The Undress AI Index A data desk on the synthetic-intimate-imagery market

Home › FAQ

Reader queries · November 2026

The questions, answered in numbers

Every answer here leads with a figure — because in this market the numbers are the honest version.

By Henrik Dahl, data editor · Last updated: November 2026

Q1 · DefinitionWhat is undress ai, in one sentence?

A category of services that use diffusion inpainting to synthesize a fake nude body into a clothed photograph of a real person — generation, not revelation: nothing underneath was ever captured. The three-stage pipeline is quantified on the methods page.

Q2 · LegalityIs undress ai legal in 2026?

Distribution of nonconsensual intimate synthetics is criminal under 1 US federal statute plus ~48 states/DC, the UK, most EU states and Australia — roughly 58 jurisdictions in our cumulative count. Creation is criminal in the UK since 2025 and a growing minority of US states. Depictions of minors are CSAM everywhere. Full table: legal risk.

Q3 · Market sizeHow many undress ai services exist?

WHOIS sampling shows 40+ distinct undress-ai-branded domains since 2023, with a median brand lifespan of about 4 months. Counting unbranded churn domains, the operator set is materially larger — branded counting understates the market by design. See Fig. 1 in the data briefing.

Q4 · PricingWhat does undress ai output cost — and cost to produce?

Consumer pricing runs $0.10–$5.00 per image; operator GPU cost is roughly $0.02–$0.10. That 10×–250× gross margin — not technical differentiation — is what the branded sites sell. Free tiers monetize the upload itself. Figures in the data briefing, Table 1.

Q5 · PrivacyDo undress ai sites really delete uploads?

About 85% of sampled services claim auto-deletion; 0% offered any audit or proof. With domains rotating quarterly and documented archive breaches in the category, the working assumption should be permanent retention — treat every upload as public-record-adjacent.

Q6 · DetectionHow often can undress ai output be detected?

Forensic evaluations find diagnostic artifacts in the majority of samples: mask-boundary seams in ~70–80%, missing skin detail in ~60–70%, provenance stripping in 95%+. Expert review catches most output — the problem is that distribution outpaces review. Artifact table: methods, Fig. 3.

Q7 · RemovalHow fast can a victim get images removed?

48 hours is the statutory ceiling for TAKE IT DOWN requests on covered platforms; StopNCII hash-matching propagates across partner platforms in roughly 24–72 hours; NCMEC's Take It Down covers depictions made during minority. Acting within the first 48–72 hours measurably improves suppression outcomes — full scoring in legal risk, Table 2.

Q8 · InputsHow much source material does undress ai need?

Exactly one photograph. Inpainting pipelines need only the victim's clothed image; identity-injection siblings need one to a handful. Quality degrades only below roughly 512px inputs — standard social-media resolution sits well inside the viable band.

Q9 · User riskWhat is the risk to someone who uploads a photo?

Three identification vectors are documented — payment records, platform/IP logs, breach dumps — and each has surfaced in real investigations. Add criminal exposure tracking the act itself (creation is now an offense in the UK and a growing minority of US states) plus possession as evidence. "Anonymous" survives contact with none of the three vectors. See measured.

Q10 · ContextWhat share of synthetic intimate content is undress ai vs. face swaps?

Intimate imagery of all kinds constitutes ~96% of deepfake video online (2019–24 study synthesis). Within the image segment, undress-style inpainting and face-swap composites split the majority, with undressing output the faster-growing share since 2023 as single-photo pipelines spread.

The full dataset: market figures in the data briefing, pipeline measurements in methods, statute tables in legal risk, archetype scores in measured, and sourcing on the methodology page.