About ImgFact

Built for the professionals who can't afford to be wrong.

ImgFact gives legal teams and insurance investigators forensic-grade image authentication — without the $5,000 lab fee or the forensics degree.

Start a verification Contact us

Market problem

$308B

Lost to insurance fraud annually in the US. AI-generated claim photos are a fast-growing share of that number.

Scope

20–30%

Of P&C claims now include AI-altered images or documents, according to Shift Technology 2025.

Speed

<15s

Time to get a forensic verdict with ImgFact. Fast enough for real workflow use at scale.

Why ImgFact exists

AI-generated images are now indistinguishable from real photographs to the human eye.

Tools like DALL-E, Midjourney, and Stable Diffusion can produce photorealistic images of accidents, property damage, and people that never existed — in seconds, for free. In 2025 alone, over 518 documented cases of AI-generated content appeared in US courts. Louisiana, California, and eight other states now require attorneys to exercise "reasonable diligence" in verifying evidence authenticity — or face sanctions.

Insurance fraud involving AI-generated claim photos has reached an estimated 20–30% of all submitted claims. The existing tools are either $5,000+ forensic lab software requiring specialist training, or consumer apps with no chain-of-custody documentation.

ImgFact is the affordable, self-serve middle layer built specifically for the legal and insurance workflows that need it most.

No competitor offers side-by-side forensic comparison of two images. ImgFact's Comparison Mode — powered by Claude AI — is the only tool in the market that analyzes two images simultaneously and produces a plain-English discrepancy report.

Forensic methodology

Eight independent signals. One verdict.

Every ImgFact scan runs eight forensic signals in parallel, then synthesizes them into a single verdict using weighted AI analysis. No single signal determines the outcome — the strength of the verdict comes from consensus across multiple detection methods.

Error Level Analysis (ELA)

Re-saves the image at a known compression level and measures which pixels changed most. Edited or AI-generated regions compress differently than originals.

EXIF Metadata Forensics

Every real camera photo contains hidden metadata — make, model, GPS, timestamp. Missing or inconsistent EXIF is the single strongest indicator of a fake.

Noise Pattern Analysis

Real camera sensors produce consistent random noise. AI images and edited photos have unnatural noise — too uniform, or mismatched between regions.

Perceptual Hash Detection

Detects if an image has been cloned, copy-pasted, or is a near-duplicate of another image — even after resizing or color adjustments.

Power-Law Frequency AnalysisImgFact™

Measures adherence to the 1/f natural image law using power-law slope fitting. Real photographs exhibit characteristic frequency decay that AI-generated images violate.

Complexity VarianceImgFact™

Analyzes spatial complexity variation across image blocks using four descriptors. Diffusion models distribute detail more uniformly than real photographs.

Color Channel Correlation

Real cameras use a Bayer filter array that creates predictable inter-channel dependencies. AI models have no knowledge of camera physics and often violate these dependencies.

Claude Vision AI

Anthropic's Claude Vision analyzes lighting consistency, shadow physics, texture realism, anatomical accuracy, and GAN artifact signatures.

Verdict synthesis

Heuristic signals and Vision AI are weighted and merged into a final ImgFact score (0–100) with a plain-English explanation of what was found. The Authentic Gate requires positive forensic evidence before returning an Authentic verdict — preventing false confidence on ambiguous images.

Data privacy & confidentiality

Designed for privileged and confidential material.

Legal and insurance professionals handle sensitive material. ImgFact is built around strict data minimization from the ground up.

Images are permanently deleted from our servers immediately after analysis completes. We never store your uploaded images.
We never use your images to train AI models or improve our systems. Your images stay yours.
Scan results and forensic reports are stored only in your account and protected by row-level security — no one else can access your results.
All data is encrypted in transit via TLS 1.3 and at rest via Supabase's enterprise-grade infrastructure.
We do not sell, share, or expose your data to third parties under any circumstances. HIPAA BAA available — contact hello@imgfact.com.

Honest limitations

A powerful first-line screening tool. Not a replacement for expert testimony.

What you should know before using ImgFact in high-stakes situations

No detection method is 100% accurate. The confidence score on every result communicates how certain the analysis is — verdicts above 85% confidence are highly reliable, while lower scores indicate the image warrants further investigation or human expert review. ImgFact is designed to support diligence and documentation, not replace a certified forensic expert in court proceedings. For formal legal submission, consult with a qualified forensic analyst.

The team

Founded in Michigan.

Priyadarshi Gavali

Founder & CEO — ImgFact LLC (KPZG LLC)

Priyadarshi built ImgFact after identifying a clear gap in the legal and insurance markets for affordable, self-serve forensic image verification. He holds a Master's in Computer Information Systems with a focus on Business Information Systems from Colorado State University. He brings five years of enterprise sales experience and a background in AI product development, database design, and systems analysis. ImgFact is based in Kalamazoo, Michigan.

MS Computer Information SystemsColorado State UniversityKalamazoo, MIhello@imgfact.com

Start verifying images today.

Join legal teams and insurance investigators using ImgFact to catch fake evidence before it costs them a case or a claim.

Start a verification →Contact us