The 2026 State of AI Photo Scams: A Data Roundup

All the sourced statistics on AI-generated image fraud, romance scams, elder fraud, deepfakes, and fake profiles — in one place, linked to primary sources. Updated August 2026.

Quick Answer

$1.14 billion lost to romance scams in 2023 (FTC). $3.4 billion in online fraud losses for adults 60+ in 2023 (FBI IC3). 1 in 4 job applicants projected to be fake by 2028 (Gartner, Nov 2024). 88-98% AI image detection accuracy on standard test images (2026). No detector exceeds 45-60% accuracy against specifically evasion-optimized images (Imagera, 2026).

Romance scam losses

$1.14B

Romance scam losses reported to FTC, 2023 — highest fraud category
FTC Sentinel 2023

$2,000

Median individual romance scam loss, 2023
FTC Consumer Alerts

$9,000

Median loss for victims 70+ — 4.5x overall median
FTC Consumer Alerts

$41M

Grandparent scam losses reported to FTC, 2023
FTC Consumer Alerts

Elder fraud and online fraud losses

$3.4B

Total online fraud losses for Americans 60+, 2023
FBI IC3 Elder Fraud 2023

#1

Age group by total online fraud loss — adults 60+ exceed all other groups
FBI IC3 Elder Fraud 2023

$308.6B

Estimated total US insurance fraud annually (all types)
Coalition Against Insurance Fraud

Fake profiles and fake job applicants

Statistic Figure Source
Fake job applicants by 2028 1 in 4 applicants Gartner, November 2024
Dating app users who've encountered fake profiles Significant majority Pew Research Center, 2023 dating survey
AI-generated headshots: primary advantage over stolen photos No reverse image search results Operational advantage: AI faces have no prior internet presence

AI image detection accuracy benchmarks (2026)

Tool / API Stated accuracy (standard test) False positive rate Notes
Hive Moderation (API) 89-98% ~3% Enterprise API; no consumer extension
Illuminarty ~91% Not publicly stated Self-reported; heatmap explainability
FauxSpy / Sightengine genai ~88% ~3% Sub-second processing; no account free
All tools (evasion-optimized images) 45-60% max — Imagera, 2026 adversarial test

Accuracy figures are from tools' own benchmarks unless otherwise noted. Independent benchmarks vary by test methodology.

C2PA and regulatory developments

Development Date Significance
EU AI Act Article 50 enforcement August 2, 2026 EU mandate: AI-generated content must be machine-readably marked
OpenAI: C2PA + SynthID dual-layer May 19, 2026 Metadata + pixel watermark; some evasion resistance
Social platforms: C2PA metadata stripped on upload Ongoing as of 2026 Instagram, X, Facebook, WhatsApp all strip metadata; credentials don't survive social sharing

Sources and further reading

Check any photo — free, no account, no signup

🕵️ Add to Chrome — Free 🦊 Add to Firefox — Free