
Oct 6, 2026
What Changed in the 2026 HTS: Annual Schedule Updates Importers Missed
The 2026 Harmonized Tariff Schedule has gone through several updates during the year, which makes it important for importers to keep track of the exact version they are using. By September 9, the latest version was Revision 18, published on September 2. Each revision can introduce changes that affect how products are reported, how duties are applied, or whether additional tariff provisions need to be included.
A review of the official change records from the Basic Edition through Revision 18 found 2,072 recorded change events. Many of these did not involve a change to the main HTS classification. Instead, they included new or discontinued statistical suffixes, duty rate changes on existing tariff numbers, and updates to Chapter 99 provisions. These changes can be easy to miss because the main classification may still look the same, even though the reporting or duty treatment has changed.
Key 2026 changes teams may have missed
The first statistical wave did not land on January 1. It landed February 1, after the Fall 2025 section 484(f) cycle was delayed.
432 ten-digit reporting numbers were established and 105 discontinued, none of which changed a legal classification.
9903.04.63 had its duty rates modified on July 31 without the tariff number changing at all.
Chapter 99 turned over hard: 188 provisions appeared and 242 disappeared between the Basic Edition and Revision 18.
The official change records exclude staged FTA rate reductions, so reading them alone cannot produce a complete annual comparison.
Why "The 2026 HTS" Is Not a Precise Citation
Between January 16 and September 2, the US International Trade Commission published 18 numbered revisions. Each supersedes the last and reproduces the entire schedule, so a reference to "the 2026 HTS" identifies a year rather than a document. For a record intended to survive review, the schedule needs identifying the way a software build does: Harmonized Tariff Schedule of the United States (2026), Revision 18, published September 2, 2026, checked on a stated date.
Publication date and effective date are separate facts, and in 2026 they came apart repeatedly. Revision 18 was published September 2 but carries provisions effective September 1 and September 3, plus conforming changes reaching back to July 1. Any single change therefore needs four details recorded:
Revision: The revision that introduced the change
Publication date: When that revision was published
Effective date: When the provision legally took effect
Legal instrument: The proclamation, notice, or other authority behind the change
The distinction that drives most of what follows is between the eight-digit legal subheading and the ten-digit statistical reporting number. A change to the first alters your legal classification. A change to the second alters what you must transmit without touching the classification at all, which is precisely why it goes unnoticed. Teams that store only the eight digits, treating the last two as a formatting detail rather than a legal one, have no field in which the change can register.
Where the 2026 Changes Concentrated
Counting change-record rows attached to numbers outside Chapter 99, the distribution skews heavily toward four chapters:
Chapter | Change events | Concentration |
|---|---|---|
30 | 174 | Medicaments and pharmaceutical reporting splits |
84 | 108 | Machinery, pumps, engines, bearings |
29 | 62 | Organic chemicals and pharmaceutical ingredients |
85 | 58 | Electrical equipment, batteries, semiconductors |
09 | 32 | Coffee reporting-number restructuring |
18 | 24 | Cocoa and chocolate products |
20 | 23 | Prepared vegetables and fruit |
Almost all of the Chapter 29, 30, 84, and 85 comes from section 484(f) statistical work rather than changes to the underlying six- or eight-digit classification. Importers in pharmaceuticals, chemicals, and machinery saw the most churn while their legal classifications held steady, the least intuitive combination to catch.
Broken down by what actually happened to each provision, ten-digit statistical activity dominates:
Level | Change type | Events |
|---|---|---|
Ten-digit statistical | Established | 432 |
Ten-digit statistical | Discontinued | 105 |
Ten-digit statistical | Annotated | 65 |
Ten-digit statistical | Deleted | 63 |
Ten-digit statistical | Modified | 46 |
Eight-digit legal, predominantly Chapter 99 | Added | 297 |
Eight-digit legal, predominantly Chapter 99 | Modified | 229 |
Eight-digit legal, predominantly Chapter 99 | Deleted | 226 |
These count events, not unique provisions: a single provision can appear in more than one record across the year.
The Two Statistical Waves, and Why the First One Moved
The February 1 Statistical Wave
Revision 2, published January 30, brought the first major wave of the year. It lists section 484(f) Committee changes as its only modification source and affects chapters spanning food, chemicals, base metals, machinery, electrical equipment, and the Statistical Annexes. It took effect February 1 rather than the customary January 1, because the Fall 2025 Committee cycle had been delayed. A compliance calendar built around a January 1 assumption simply missed it.
CBP deployed the associated changes through HSU 2602, reporting 5,405 harmonized-tariff records and 25,233 ABI records. Those are system-record counts, not counts of newly created HTS numbers, and the gap between the two is where the false comfort lives. A deployment of that size updates what ACE will accept; it decides nothing about which successor suffix fits a given SKU.
The July 1 Statistical Wave
The second wave arrived with Revision 11 on July 1, again driven solely by section 484(f) changes, reaching chapters from food and chemicals through machinery, vehicles, and furniture. CBP's HSU 2614 carried 310 harmonized-tariff records and 1,287 ABI records. The July 1 approved list shows exactly what a transfer looks like in practice: 0704.20.0000 was divided into 0704.20.0020 and 0704.20.0040 for Brussels sprouts.
That division is the whole problem in miniature. One old basket became two reporting categories, and no system can pick between them without a product fact it may never have been asked to store. Where a master holds only the eight-digit code, software will often select a valid default suffix without asking whether the item is organic, pharmaceutical-grade, or a specific machinery component. The entry is accepted. The reporting is wrong.
The Changes That Move Duty Without Moving a Code
Three 2026 examples show duty shifting where a search for new or deleted tariff numbers finds nothing.
A rate change on an unchanged number: Revision 15 contains exactly one listed change: 9903.04.63 was modified as to its rates of duty, effective July 31, 2026, under the Notice of Reduction of Tariffs on Patented Pharmaceuticals and Pharmaceutical Ingredients for Products of the United Kingdom at 91 Fed. Reg. 49406. The number is identical before and after. Only the duty moved.
A legal effective date preceding publication: Revision 1 implemented Presidential Proclamation 11002, imposing a 25% additional duty ad valorem on covered advanced computing chips and derivative products for entries on or after 12:01 a.m. EST on January 15, 2026. USITC published Revision 1 on January 16, one day after the duty took legal effect. Entries filed in that window were governed by a schedule that had not yet been published, a pattern that recurs whenever a Section 232 action moves faster than the compiled schedule.
Rate movements the change record does not list: The Basic Edition change record states explicitly that it does not include previously proclaimed staged-duty reductions taking effect January 1, 2026, even though the release page names staged FTA rates as a modification source. A normalized comparison of common tariff-number rows between the 2025 and 2026 Basic Edition datasets found 209 shared provisions with changed Special-column text, including 59 where the prior Korean rate presentation moved into the Free country group and 84 where Panama did. Those figures come from the official edition files but are a controlled data comparison, not an agency-published aggregate.
Chapter 99 turnover compounds all of this. The 2026 Basic Edition added 244 Chapter 99 eight-digit provisions relative to 2025 and no longer contained 451 provisions under headings 9912 and 9913. Between that Basic Edition and Revision 18, a further 188 provisions appeared and 242 disappeared. Because Chapter 99 requires reporting the applicable provision in addition to the ordinary ten-digit number, an importer whose Chapters 1 to 97 classification is perfectly correct can still file a non-compliant entry by missing the overlay line.
What Broker Software Is Least Likely to Flag
Ranked by how quietly each change type passes through a workflow:
Rank | Change type | Why it passes unnoticed | Consequence |
|---|---|---|---|
1 | Same-number rate change | No classification key changes | Underpayment, overpayment, wrong landed cost |
2 | One-to-many statistical transfer | Software substitutes a valid default suffix | Accepted entry, inaccurate reporting |
3 | Note-only scope change | No tariff number necessarily changes | Wrong inclusion, exclusion, or timing decision |
4 | New Chapter 99 overlay | Ordinary code remains valid | Missing additional-duty or exemption line |
5 | Effective date preceding publication | Database cannot support the legal date | Entries filed under superseded treatment |
6 | Unit-of-quantity change | Code and description stay stable | Quantity rejection or inaccurate reporting |
7 | Renumbered note subdivision | Saved decisions point to obsolete text | Unsupported classification analysis |
8 | Deleted legal subheading | Usually produces a hard system alert | Clear rejection, so rarely missed |
Neither USITC nor CBP publishes a specification for how broker software generates alerts, so this ranking is workflow analysis rather than a claim about any vendor. The pattern holds regardless: changes that break loudly get fixed, and changes that break nothing get filed.
How This Comparison Was Built
This analysis uses two different counting methods:
Method | What it counts | What it measures |
Change-record count | Every row in the Basic Edition and each noncumulative revision record | Change events, so one provision can appear more than once |
Schedule-diff count | Normalized HTS number fields between two exact editions | Net additions and removals between those editions |
The 2,072 figure above is a change-record count from the 2026 Basic Edition through Revision 18. Because the source records group note references and wrap text across lines, it is a reproducible research count rather than an official USITC aggregate. The Chapter 99 and staged-rate figures are schedule diffs against the official USITC datasets.
One limitation is worth stating plainly. A September 8 proclamation announced further Canada-related modifications effective September 15, 2026, not incorporated in Revision 18 as at this analysis. The superseding revision and its final text were still pending.
The Ten Questions a 2026 Update Project Should Answer
Reconciling a product master against the current schedule differs from confirming a broker's database is current. The second tells you nothing about your own catalogue.
Which HTS edition and revision is stored in the product master?
Does the master store six, eight, and ten digits separately?
Can it assign a Chapter 99 code by origin, entry date, supplier, declared use, and exemption status?
Were the February 1 and July 1 suffix transfers mapped at SKU level rather than defaulted?
Were units of quantity and secondary-quantity requirements compared?
Were duty fields on unchanged numbers compared?
Were staged FTA rates loaded, given the change record omits them?
Were note citations in rulings and internal memoranda checked for renumbering?
Can the broker return the actual HTS revision identifier used on an entry?
Were entries filed between a legal effective date and the schedule's publication date reviewed separately?
That last question separates a real reconciliation from a database refresh, and it is unanswerable without version-level records. CBP's acceptance of a ten-digit number confirms the number is operational in ACE. It confirms nothing about whether the right successor suffix was chosen, and that gap is where an accepted entry quietly becomes an expensive one.
Gaia helps teams keep HTS revisions, successor suffixes, classification records, and product data connected across the update process. Teams can start a free trial to see how Gaia supports ongoing HTS reconciliation.
FAQ
Does a discontinued suffix mean my classification was wrong?
No. A ten-digit suffix is a statistical reporting number and does not carry the legal classification, which sits at the six- and eight-digit levels. A discontinued suffix means the reporting category was retired or split, not that the analysis failed. It still needs correcting: it can fail ACE validation, and a split needs a product-specific decision.
Why did the statistical update land on February 1?
The Fall 2025 section 484(f) Committee cycle was delayed, so USITC targeted the approved reporting numbers for February 1 rather than the customary January 1. Revision 2, published January 30, carried the changes, deployed by CBP through HSU 2602. Calendars assuming a fixed January 1 date missed the wave.
Do thousands of deployed ACE records mean thousands of codes changed?
No. HSU 2602 contained 5,405 harmonized-tariff records and 25,233 ABI records; HSU 2614 contained 310 and 1,287. Those count system records touched in an ACE deployment, not tariff numbers established or discontinued. Deployment volume runs far larger than what a classifier reviews.
Do the change records capture every 2026 change?
No. The official change records expressly exclude previously proclaimed staged duty reductions, endnotes, and minor formatting changes. They are also noncumulative for in-year revisions, so the full picture requires reading every intervening record or directly comparing two complete machine-readable editions.







