How Will AI Shape the New Public Charge Rule for Spouse Visas?

The July 16, 2026 public charge rule did not simply change a test. It removed much of the regulatory vocabulary that attorneys had been using to prepare clients for that test. In spouse visa and marriage-based green card applications, the immediate problem is not only whether a U.S. citizen’s spouse has used a benefit, has a medical condition, or has a thin credit file. The problem is how those facts will be located, labeled, and elevated inside a USCIS workflow before an officer writes a decision.

USCIS announced that DHS is rescinding the 2022 public charge regulation effective September 18, 2026, removing the “primarily dependent” framework, requiring applicants to use a new Form I-485 after that date, and replacing the prior regulatory definitions with a discretionary standard to be developed further through future policy guidance.[1] Fragomen’s same-day analysis reads the change the same way: the rule eliminates the existing provisions at 8 CFR 212.21 and related sections, deletes definitions including “public charge” and “primarily dependent,” and leaves interpretation to sub-regulatory guidance that had not been published as of July 19, 2026.[2]

That gap matters because adjudication does not pause while everyone waits for cleaner language. Files still arrive. Forms still need intake. Exhibits still need to be sorted. Officers still need to decide which parts of a marriage-based adjustment packet deserve attention. When a rule stops defining its own key terms, the surrounding systems can begin to define them operationally.

The Rule Creates Space for Workflow to Matter

Marriage-based green card cases often look straightforward from the outside: a U.S. citizen spouse, an intending immigrant spouse, a bona fide relationship, and a packet of civil records, tax documents, medical forms, employment history, and financial evidence. In practice, the file is a dense evidentiary record. It may include years of address history, inconsistent job dates, health insurance gaps, prior benefit-related correspondence, old immigration filings, bank statements, affidavits, medical exam results, and explanations of household income.

Under a tightly defined public charge rule, practitioners can at least organize the file around known legal categories. Under the July 2026 framework, the harder question is what counts enough to matter. USCIS has said more guidance will come, but the rule itself shifts the center of gravity away from fixed regulatory definitions and toward officer discretion.[1][2]

That discretion will not operate in a paper vacuum. USCIS already uses technology to help process immigration submissions. GT Law’s 2026 discussion of AI in immigration adjudications describes USCIS’s Evidence Classifier as a machine-learning tool that automatically categorizes and tags documents submitted with petitions, with the stated purpose of reducing processing times.[3] That is a mundane-sounding function until the legal standard becomes unusually open-ended.

Document classification is not a denial. A tag is not a finding. But anyone who has assembled an adjustment packet knows that the first organization of the record can shape the review that follows. If medical documentation, benefit history, insurance information, employment gaps, or prior filings are surfaced early and grouped as relevant to public charge, the officer’s individualized review begins from a file that has already been made legible in particular ways.

How a Marriage-Based File Becomes Machine-Readable

A spouse’s adjustment filing is not a single story. It is a stack of structured forms and unstructured proof. The forms ask for dates, addresses, employment, immigration history, household information, and admissibility answers. The exhibits try to explain the human mess underneath: why one spouse stopped working, why the household used a public program, why an illness affected income, why a sponsor’s tax transcript does not match the current pay stubs, or why a previous filing listed an old address differently.

In an AI-assisted workflow, the first move is likely not a dramatic risk score. It is categorization. A lease goes into one bucket. A tax transcript goes into another. A medical record, insurance card, benefits notice, affidavit, or prior approval notice may be tagged according to document type or content. GT Law’s description of the Evidence Classifier supports that narrower point: the tool is used to automatically categorize and tag submitted documents, not to publicly announce final eligibility determinations.[3]

Once the file is classified, the public charge issue can be built from several kinds of signals:

• Benefit-related signals: records or database checks may indicate past or current use of public benefits, even where the legal significance of a particular benefit remains disputed or unclear.
• Medical signals: Form I-693 materials, insurance documentation, or applicant explanations may identify a condition that becomes relevant to financial self-sufficiency analysis.
• Financial signals: tax transcripts, pay stubs, bank records, debts, credit history, and sponsor evidence may be grouped in ways that highlight household instability.
• Consistency signals: old filings, address histories, employment dates, and benefit records may create apparent discrepancies that require explanation.
• Omission signals: missing insurance proof, incomplete employment history, or thin documentation may be treated as a reason to request more evidence or to scrutinize the file more closely.

This is where the public charge rule’s breadth becomes operationally important. Anwari Law’s July 16, 2026 client alert describes the new framework as allowing officers to consider any information deemed relevant, including medical conditions, insurance coverage, employment gaps, debts, credit history, and prior immigration filings.[4] That is a practitioner reading of the rule, not binding agency guidance. Still, it captures the practical anxiety: if the officer may consider almost any relevant fact, then the systems that locate “relevant” facts become part of the legal risk.

The SAVE system adds another layer. NILC has documented that SAVE verifies benefit receipt across federal and state databases and has been expanded under the current administration. NILC also notes that SNAP, Medicaid, and housing assistance had not been part of the public charge test in more than a century of practice.[5] The July 2026 rule does not by itself tell practitioners exactly how each benefit will be weighed. In the meantime, the existence of a verification pathway means benefit information can move into the adjudication environment even before the legal treatment of that information is fully explained in public guidance.

The important distinction is adoption versus outcome. Evidence Classifier and SAVE can help USCIS process and verify information. The available materials do not prove that those tools make final public charge determinations, assign binding scores, or automatically deny spouse-based green card cases. The better-supported concern is subtler: automated categorization and verification can decide what is made conspicuous, what is routed for attention, and what the applicant may later have to rebut.

Individualized Discretion Can Still Begin with Preselected Facts

The public-facing legal promise is individualized discretion. The officer looks at the totality of the circumstances, weighs the facts, and decides whether the applicant is likely to become a public charge. That is familiar language in immigration practice. It is also language that depends heavily on what reaches the officer’s attention.

A file can be individualized and still be triaged. Those are not opposites. An officer may retain final authority while reviewing a record that has already been sorted by document type, database return, inconsistency flag, or category label. If the system highlights benefits, medical issues, employment gaps, debts, or prior filings, the officer’s discretion may be exercised around those highlighted items rather than around the applicant’s full narrative in the order counsel intended.

That is a workflow problem, not a science-fiction problem. In a cleanly prepared marriage-based adjustment packet, counsel may put a careful explanation of Medicaid history next to evidence of current private insurance, sponsor income, household support, and medical stability. If the benefits notice and the medical record are classified separately and surfaced as public charge-relevant items, the explanation may still be in the file, but it may not be the first thing that structures the review.

The May 2026 USCIS policy memorandum described in Abachi Law’s analysis adds another discretionary overlay by reframing adjustment of status as extraordinary relief and a matter of discretion, not merely a mechanical approval after statutory eligibility is met.[6] That memorandum is not the public charge rule. But for in-country spouse-based adjustment cases, the timing matters. A broad public charge standard and a broader discretionary framing of adjustment can operate in the same file.

The result is a dual system. On paper, the officer performs an individualized review. In operation, the file may arrive already shaped by classification, verification, and triage. The harder a rule is to pin down, the more consequential those preliminary steps become.

Why Spouse Cases Deserve Particular Attention

Spouse-of-citizen cases are not a niche corner of immigration practice. They are common, emotionally high-stakes, and document-heavy. They also tend to involve household facts that do not fit neatly into eligibility boxes: caregiving, pregnancy, illness, interrupted employment, shared debt, mixed immigration histories, and periods when one spouse relied on family or public support.

The Manhattan Institute’s November 2025 issue brief estimated, using an FY2019 baseline, that 59,290 out of 304,338 spouse-of-citizen visa applicants, or 19.5 percent, could be affected under the points-based public charge framework it proposed.[7] That figure should be handled carefully. It is not DHS policy, and it does not predict denial rates under the July 2026 rule. Its value is narrower: it shows why spouse applicants could be a high-impact population if public charge screening becomes stricter or more heavily operationalized around financial and benefit-related indicators.

For attorneys, the spouse context also changes the counseling conversation. A family may understand that income matters. They may not understand why an old benefits notice, a gap between jobs, a medical diagnosis, or a mismatch between a prior filing and a current form could become public charge material. They are even less likely to know whether that fact was noticed by an officer, a classifier, a database check, or some combination of all three.

What Practitioners Can Watch Before Guidance Arrives

Until DHS publishes the promised sub-regulatory guidance, responsible advice has to stay inside the uncertainty. It is too early to say that the 2026 framework will recreate any prior public charge regime in full. It is also too easy to say that nothing has changed because officers have always had discretion. The practical change is that discretion is now operating alongside increasingly automated ways of organizing the file.

The immediate work is not to guess a secret score. It is to prepare and review spouse-based filings with the assumption that certain categories of evidence may be extracted, grouped, or checked outside the narrative sequence counsel creates. That changes how explanations should be drafted and how adverse notices should be read.

Source verification becomes especially important. If an adverse reference appears to describe benefit use, counsel needs to know whether the information came from the applicant’s own filing, a SAVE return, a state record, a prior immigration filing, or an inference from submitted documents. Each source creates a different rebuttal problem. A database mismatch is not answered the same way as a correctly identified benefit with disputed legal relevance.

The same is true for medical and financial evidence. A diagnosis is not the same as an inability to work. A lapse in insurance is not the same as future dependency. A sponsor’s old tax return may not reflect current income. A debt entry may have a payment plan, a co-signer, or a context that never appears if the file is reduced to category labels. The legal response has to pull the review back from the signal to the record.

The Challenge Is Making the Hidden Layer Reviewable

Public charge advocacy after July 2026 will not be only about collecting more evidence. It will also be about preserving the path of reasoning. If USCIS issues a request for evidence, notice of intent to deny, or denial that appears to mirror classifier categories, counsel should look for what is missing: the source of the fact, the legal standard applied to it, the weight assigned, and the countervailing evidence considered.

That does not require treating every automated tool as hostile. A well-functioning classifier can make a buried exhibit easier to find. A verification system can confirm information that otherwise would take weeks to document. The due process problem begins when a label, database return, or triage flag becomes practically decisive while the applicant receives only a generalized explanation.

For now, the most defensible interim judgment is narrow but important: the hidden operational layer is part of the risk environment for spouse visa and marriage-based green card counseling. DHS has rescinded the 2022 rule, stripped out key definitions, required a new I-485 after September 18, 2026, and promised guidance that had not arrived as of July 19, 2026.[1][2] USCIS also operates tools that classify evidence and verify benefit-related information.[3][5] Until the agency explains how those systems interact with the new discretionary public charge standard, practitioners should treat unexplained public charge concerns as both legal issues and workflow issues.

This is not legal advice for any specific family. It is a warning about where the record may be shaped before the family ever sees the government’s reasoning.

References

1. U.S. Citizenship and Immigration Services Rescinds 2022 Public Charge Regulation, USCIS, July 16, 2026.
2. United States: DHS Replaces Public Charge Regulation with Policy Guidance and Broad Discretion for Adjudicators, Fragomen, July 16, 2026.
3. The US Government's Expanding Use of AI in Immigration Adjudications, GT Law, 2026.
4. Client Alert: Important Change to the Public Charge Rule, Anwari Law, July 16, 2026.
5. Public Charge: What Advocates Need to Know About the November 2025 Proposed Rule, NILC, November 2025 and April 2026.
6. Marriage Green Card Scrutiny 2026: US Citizen Spouses, Abachi Law, July 14, 2026.
7. A Points-Based Public Charge Rule, Manhattan Institute, November 2025.