U.S. immigration agencies are rapidly expanding their use of Artificial Intelligence (AI) in case processing, fraud detection, and security screening. While these systems are intended to improve efficiency, they are also contributing to a measurable rise in Requests for Evidence (RFEs), Notices of Intent to Deny (NOIDs), denials, and erroneous rejections across multiple visa categories.
This alert summarizes what employers, HR teams, and foreign national employees need to know.
1. USCIS Is Now Using AI Throughout the Adjudication Process
Evidence Classification and Document Review
USCIS deploys the ELIS Evidence Classifier, a machine‑learning tool that automatically tags uploaded evidence and determines which documents adjudicators see first.
- The system has processed over 24 million page scrolls and significantly altered how officers review filings.
- USCIS has not published any error‑rate data, and practitioners report RFEs for documents that were in fact submitted, consistent with classifier mis‑tagging.
AI‑Powered Document Translation
USCIS uses an AI translation service that produces near‑instant English translations of foreign‑language documents.
Identity and Data‑Matching Systems
The Verification Match Model uses machine learning to match names, dates of birth, and identifiers across E‑Verify and SAVE. Minor inconsistencies can trigger automated flags that lead to RFEs or delays (and even erroneous rejections of filed petitions).
Fraud Detection and Pattern Analysis
USCIS and FDNS use AI systems to detect anomalies, cross‑reference filings, and identify potential fraud indicators.
2. AI Use Is Expanding at Consulates and Ports of Entry
Social Media and Open‑Source Screening
Customs and Border Protection (CBP) uses Babel X, an AI‑enabled tool that analyzes social media and open‑source data, including sentiment analysis and identity resolution.
This tool is used to screen travelers, including visa holders and U.S. citizens, at ports of entry.
Visa Revocations and Security Vetting
State Department and DHS systems increasingly rely on AI to identify “derogatory” or security‑related content, which has led to visa revocations and heightened scrutiny.
3. Correlation: AI Expansion and Rising RFE/Denial Rates
Documented Increase in Scrutiny
- EB‑2 NIW denial rates have climbed to nearly 40%, and EB‑1A petitions face heightened scrutiny.
- AI‑assisted screening is explicitly cited as a driver of increased RFEs and NOIDs.
How AI Contributes to Higher RFEs
- Mis‑tagged evidence → RFEs for “missing” documents that were submitted.
- Cross‑document mismatch detection → small discrepancies in job titles, dates, or signatures trigger automated flags.
- Template or “ghost text” detection → AI reads hidden text layers in PDFs, causing incorrect RFE assertions.
- AI‑generated boilerplate RFEs → vague, repetitive, or contradictory requests linked to classifier‑generated suggestions.
Systemic Impact: Student Status Terminations
More than 1,200 international students recently lost status due to automated database terminations, requiring nationwide litigation and government reversal.
This demonstrates how AI‑driven mismatches can produce large‑scale immigration consequences.
What does this all mean?
The expansion of AI across USCIS, DHS, and the State Department is already reshaping how cases are screened and adjudicated. Employers should expect the following practical impacts:
- Social‑media screening is now standard and AI‑assisted.
- Applicants must list all social‑media identifiers from the past five years, and consulates use AI tools to flag posts for additional review. This is causing significant visa‑stamping delays, sometimes lasting months.
- Online activity can directly affect visa outcomes.
- Posts that appear inconsistent with employment details, immigration status, or security‑related criteria can trigger 221(g) holds, refusals, or revocations. Employers should prepare employees accordingly.
- Expect more intake errors and erroneous rejections.
- AI‑driven data‑matching systems are producing false mismatches, resulting in rejections of properly filed petitions.
- AI‑based evidence classification and mismatch detection are generating more RFEs, even for strong, well‑documented cases.
- AI‑flagged cases move more slowly, and visa applicants should build in additional time for stamping, travel, and onboarding.
- RFEs will continue to rise across categories.
- Employers must plan for longer, less predictable processing timelines.
About the authors:
Scott Bettridge currently serves as the chair of Cozen O’Connor’s Immigration Practice and vice chair, office managing partner of the Miami office. He represents both corporate and individual clients in all aspects related to U.S. immigration law. Scott’s corporate clients include major global organizations in the financial services, hospitality, fashion, media, IT, energy, agriculture, insurance, and health care sectors, along with organizations specializing in professional services and private club industries.
David Adams focuses his practice on corporate immigration law. He has experience advising multinational corporate clients, including dozens of Fortune 500 companies on a variety of immigrant and nonimmigrant visas, PERM, permanent residency, and consular processing issues. David provides enterprise-wide advisory services to support corporate clients on a programmatic level and works alongside clients’ legal departments, mobility groups, HR, talent acquisition, and business stakeholders to help them understand policy changes, assess the impact of these changes to their workforce, and modify their immigration program accordingly.