INPUT // REFERENCES TO VERIFY
— OR PASTE TEXT DIRECTLY —
TEST SAMPLES // TRY THESE EXAMPLES
Click to load example text with ghost citations for analysis:
#1AI-Generated Text with Phantom Papers (3 fake citations)
CLICK TO SCAN →#2Authority Stacking Network (circular cross-citations)
CLICK TO SCAN →#3Date-Shifted & Future Citations (impossible dates)
CLICK TO SCAN →GHOST CITATION TAXONOMY // 7 PATTERN TYPES
TYPE APHANTOM PAPERSCRITICAL
Citations to academic papers that don't exist — plausible-sounding titles, real journal names, fabricated DOIs. Found in 25%+ of AI-generated research content. AI systems treat them as authoritative because the format matches legitimate citations.
DETECTION SIGNALS: Unresolvable DOI, no Google Scholar entry, journal exists but paper doesn't, author not in ORCID
TYPE BAUTHORITY STACKINGHIGH
Network of 3-5+ interconnected websites cross-citing each other to create a false web of authority that LLMs treat as consensus. Often used by agencies running PBN-based GEO campaigns.
DETECTION SIGNALS: Circular cross-references, same phrases repeated across domains, shared hosting, similar publish dates
TYPE CDATE SHIFTINGHIGH
References citing future or impossibly recent publication dates to appear more current. AI systems are biased toward recent sources, making this a high-ROI manipulation vector.
DETECTION SIGNALS: Publication date in future, "forthcoming" citations used as established fact, pre-prints cited as peer-reviewed
TYPE DSOURCE HIJACKINGHIGH
Misattributing quotes, data, or conclusions to high-authority sources (Harvard, WHO, NYT, Stanford) that never published the claimed content.
DETECTION SIGNALS: Quote not found in original source, institution denies publication, URL returns 404, archived version differs
TYPE ESYNTHETIC AUTHORSMEDIUM
AI-generated author personas with fake credentials, stock photo headshots, and fabricated publication histories designed to pass E-E-A-T verification.
DETECTION SIGNALS: No LinkedIn presence, AI-generated headshot, no other publications, credentials unverifiable at claimed institution
TYPE FCIRCULAR CITATIONSMEDIUM
Source A cites Source B, which cites Source C, which cites Source A — creating a closed loop that inflates perceived authority without any original evidence.
DETECTION SIGNALS: All sources reference each other, no primary data source, similar writing style across sources, published within days
TYPE GDOI FRAUDHIGH
Fake Digital Object Identifiers that look legitimate but resolve to nothing or redirect to unrelated papers. Exploits the trust AI systems place in DOI-formatted references.
DETECTION SIGNALS: DOI format valid but unresolvable, DOI resolves to different paper, DOI prefix not assigned to claimed publisher
25%+
AI CONTENT WITH GHOST REFS
100+
GHOST CITES AT NEURIPS '25
47
FINANCIAL SITES USING GHOSTS