# ERC-7007: Verifiable AI-Generated Content Token

Source: https://www.eipsfordesigners.com/standards/ERC-7007
Agent brief: https://www.eipsfordesigners.com/standards/ERC-7007/agent.md
Machine-readable JSON: https://www.eipsfordesigners.com/api/standards/ERC-7007
Last reviewed: 2026-05-25
Last updated: 2026-05-25

| Field | Value |
| --- | --- |
| Status | Draft |
| Chain | both |
| Category | Security & Trust |
| Journey stages | Specialized Interactions |
| Detailed guide | Yes |
| Official specification | https://eips.ethereum.org/EIPS/erc-7007 |
| Discussion search | https://ethereum-magicians.org/search?q=ERC-7007 |

## UX Impact

AI-generated NFTs include cryptographic proof linking output to specific model and prompt — verifiable provenance for AI art. Design implications: show 'AI-Generated' label with model info, display original prompt, add 'Verify Proof' button. Design decisions: how prominent to make AI disclosure — regulatory requirements vs aesthetic concerns, whether to filter/highlight AI content separately.

## Summary

AI-generated NFTs include cryptographic proof linking output to specific model and prompt — verifiable provenance for AI art.

## For Designers

- You can show AI-Generated labels with model name and a Verify proof action on every eligible item.
- Your detail pages can display the original prompt behind an expandable section for collectors who care.
- You can offer marketplace filters for AI vs human-made without shaming either category.

## Applicability

### When to Use

- Platform sells or displays AI-generated NFTs.
- Regulatory or community norms require AI disclosure.
- Provenance verification is a collector feature.

### When to Avoid

- Collection is exclusively human-made with no AI pipeline.
- Proof infrastructure is not wired to your indexer.
- Draft spec risk is unacceptable for production claims.

## Problems It Solves

### Fake human-made claims for AI art

Impact: high

Old way: Metadata says artist with no proof

New way: Verifiable link from token to model and prompt

### Regulatory AI disclosure gaps

Impact: high

Old way: Platforms guess from metadata

New way: Standardized on-chain provenance fields

### Collectors cannot audit origin

Impact: medium

Old way: Trust the marketplace description

New way: One-click verify proof on detail page

## Anti-Patterns

### AI label without verify path

Severity: critical

Unverifiable claims are worthless

Instead: Show Verify proof or omit verified badge

### Hiding AI status in metadata only

Severity: high

Buyers never see disclosure before purchase

Instead: Thumbnail chip visible in browse and checkout

### Scare-banner overlay on all AI art

Severity: medium

Alienates legitimate AI creators

Instead: Neutral AI-Generated chip with optional detail

## Design Decisions

### How prominent should AI labels be?

Recommendation: Chip on thumbnail plus detail panel; avoid full-overlay warnings on art.

Rationale: Disclosure must be visible without vandalizing the artwork.

### Show prompt by default?

Recommendation: Collapsed behind View prompt; never auto-expand on gallery cards.

Rationale: Prompts can be long, personal, or spoil the piece.

### Filter AI content separately?

Recommendation: Offer optional filter and sort, default to show all with labels.

Rationale: Hiding by default angers creators; unlabeled browsing hides disclosure.

## States to Design

### Verified AI provenance

Trigger: Proof checks pass.

User need: Trust the AI label.

Design response: Verified badge with model info and optional prompt preview.

### Unverified or missing proof

Trigger: Token lacks valid proof.

User need: Not be misled about origin.

Design response: No AI badge; optional Unknown provenance warning for claimed AI pieces.

### Proof verification failed

Trigger: Hash or model mismatch.

User need: Understand label is unreliable.

Design response: Verification failed banner; do not show verified AI chip.

### Prompt hidden by creator

Trigger: Proof valid but prompt redacted.

User need: Know proof exists without seeing prompt.

Design response: Verified with prompt private note.

### Regulatory filter active

Trigger: User filters to human-only art.

User need: Clean results without AI items.

Design response: Filter excludes unverified and verified AI; show count removed.

## Vocabulary

- Use "AI-generated" instead of "Synthetic token mint": Plain disclosure language.

- Use "Verify proof" instead of "Validate zk attestation": User action, not cryptography lecture.

- Use "Created with [model]" instead of "Inference pipeline ID": Human-readable model naming.

## What to Prototype First

### AI provenance detail panel

Collectors decide authenticity at the item page.

Covers: Verified proof, Missing proof, Tampered proof

- Model name
- Prompt expander
- Verify button
- Proof status chip

### Gallery with AI filter

Browsers need to discover or avoid AI content intentionally.

Covers: Filter on, Filter off, Mixed grid

- AI-Generated chip
- Filter toggle
- Empty filter state

### Mint disclosure step

Creators must attest AI use before listing.

Covers: AI mint, Non-AI mint

- Disclosure checkbox
- Model selector
- Preview label

### Verify proof result

Verification must feel definitive, not technical.

Covers: Valid, Invalid, Unavailable

- Green verified or red failed
- Plain explanation
- Link to spec

## Mental Model

### Generation record

Model + prompt + parameters produce an output that gets hashed and linked to the token.

### On-chain proof

The token carries a verifiable claim that this output matches that generation record.

### Verification

Anyone can re-check the proof — trust moves from brand promise to math.

### Disclosure

AI-Generated is a transparency label, not a quality judgment.

### Creator attestation

Minter submits proof at mint; marketplaces inherit the label from chain data.

## Seen in the Wild

- OpenSea: Content credentials and AI disclosure patterns emerging on major marketplaces. (https://opensea.io/)

- Foundation: Creator-centric provenance display sets bar for authenticity UI. (https://foundation.app/)

- Art Blocks: Generative art platforms already expose algorithm provenance to collectors. (https://www.artblocks.io/)

- C2PA: Cross-industry content credentials inform AI disclosure UX. (https://c2pa.org/)

## On Monad

### Verification cost

Ethereum: On-chain verify may be gas-heavy

Monad: Lower fees enable verify-on-view in gallery

Design implication: Offer inline verify without discouraging clicks.

### AI mint throughput

Ethereum: Batch AI drops congested

Monad: Higher throughput supports large generative drops

Design implication: Design for rapid mint status updates during AI launches.

## UX Patterns

### AI Provenance Panel

Model, proof status, and verify action on NFT detail.

Components: AIChip, ModelRow, VerifyButton, PromptExpander

User flow:

- User opens NFT
- Sees AI-Generated
- Taps Verify
- Result shown

Mockup registry key: `concept/agent-task` (React UI on the live standard page).

### AI Content Filter

Gallery toggle for AI-labeled items.

Components: FilterToggle, LabeledThumbnail, EmptyState

User flow:

- User opens gallery
- Toggles AI filter
- Grid updates
- Labels remain on items

Mockup registry key: `concept/nft-gallery` (React UI on the live standard page).

## Related Standards

- ERC-721: NFT base extended with AI provenance fields

- ERC-4906: Metadata updates when proof is amended

## Technical Notes

ERC-7007 is draft; pair visible disclosure with working verify infrastructure before production claims.