapt this template: For cataract surgery, YAG capsulotomy, or other surgical PA denials, replace the step-therapy paragraph with documentation against the payer’s specific surgical medical-necessity criteria (e.g., visual acuity thresholds, functional impairment, or glare testing results the payer’s own policy requires), and keep the clinician-review and time-sensitivity language largely as-is.

Denial Type vs. Appeal Strategy at a Glance

Denial reason	Likely cause	Best first move	Typical strength on appeal
Step therapy / “fail bevacizumab first”	Payer policy requires documented Avastin trial before Eylea/Lucentis/Vabysmo	Submit trial dates, dosing, and documented lack of response or contraindication	Strong if trial is documented; moderate if arguing against step therapy in principle
“Not medically necessary”	Payer criteria not clearly matched in the original submission	Pull payer’s own clinical policy bulletin; match documentation to each listed criterion	Strong when payer’s own criteria are objectively met
Non-covered/off-label diagnosis (bevacizumab)	Payer-specific payable-diagnosis list for off-label use	Cite community standard of practice and relevant clinical literature	Moderate — payer-policy-dependent
Diagnosis/CPT/modifier mismatch	Coding error, not true clinical dispute	Correct and resubmit; this is often not a true “appeal” case	Very strong once corrected
No authorization on file / lapsed	Missed pre-service authorization step	Confirm payer’s retroactive-authorization policy; usually not appealable	Weak — largely unrecoverable
Reauthorization denied with no clinical change	Possible algorithmic/administrative error	Request peer-to-peer immediately; flag as likely error given established treatment tolerance	Strong — reauthorizations are approved at very high rates historically

Common Mistakes That Sink an Appeal

• Writing a general clinical-necessity letter instead of rebutting the exact denial reason. Payers respond to specificity; a letter that doesn’t quote and address the stated reason reads as generic and is easier to deny again.
• Missing the peer-to-peer window because staff focused only on the longer written-appeal deadline. Track both clocks from day one.
• Assuming CMS-0057-F protections apply to a drug-based denial without checking the payer’s own policy — see the exclusion discussed above.
• Citing a state AI-oversight statute without first confirming the plan is state-regulated and not ERISA self-funded — this undermines the letter’s credibility.
• Treating a coding/modifier denial as a clinical dispute, which wastes the appeal window on a case that should have been a same-day corrected resubmission instead.
• Letting an AI drafting tool fabricate or misstate a clinical guideline citation. Every citation in a submitted appeal should be verified against the actual source before it goes out under a physician’s name.

Edge Cases and Exceptions

• Emergent or urgent situations (e.g., an acutely progressing retinal detachment or a rapidly worsening exudative process) generally qualify for expedited review timelines under both federal and most state/plan rules; explicitly label the request and appeal as urgent and cite the clinical basis for urgency.
• Medicaid managed care and dual-eligible patients may have additional state-specific protections and shorter mandated timelines than commercial plans — check state Medicaid guidance in addition to the payer’s own policy.
• Retroactive authorization is the narrowest exception across nearly all payers; if a service was rendered without authorization due to a true emergency, document the emergent nature clearly and immediately, since standard appeal arguments about medical necessity will not overcome a retroactive-authorization denial on their own.
• ERISA self-funded plans follow federal ERISA claims-procedure rules rather than state insurance law; appeals should reference the plan’s own summary plan description and ERISA claims-procedure regulations instead of state AI statutes.

Where AI Legitimately Helps the Appeal — and Where It Can Hurt You

AI-assisted appeal-drafting tools (used by the practice, not the payer) can meaningfully cut the time to assemble a documentation-heavy appeal — pulling relevant chart notes, matching them to a payer’s stated criteria, and producing a first-draft letter in minutes rather than the many hours of staff time historically reported per anti-VEGF PA request in the literature. That is a legitimate, useful application. The risk is treating the AI-drafted letter as final: any tool can misstate a clinical guideline, cite an inapplicable regulation, or misquote the payer’s own policy. A physician or qualified staff member should verify every factual and regulatory claim in an AI-drafted appeal before it is submitted, exactly as this article recommends verifying every figure and rule cited here against current payer and regulatory sources.

FAQ

Does CMS-0057-F guarantee my anti-VEGF prior authorization will be decided within 7 days? Not necessarily. CMS-0057-F’s 72-hour/7-day turnaround and specific-denial-reason requirements apply to non-drug items and services for impacted payers; the rule explicitly excludes prior authorization for drugs, which is how most anti-VEGF agents are categorized. A separate proposed rule (CMS-0062-P) would extend similar protections to drug PA starting as early as October 1, 2027, but it is not yet final. Check the specific payer’s own policy for the actual applicable timeline.

Can I demand to know if AI was used in my denial? You can generally request the name, credentials, and specialty of the clinician who reviewed and issued the denial, and in some states you can request the specific criteria an automated system applied. Whether the payer must disclose that AI was involved at all varies by state and payer; industry commitments lean toward disclosure, but this is not yet a uniform legal requirement everywhere.

Is it worth appealing if the historical overturn rate looks high? Historical data generally supports appealing — a large majority of appealed Medicare Advantage denials have been overturned in recent years — but that figure describes appeals across all specialties and denial types, not a guarantee for any individual ophthalmology case. Base your decision on whether your documentation actually addresses the payer’s stated criterion, not on the base rate alone.

What if the denial letter doesn’t name a reviewer at all? Request that information in writing as a first step, separate from your substantive appeal — you are generally entitled to it, and the absence of a named, credentialed reviewer is itself useful evidence if you later need to argue the review was inadequate.

---

Author/reviewer note: this article should be reviewed by a supervising ophthalmologist or certified coding specialist (e.g., COE/CPC) before publication, with regulatory and coding details verified against current CMS, AAO, and AAPC primary sources. Prior authorization policy — federal, state, and payer-specific — changes frequently; confirm all figures and deadlines directly with the relevant payer before relying on them in a live appeal.

Sources referenced: CMS-0057-F Fact Sheet · CMS WISeR Model overview · California SB 1120 (“Physicians Make Decisions Act”)

Last verified: August 2026.