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.
$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
$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
| 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 |
Vendors publish accuracy figures from their own test sets, so they aren't comparable between tools, and real-world images (compressed, cropped, edited) are harder than any test set. Treat a detector's result as one strong signal, not proof.
Accuracy figures are from tools' own benchmarks unless otherwise noted. Independent benchmarks vary by test methodology.
| 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 |