State-by-state tax research compliance requires verifying nexus thresholds, apportionment methods, and IRC conformity positions against each state's current statutes, because generic AI tools frequently return outdated or uncited answers that can't be defended in professional practice.
Your client operates in eight states. Each one sets its own nexus threshold, its own apportionment formula, and its own conformity position to the federal tax code. A general-purpose AI tool will answer confidently, but confidence isn't the same as accuracy, and an uncited answer isn't one you can put in front of a partner.
This guide breaks down where AI tax research tools commonly fail on multi-state compliance, what to check before you rely on a state-level position, and the criteria that separate professional-grade SALT research from expensive guesswork.
The United States has thousands of state and local tax jurisdictions. Every state sets its own statutes and administrative guidance, and many delegate additional taxing authority to counties, cities, and special districts. Some have no corporate income tax, some have no personal income tax, and a handful have no state sales tax at all.
South Dakota v. Wayfair, Inc., 585 U.S. 162 (2018), eliminated the physical presence requirement for sales tax nexus and let states impose economic nexus standards based on revenue or transaction volume instead. States adopted the new standard differently. Some kept a transaction-count threshold, some never adopted one, and a growing number have dropped the transaction count entirely.
According to Avalara's state-tracking summary, as of April 21, 2026, 16 states have eliminated a 200-transaction threshold since Wayfair and moved to revenue-only tests. Kentucky is next: under House Bill 757, enacted April 14, 2026, its transaction-count prong falls away effective August 1, 2026, leaving a $100,000 revenue-only test. The trend is clear, but the specific threshold still has to be checked state by state.
Generic AI tools treat this as a single-source problem. They return a confident answer without tracing it back to the specific state statute, regulation, or department guidance that supports it. For a practitioner building a position under Circular 230 standards, an uncited or outdated fact creates real professional risk.
State tax codes change constantly. A tool trained on older data won't know that Kentucky enacted HB 757 in April 2026, or that Illinois eliminated its 200-transaction threshold effective January 1, 2026. When you ask about current law and get last year's answer, the position is wrong before you finish the memo.
Example: Ask a general AI tool about California's economic nexus threshold and you may get "$100,000 or 200 transactions," which is the old Wayfair safe harbor model, not California's actual statute (Cal. Rev. & Tax. Code § 6203(c)(4) sets a $500,000 revenue-only test with no transaction count).
An answer without a citation isn't something you can hand a partner for review. A defensible SALT position needs to trace back to a specific statute, regulation, or reliable department guidance.
Example: "Texas has a $500,000 economic nexus threshold" is correct, but without citing 34 Tex. Admin. Code § 3.286(b), you can't put it in a memo. If the tool doesn't surface the citation, you have to find it yourself, which defeats the purpose of using the tool.
Multi-state work is inherently comparative. You need to know how California's single-sales-factor apportionment differs from Alaska's three-factor formula, or why Texas sources its franchise tax differently than a corporate income tax state. Most AI tools answer one state at a time and can't structure a cross-state comparison with parallel citations.
Some AI tools invent a change that never happened. California, Texas, and Pennsylvania all run revenue-only economic nexus tests today, but describing them as states that "recently repealed" a transaction-count prong is wrong unless you've confirmed the specific legislative history. Washington actually did remove its 200-transaction threshold in 2019 (SSB 5581). Confusing that pattern with a state that never had one produces confident-sounding misinformation.
Example: A general AI tool states "Pennsylvania eliminated its 200-transaction threshold in 2024." That specific claim needs to be verified against PA Act 13 of 2019 and Sales & Use Tax Bulletin 2019-01. Getting the legislative history wrong means your memo cites the wrong authority.
Economic nexus thresholds range from $100,000 to $500,000 in revenue depending on the state, and at least 16 states now use a revenue-only test with no transaction count. The thresholds and tests differ by state, and local home-rule nuances or marketplace-only rules should be checked in edge cases.
Thresholds change. Verify the current amount against the state's Department of Revenue or statute before you rely on it in a client memo.
Most states use a single-sales-factor apportionment formula for corporate income tax, but Alaska and a few others retain the traditional three-factor approach, and Texas uses an entirely different margin-tax framework.
California and Pennsylvania both generally use single-sales-factor apportionment. For California, this applies to business income for tax years beginning on or after January 1, 2013, with exceptions for agricultural and extractive activities. For Pennsylvania, the sales-factor-only formula applies for tax years beginning after December 31, 2012 (72 P.S. § 7401(3)2.(a)), with special methods for certain industries.
Alaska retains an equally-weighted three-factor formula (property, payroll, and sales) for many corporate taxpayers under its adoption of the Multistate Tax Compact.
Texas has no corporate income tax. Its franchise tax (margin tax) uses a single-receipts-factor sourcing method, but the tax base is margin, not net income, so it isn't a true corporate income tax comparison. Treat Texas as its own category when researching apportionment.
Whether a state uses rolling or static conformity to the IRC determines which federal provisions apply for state tax purposes. AI tools that ignore this distinction produce wrong answers.
Rolling conformity states generally adopt federal changes as they happen, subject to decoupling rules. New York is a common example, though it applies specific decoupling provisions, so the real research question is which provision, which tax, and which year.
Static conformity states adopt the IRC as of a fixed date and need separate legislation to update. California conforms to the IRC as of January 1, 2015, but selectively conforms to many post-2015 provisions and modifies numerous IRC sections on its own. An answer that applies current federal law to a California question without checking the state's specific conformity position will be wrong. For a deeper look at where California diverges, this guide to California state tax conformity covers the traps.
Kentucky illustrates how quickly a static conformity date can move: under HB 757, Kentucky updated its IRC conformity date to December 31, 2025 for taxable years beginning on or after January 1, 2026, while retaining state-specific decoupling provisions.
Filing deadlines, extension rules, and estimated payment requirements differ by state. Most states follow the federal April 15 deadline for individual returns, but corporate deadlines, composite return rules, and pass-through entity election windows vary. Missing a state-specific deadline because a research tool only surfaced the federal date is a preventable error.
Not every AI tax research tool is built for SALT. Five criteria separate the ones that are:
Bizora checks these boxes with a curated primary authority database covering all 50 states, View Steps reasoning transparency, and Vault for storing compliance project files and running research against client documents. For a side-by-side comparison of how other platforms handle SALT, this breakdown of the top AI tax research tools covers the specifics.
Multi-state compliance is where the gap between generic AI and professional-grade research is widest. Every state sets its own nexus rule, apportionment formula, and conformity position. A tool that blends them together or invents a change that never happened creates real risk for the practitioner relying on it.
The right process traces the specific statute or department guidance behind every number, documents the reasoning, and checks the position again as state law changes. For a broader look at how firms are integrating AI into multi-state workflows, this guide to AI for tax professionals covers what top firms are doing differently.
If you want to test what defensible SALT research looks like in practice, you can try Bizora's coverage directly at bizora.ai.
Some can. Most can't. The dividing line is citation quality: does the tool cite the actual state statute (like Cal. Rev. & Tax. Code § 6203(c)(4) for California's $500K threshold), or does it give you a number with no source? If there's no citation, there's no defensible position.
Applying the wrong nexus test to the wrong state, relying on an outdated threshold, and missing a state's conformity position. California's static conformity to the 2015 IRC means dozens of post-TCJA provisions don't apply without specific legislative adoption. One wrong assumption cascades through the entire state return.
Test it on your hardest question. Ask about a state's current nexus threshold, then apportionment, then conformity, all in the same session. Check whether it cites statutes, maintains context across follow-ups, and correctly distinguishes between states. If it can't show you the authority behind a number, it isn't ready.
Rolling conformity (New York) means automatic adoption of IRC changes as enacted, with state-specific decoupling. Static conformity (California, tied to 2015 IRC; Kentucky, recently updated to December 31, 2025 under HB 757) means the state freezes at a date and selectively adopts later provisions. The practical impact: you must check which IRC version a state applies before relying on any federal provision for state purposes.
At least 16 as of April 2026 (per Avalara's tracking), with Kentucky scheduled as the 17th on August 1, 2026 (HB 757). California, Texas, and Pennsylvania run revenue-only tests today. Verify each state's current standard against its statute or Department of Revenue guidance, since this count changes as legislatures act.
No. Most use single-sales-factor for corporate income tax (California, Pennsylvania, and 20+ others). Alaska retains a three-factor formula under the Multistate Tax Compact. Texas doesn't fit either category because its franchise tax is a margin tax with its own sourcing rules. Always confirm the formula and check for industry-specific carve-outs.
Related articles