Skip to content
Case Study

1 in 3 California college
applications were fake.

$13M gone to students who never existed. AI-generated ghost students enrolled, submitted coursework, collected financial aid, and disappeared. suss. sees what admissions can't.

116 campuses. 2.1 million real students competing for slots and aid against an army of synthetic identities.

The scale of the problem

$13M
Lost in state and federal funds
116
Campuses affected statewide
223K+
Fraudulent enrollments flagged

How ghost student schemes work

1. Fake application at scale
Fraudsters use stolen and synthetic identities to submit hundreds of applications programmatically. At some colleges, 50 applications arrived within two seconds.
2. AI-generated coursework
Ghost students “attend” class by submitting AI-written assignments. They look active enough to maintain enrollment and avoid triggering academic flags.
3. Financial aid disbursement
Pell Grants and state aid are disbursed to newly opened accounts. Maximum aid of $7,395/semester per ghost student, multiplied across thousands of fake enrollees.
4. Disappear
Once aid is cashed out, the ghost student vanishes. The identity was synthetic. There's no one to find. The money is gone.

Sources: ABC News, CalMatters, U.S. Dept. of Education OIG

How suss. catches ghost students

We reconstructed a ghost student application and enrollment flow and ran it through our API. Here's what fired.

90%
High Risk
Ghost Student Enrollment Fraud Detected

Text scan -- what suss. catches immediately

These signals fire from the application text alone, with no enrollment system access.

90%
SSN disclosed in application text
ssn_request
85%
Bulk submission pattern detected (application 47 of 48)
ghost_student_enrollment_fraud

With enrollment system integration -- detection multiplies

When suss. receives metadata from the enrollment system (submission velocity, account age, aid amounts), these additional signals fire. This is the difference between catching one application and catching the entire ring.

90%
48 applications from same IP block in 120 seconds
velocity_anomaly
85%
Financial aid routed to account opened 3 days ago
new_account_disbursement
80%
Maximum Pell Grant disbursement to Chime (neo-bank drop)
max_aid_fintech_routing
80%
No prior credit file, no employment history, PO box address
synthetic_identity_indicators

Recommended actions

  1. 1HOLD all financial aid disbursement for this application
  2. 2Flag IP block for velocity analysis -- 48 applications in 120 seconds is not human behavior
  3. 3Verify SSN through the Social Security Administration before processing enrollment
  4. 4Cross-reference mailing address against known commercial mail drops (UPS Store, PO boxes)
  5. 5Run AI detection on personal statement and submitted coursework
  6. 6Report to U.S. Department of Education Office of Inspector General

Why this is hard to catch without AI

Ghost students look real
They enroll in classes, submit assignments, and maintain activity. From the registrar's perspective, they're indistinguishable from real students.
AI-generated work passes human review
Coursework is generated by large language models and submitted on schedule. Individual professors can't detect synthetic work across hundreds of students.
Synthetic identities have no prior record to check against
These aren't stolen identities with fraud alerts. They're fabricated from scratch, with SSNs that pass basic validation but have no credit history.
Volume overwhelms manual verification
Community colleges process hundreds of thousands of applications. Manual identity verification at this scale is not feasible.
Speed of submission is invisible without instrumentation
50 applications in 2 seconds looks identical to 50 individual applications in a database. Without velocity analysis, the pattern is invisible.

Impact beyond the $13 million

Real students displaced

Ghost students fill enrollment slots in impacted classes and programs. Legitimate students are waitlisted or denied admission entirely.

Financial aid pools drained

Pell Grant and Cal Grant funds are finite. Every dollar disbursed to a ghost student is a dollar unavailable to a real student in need.

Faculty time wasted

Professors grade AI-generated assignments, respond to nonexistent students, and manage inflated class rosters that distort resource allocation.

Institutional credibility at stake

Campuses that can't distinguish real from fake students face scrutiny from accreditors, the Department of Education, and state legislatures.

Purpose-built enrollment fraud detection

Synthetic Identity Detection

Flags SSNs with no credit history, mismatched contact information, commercial mail drop addresses, and VoIP phone numbers registered within days of application.

Velocity & Behavioral Analysis

Detects programmatic submission patterns, same-IP-block clustering, and application timestamps that are impossible for human applicants.

AI-Generated Content Detection

Identifies machine-generated personal statements, coursework, and identity documents using content provenance analysis and AI detection models.

Financial Aid Anomaly Detection

Flags maximum-grant requests to newly opened accounts, disbursement routing to payment apps, and patterns consistent with aid harvesting at scale.

With suss. vs. without

Without suss.

  • Application arrives, passes basic validation
  • Synthetic SSN has no fraud alert on file
  • Student enrolls and submits AI-written work
  • Pell Grant disbursed to newly opened account
  • Ghost student cashes out and disappears
  • $13M drained across the system before anyone notices

With suss.

  • Application flagged at 90% risk on submission
  • Synthetic SSN and missing digital footprint detected
  • Velocity anomaly: 48 apps from same IP in 2 minutes
  • AI-generated personal statement flagged before enrollment
  • Financial aid hold triggered before disbursement
  • $13M protected, real students get their slots and aid

What institutions should do now

1
Instrument application pipelines for velocity
Track submission timestamps and source IP blocks. Human applicants don't submit 50 forms in 2 seconds. This is the single highest-signal detection layer.
2
Verify identity before disbursing aid
Cross-reference SSNs with the Social Security Administration. Check mailing addresses against commercial mail drop databases. Flag accounts opened within 30 days of disbursement.
3
Deploy AI content detection on submitted coursework
Run AI detection across enrollment essays, assignments, and discussion posts. Ghost students use AI-generated content that is detectable at scale even when individual assignments appear plausible.
4
Integrate trust scoring at the enrollment layer
suss. can score applications at submission time, flagging synthetic identities, velocity anomalies, and content provenance issues before any enrollment or aid is processed.

Built for the scale of the California system

116 campuses, 2.1 million students, one trust layer. Detect ghost students at the application layer, before enrollment, before disbursement, before it costs another $13 million.

See the campus solutionView enterprise plans

One API call per application. Real-time scoring. No IT integration required.

Free pilot for qualified California community colleges and state institutions

508
Scam signals
14
Campus-specific
94.5%
Precision
93.2%
Recall

Sources: ABC News | U.S. Department of Education Office of Inspector General | CalMatters | LACCD

made by humans who've been scammed.

© 2026 suss. AI, Inc.