AI Provenance Metadata by Platform (2026)
TL;DR
AI provenance is a model-level feature, not a platform choice. Google's Nano Banana family embeds IPTC “Made with Google AI” metadata[1]in every output. Four platforms are affected: Flora, Lovart, Virtuall, and Fotor — all via Nano Banana 2. Most other models and platforms produce untagged outputs.
Recommended Benchmarks
- AI Platform Content Rights (2026)Figma Weave scores 5/5 with full commercial rights, no training on user data, and enterprise indemnity. Most platforms reserve training rights.
- AI Platform NSFW Moderation (2026)Flora and VEED lead with intelligent moderation (4/5). SeaArt silently blocks prompts with zero feedback. We tested 13 platforms with one prompt.
- Best Creative AI Platform 2026: 14 RankedFotor and Flora tie at 3.85/5 in our 14-platform benchmark. Full rankings with trust scores and segment breakdowns for every use case.
- Best Creative AI Platform for E-Commerce (2026)Fotor leads e-commerce at 4.2/5 with a dedicated e-commerce suite — Virtual Model, Product Shot, AI Marketing Video, batch BG removal, and 100K+ templates from £2.91/mo.
Which Platforms Produce Tagged Images?
4 platforms produce IPTC-tagged outputs as of April 2026: Flora, Lovart, Virtuall, and Fotor — all via Google's Nano Banana 2 model. Provenance is model-level, not platform-level: when a platform uses a Google model, outputs carry the “Made with Google AI” tag regardless of the platform's own policies.
| Platform | Google Model Used | IPTC Tagged |
|---|---|---|
| Flora | Nano Banana 2 | Yes |
| Lovart | Nano Banana 2 | Yes |
| Virtuall | Nano Banana 2 | Yes |
| Fotor | Nano Banana 2 | Yes |
| All other platforms | Non-Google models | No |
What the Metadata Contains
Google's IPTC metadata embeds a “Made with Google AI” tag in the image file's EXIF/IPTC data as of April 2026. The tag persists across common workflows and is detectable by automated tools:
- • Survives file conversions — PNG to JPG, resizing, and compression do not strip the tag
- • Detectable by platforms — content verification tools and social media (Facebook, X/Twitter) can flag tagged images
- • No prompt data — does not contain the specific prompt, generation parameters, or seed values
- • Model-level only — indicates AI generation but not which platform (Flora, Fotor, etc.) was used
Why This Matters: Three Perspectives
Regulatory Compliance
The EU AI Act and emerging US state laws require disclosure of AI-generated content in advertising, political content, and news as of April 2026. IPTC provenance metadata from Nano Banana 2 provides automatic compliance. Platforms using non-Google models (WaveSpeed, Ideogram, OpenArt) produce untagged outputs — requiring manual disclosure processes.
Content Authenticity
Media organizations and social platforms (Facebook, X/Twitter) increasingly flag AI-generated content using metadata. Images from Flora, Lovart, Virtuall, and Fotor will be automatically flagged. Images from WaveSpeed, Ideogram, Picsart, and others will not — creating an uneven transparency landscape across the industry.
Commercial Safety
Provenance metadata is a double-edged sword for e-commerce. It proves compliance with disclosure requirements, but it also makes AI-generated product photos and marketing assets detectable by competitors and verification tools. Commercial users who want untagged outputs should use platforms with non-Google models (WaveSpeed at $0.07/gen, Ideogram, OpenArt).
The Broader Provenance Landscape
Google is the only model provider consistently embedding provenance metadata as of April 2026. Other major providers — FLUX, Ideogram, Seedream — do not embed IPTC tags. This means provenance is voluntary and inconsistent: 4 platforms produce tagged outputs while 10+ platforms produce untagged outputs.
The C2PA (Coalition for Content Provenance and Authenticity)[3] standard is gaining adoption, but as of April 2026, no consumer AI image platform has implemented it. Google's IPTC approach remains the only production-level provenance system in the market. C2PA limitation: no consumer platform support yet; adoption depends on browser and operating system integration.
Our Recommendation
If you need provenance for compliance: Use Flora (free tier available), Fotor (£2.91/mo), Lovart, or Virtuall — all use Nano Banana 2 and embed automatic IPTC proof of AI generation as of April 2026.
If you want untagged outputs: Use WaveSpeed ($0.07/gen, API-only), Ideogram (2 proprietary models), OpenArt, or any non-Google-model platform. Be aware that the EU AI Act and emerging US state regulations may require manual disclosure even without embedded metadata.
Sources & References
All external sources were verified as of April 2026. Ratings and metrics reflect the most recent data available at time of review.
- Google - AI-Generated Content Metadata (SynthID)(deepmind.google)
- IPTC - AI Metadata Standards(iptc.org)
- C2PA - Content Authenticity Initiative(c2pa.org)
- Flora AI - Privacy Policy(flora.ai)
- Fotor - Privacy Policy(fotor.com)
Related Vibedex Benchmarks
AI Coding Tool Pricing: Type A vs Type B (2026)
Bolt burns 100k tokens per prompt; Replit hit $1,000 a week. We split AI coding tool pricing into Type A (structural) vs Type B (usage) so you can budget.
Deep DiveZapier vs n8n 2026: Breadth vs Self-Host Freedom
Zapier: 8,000+ integrations, Copilot for SMB ops. n8n: free self-host, Code node, dev-native escape hatches — and 4 critical 2026 CVEs. Which one breaks your ops first?
Deep DiveWorkflow Automation Security Compared (2026)
n8n shipped 4 critical RCEs in Q1 2026. Make ran a $12K-loss outage. Codewords has no independent audit. 6 platforms compared on CVEs, SOC 2, and self-host.
Methodology: Rankings and scores in this article are based on VibeDex's independent benchmarks. Models are evaluated by AI-powered judges across multiple quality dimensions with scores weighted by prompt intent. See our full methodology
FAQ
Which AI platforms embed provenance metadata?
This is model-level, not platform-level. Any platform using Google's Nano Banana family of models (Nano Banana, Nano Banana 2, Nano Banana Pro) embeds IPTC "Made with Google AI" metadata. Affected platforms: Flora, Lovart, Virtuall, and Fotor.
Can I remove AI provenance metadata from images?
Technically yes — re-saving as a new file, taking a screenshot, or using certain export tools can strip IPTC metadata. However, doing so may violate platform terms of service and emerging regulations around AI content disclosure.
Why does AI provenance metadata matter?
Regulatory compliance (EU AI Act, state-level US laws), content authenticity for media organizations, and commercial liability. As regulations tighten, provenance metadata becomes a compliance requirement rather than an optional feature.
Find the best model for your prompt
VibeDex analyzes your prompt and recommends the best AI image model based on what your specific image demands.
Try VibeDex →