
Cycle time, review passes per return, and recurring-comment rate are three numbers most firms can already pull from their workflow. Here's exactly how each one gets faked the moment it becomes a target instead of a diagnostic.

The IRS publishes exactly how long a return takes to prepare, broken down by activity, for every return type it processes. Review isn't one of those categories — here's the honest range that exists, and what would have to be measured to get a real one.

Review by exception means most of a return gets no second look at all — only what's flagged does. When that scope is defensible under Circular 230, and when it's a diligence failure wearing an efficiency label.

A reviewer's note on a return isn't deleted when the season ends — it's archived. What actually happens to it, and why archived turns out to be a different thing from findable.

Four structural reasons a tax firm's review knowledge disappears — and why turnover, the one everyone blames first, explains the least of them.

Why the judgment behind a tax return outlives no ledger entry, why the pipeline that used to replace it is shrinking, and what a firm can do structurally before the next departure takes it — the hub for Pernee's Institutional Knowledge cluster.

In June the IRS told tax preparers to credit clients for whatever AI actually saves them, through lower fees. In September the AICPA formally asked the IRS to clarify that instruction — arguing it ignores what AI costs the firm before any hour is saved.

Form 1099-DA started reporting gross proceeds on digital-asset sales for the first time this filing season — and, per Treasury's own phased timeline, without cost basis. A reviewer who treats it like a 1099-B is one step from signing off on the wrong gain.

On February 9, 2026, the U.S. Tax Court found that three of the four cases cited in a taxpayer's brief don't exist — invented, the judge concluded, by a large language model. It didn't sanction anyone. This is what nobody checked first, and what already requires someone to.

Cycle time, review passes per return, and recurring-comment rate are three numbers most firms can already pull from their workflow. Here's exactly how each one gets faked the moment it becomes a target instead of a diagnostic.

Form 1099-DA started reporting gross proceeds on digital-asset sales for the first time this filing season — and, per Treasury's own phased timeline, without cost basis. A reviewer who treats it like a 1099-B is one step from signing off on the wrong gain.

Why the judgment behind a tax return outlives no ledger entry, why the pipeline that used to replace it is shrinking, and what a firm can do structurally before the next departure takes it — the hub for Pernee's Institutional Knowledge cluster.

The IRS publishes exactly how long a return takes to prepare, broken down by activity, for every return type it processes. Review isn't one of those categories — here's the honest range that exists, and what would have to be measured to get a real one.

On February 9, 2026, the U.S. Tax Court found that three of the four cases cited in a taxpayer's brief don't exist — invented, the judge concluded, by a large language model. It didn't sanction anyone. This is what nobody checked first, and what already requires someone to.

Review by exception means most of a return gets no second look at all — only what's flagged does. When that scope is defensible under Circular 230, and when it's a diligence failure wearing an efficiency label.

OPR Alert 2026-19 maps a practitioner's existing Circular 230 duties onto AI-assisted tax work — due diligence, competence, honest billing, client confidentiality. It never asks the practitioner to tell the client AI was involved.

Four of the five tax return review checks test this year's return. The fifth tests whether it agrees with last year's — and a carryforward that was wrong once stays wrong until someone actually looks.

Four kinds of errors that keep reaching review at CPA firms — carryforward mismatches, unauthorized elections, source disagreement, and disclosure gaps — and the structural reason each one survives preparation instead of getting caught there.