Confidence Dashboard¶
Phase: 3 — Synthesis (meta) Project: likeness Date: 2026-05-09
Overview¶
Where the analysis stands on solid ground vs. thin ice.
The strongest evidence is on regulatory trajectory (Tier 1 legal sources, multiple corroborating analyses) and macro market signals (Tier 2 financial press with multiple confirming sources). The weakest evidence is on creator-specific preferences and willingness-to-pay — there is no primary discovery data, and web search cannot fill that gap.
The founder should interpret the rest of this skill's output as: the macro story is well-supported; the micro story rests on a single load-bearing unknown (creator participation), which only structured creator interviews can resolve.
Claim-Level Confidence Table¶
| Claim | Source Tier | # Corroborating Sources | Confidence | Data Age |
|---|---|---|---|---|
| OnlyFans 2024 gross revenue $7.22B | 2 | 4+ (Variety, Euronews, Yahoo, Hypebeast) | High | Aug 2025 |
| OnlyFans creator share 80%, $5.80B paid to creators 2024 | 2 | 3+ | High | 2025 |
| Fanvue 2025 ARR $100M (+150% YoY) | 2 | 1 (Sacra) | Medium | Late 2025 / early 2026 |
| Fanvue: 15% of revenue is AI-generated content | 2 | 1 (Sacra) | Medium | 2025 |
| Vylit $2.7M Seed Sept 2025; ex-OnlyFans CEO Ami Gan | 2 | 5+ (Fortune, AlleyWatch, BusinessWire, etc.) | High | Sept 2025 |
| Loti AI $16.2M Series A April 2025 | 2 | 4+ | High | Apr 2025 |
| Vermillio AI $16M Series A March 2025 | 2 | 2-3 | High | Mar 2025 |
| TAKE IT DOWN Act signed into law May 19, 2025 | 1 | 5+ (Congress.gov, multiple law firms, Wikipedia) | High | May 2025 |
| California AB 2602 / AB 1836 in effect | 1 | 3+ (Fenwick, official sources) | High | Sept 2024 |
| TN ELVIS Act effective July 1, 2024 | 1 | 4+ (Vanderbilt, Latham, Wikipedia) | High | Jul 2024 |
| Mastercard rules extended to cover AI/synthetic adult content | 1 | 2-3 (Mastercard rules doc, corepay) | High | 2024-2025 |
| Civitai's payment processor cut them off May 2025 | 1 | 1-2 (MIT Tech Review primary) | High | May 2025 |
| 2257 applies to AI of real verified performers (covered) | 1 | 2-3 (legal analyses) | High | Standard |
| Adult AI image generation market $2.8B in 2025 | 3 | 1 (companionguide.ai) | Low-Medium | 2025 |
| Candy.ai revenue: $25M ARR vs. $310M revenue | 2 vs 3 | Conflicting | Low (range only) | 2025-2026 |
| 337+ active AI companion apps July 2025 | 3 | 1 (companionguide.ai) | Medium | Jul 2025 |
| AI companion apps $120M projected 2025 revenue | 2 | 1 (TechCrunch / Sensor Tower) | Medium-High | Aug 2025 |
| Deepfake incidents 500K (2023) → 8M+ (2025) | 2 | 1-2 (identity security researchers) | Medium-High | 2025 |
| 90% of deepfake bounties target women | 1 | 1 (MIT Tech Review primary) | Medium-High | Late 2025 / early 2026 |
| 47.5% negative public sentiment on adult deepfake abuse | 2 | 1 (Springer academic) | Medium-High | 2025 |
| Reddit deepfake sentiment: 47% positive / 36.8% negative overall | 2 | 1 (Springer) | Medium | 2025 |
| OnlyFans 12,000+ deepfake violation account bans | 2/3 | 1 (List25) | Medium | 2025-2026 |
| OnlyFans 59% revenue from one-off (PPV/customs) vs 41% subscriptions | 3 | 1 (SuperCreator industry blog) | Medium | 2023 / 2024 |
| Adult-friendly processor fees 5-15% | 3 | 2-3 (industry guides, no Tier 1) | Medium | Standard |
| C2PA v2.2 published May 2025; Samsung S25 ships C2PA | 1 | 3+ | High | 2025 |
| EU AI Act Article 50 enforcement August 2026 | 1 | 2-3 | High | Pending |
| UK Online Safety Act age verification effective July 25, 2025 | 1 | 3+ | High | Jul 2025 |
| Creator preferences for Likeness's specific positioning | n/a | 0 | N/A — no data | n/a |
| Creator WTP for $15/$25/$50/$100/$200 tier structure | n/a | 0 | N/A — no data | n/a |
| Creator pain hierarchy (defense vs. offense ordering) | Inferred | 0 | Low — inferred only | n/a |
Highest Confidence Findings¶
These are well-supported and can be relied on for downstream decisions: - The regulatory direction of travel (federal + state + UK + EU) strongly favors consent-first AI likeness platforms. - OnlyFans 2024 financials and creator economics benchmarks (multiple Tier 2 corroborating sources). - Vylit's existence, funding, founder, and explicit-content exclusion (multiple Tier 2 corroborating sources). - Loti AI / Vermillio AI funding and mainstream-only positioning. - Civitai's processor cutoff as direct precedent for processor risk. - The general scale of fan demand for AI-generated adult content.
Lowest Confidence Findings¶
These should be treated as directional, not definitive: - The $2.8B adult AI image generation market figure (Tier 3 single-source). - Specific competitor revenue figures (especially Candy.ai's $310M). - AI companion / adult-AI churn rates (mostly Tier 3). - Long-tail of 330+ adult AI generator apps — likely some are credible competitors not yet surfaced.
Critical Unknowns (these matter most for strategy)¶
- Will adult creators with existing audiences actually participate in Likeness's specific configuration?
- This is the load-bearing unknown for the entire business.
- Cannot be filled by web research. Requires structured creator discovery interviews.
-
Filling this is Phase 8 priority #1.
-
Where does creator pain hierarchy actually rank — defense (#1) vs. offense (#2)?
- Affects whether to lead with the takedown/provenance frame or the monetization frame.
-
Same as above — discovery interviews resolve this.
-
What's the actual WTP across the proposed tier structure?
- $15-$200/month tier ladder is reasonable on benchmarks but unvalidated.
-
Test in concierge phase with real fans of real creators.
-
Will multi-processor redundancy actually hold in production?
- All evidence is that adult-friendly processors approve AI-based platforms but with elevated diligence.
- Civitai's cutoff happened despite Civitai having processor relationships.
-
Test by getting actual processor diligence underway pre-launch.
-
What's Fanvue's roadmap re: real-creator AI licensing?
- Unknown.
- Could compress Likeness's competitive window from 18 months to 6.
- Watch for product announcements quarterly.
Recommendations — What to Verify First¶
| Priority | Verification | Method | Owner |
|---|---|---|---|
| 1 | Creator participation hypothesis | 10-15 structured interviews using docs/reviewer-checklist.md |
Founder + Creator Ops cofounder |
| 2 | Payment processor approval posture | Pre-application diligence conversations | CEO (post-hire) |
| 3 | Specific WTP across tier structure | Concierge cohort A/B at $15/$25/$50/$100 entry tiers | Founder + Creator Ops |
| 4 | Mainstream-press / regulatory-tailwind narrative resonance | 3-5 press conversations off the consent-first angle | CEO |
| 5 | Fanvue roadmap signals | Quarterly competitive sweep | Founder |
Source Tier Distribution Summary¶
Approximate breakdown of evidence across this skill's research:
- Tier 1 (analyst reports, government, legal-firm analysis, standards bodies): ~35% — concentrated in regulatory landscape
- Tier 2 (reputable press, Sacra, BusinessWire, Variety): ~50% — concentrated in market sizing and competitive intelligence
- Tier 3 (industry blogs, single-source aggregators): ~15% — concentrated in pricing details and market estimates that lack better sources
The dominance of Tier 1 / Tier 2 sources in the regulatory and market areas is the reason confidence on those topics is rated High. The Tier 3 concentration in adult-AI-specific market sizing is why the $2.8B figure is rated Low-Medium.
Flags¶
Red Flags: - Zero primary creator validation data. This is the single largest evidence gap and the load-bearing risk for the entire project.
Yellow Flags: - Some adult-AI-specific market figures rest on single Tier 3 sources. Treat as directional. - Regulatory enforcement trajectory is well-evidenced for direction but not specific outcomes (NO FAKES Act passage timeline is uncertain).
Sources¶
This document is a meta-synthesis of all research in 01-discovery/raw/. Per-claim citations are in those files.