Non-Custodial Protocol

Sell data without giving it away.

ai.market is a non-custodial B2B marketplace. Your data stays on your infrastructure. Buyers and AI agents find it through search or post a request, and buyers download purchased files directly from the seller.

Buying? Search the catalog or tell us what you need and let providers come to you.

Your datasets become discoverable by AI agents and visible in AI-powered search, so buyers find you, not the other way around.

  • Data stays on your infrastructure.

    You can separately choose to publish a public sample.

  • Direct buyer downloads.

    Purchased files go straight from the seller to the buyer.

  • Found everywhere.

    Listings surface in AI search worldwide, and buyers' agents can transact directly.

How it works

How non-custodial actually works

1

You list. Your files stay on your infrastructure. AIM Data sends opaque inventory automatically and file structure only after you confirm each file. Buyers download purchased bytes directly from your gateway.

2

allAI handles the middle. Questions, negotiation and checkout run through our AI. Buyers get answers fast. You never chase a lead or hand out your contact details.

3

After purchase, buyers download through your gateway's HTTPS door. Stripe moves the money. We deduct 5% and the card processing fee from the seller; the buyer pays the price shown plus any applicable tax.

Your listing is marketing you don't pay for.

Every listing is published in the formats AI assistants and search engines read. Structured metadata, machine-readable manifests, all of it, on every listing, automatically.

So when a buyer anywhere in the world asks their AI for data like yours, your listing is built to be the answer. They don't need to know ai.market exists. They don't need to visit us. Your data shows up where buyers already ask.

Companies pay real money for that kind of reach. Here it comes with the listing. Free.

Products

Two ways to participate

Find data or sell data without moving raw datasets.

Find Data

Search the catalog or post a request

Search listings or describe what you need and let providers respond.

Semantic searchData RequestsLive
Find Data

Sell Data

List from your own infrastructure

Run the open-source AIM Data gateway with Docker. It sends opaque inventory automatically; confirm each file before sharing its structure. Buyers download purchases directly from your gateway.

AIM Data gatewayYour infrastructureYou confirm each file
Sell Data

Buyer Requests

Buyer demand, live now

See what verified buyers need and respond with a relevant listing.

View All Buyer Requests →
Casino and mortgages

**Title: Request for Real US ChatGPT Conversation Data – Mortgages and/or...

**Title: Request for Real US ChatGPT Conversation Data – Mortgages and/or iGaming** WLDM is looking to acquire licensed datasets containing real, historical AI chatbot conversations from US-based users, specifically covering two categories: 1. **Mortgages:** Mortgage rates, lenders, refinancing, home loans, mortgage comparisons, and related financial questions. 2. **iGaming / Online Casinos:** Casino recommendations, platform comparisons, bonuses, licensing, gambling regulations by US state, and related questions. **Providers may offer datasets for either category individually or both categories together.** We are open to purchasing one category or a combined dataset, depending on availability, quality, and pricing. ### Data requirements - Genuine user conversations, preferably from OpenAI's official ChatGPT platform. - US-based users and English-language conversations. - Full multi-turn conversations, including original user prompts, AI responses, and follow-up messages. - Original historical responses, not regenerated answers or synthetic conversations. - Conversation timestamps, session IDs, AI platform and model identifiers. - Citations, source URLs, and background web searches where available. - Clear provenance, lawful collection practices, and documented user consent. ### Volume and delivery We are initially interested in several thousand conversations, with the possibility of larger volumes or ongoing deliveries. Preferred formats: CSV, JSONL, or Parquet, with secure file delivery or S3 transfer where available. ### Licensing requirements The licence should permit commercial research, publication of aggregated findings and appropriately anonymized excerpts, conference presentations, client deliverables, and retention of derived research results. ### Information requested from providers Please include: - Which categories you can supply: mortgages, iGaming, or both. - Number of qualifying US conversations available per category. - Original AI platform and exact model used, particularly whether conversations originated directly from ChatGPT or third-party applications using GPT models. - Collection date ranges and available metadata. - A representative sample and data dictionary. - Pricing for one-time purchases and ongoing access, including any minimum order requirements. - Applicable licensing terms and restrictions. We welcome proposals from direct data providers, authorized resellers, and data brokers capable of sourcing these datasets.

Review request →

retail

Seeking 10,000 entirely synthetic retail transaction records for testing a...

(SAMPLE)

Seeking 10,000 entirely synthetic retail transaction records for testing a sales analytics dashboard. Include fictional product IDs, product categories, transaction dates, quantities, unit prices, and store regions in CSV format. All records must be artificially generated, with no real people, businesses, payment details, or contact information. Sellers should describe the generation method and usage rights.

Review request →

financeretail

Seeking a fully synthetic retail sales dataset for testing a...

(SAMPLE)

Seeking a fully synthetic retail sales dataset for testing a demand-forecasting demo. Need 12 months of daily transactions across 20 fictional stores and 100 fictional products, with date, synthetic store ID, synthetic product ID, units sold, unit price, discount, and stock availability. CSV preferred. All records must be generated, with no real people, businesses, customer identifiers, or contact details. Please describe the generation method and permitted usage.

Review request →

How it works

Three steps to get discovered

1

Add a gateway.

In Sell Data, open Gateways for a one-time pairing code and Docker Compose file.

2

Publish your listing.

Confirm each file before sharing its structure. Manage the listing, price, and licence on ai.market.

3

Get paid.

Buyers find your listing wherever they search, on ai.market or through their AI assistant. You're paid through Stripe 48 hours after the buyer confirms delivery. Sellers on international payout (to a bank account or in stablecoin) are paid at least 30 days after the buyer confirms delivery.

Ready to get started?

Buyers search free. No account needed to look. Sellers list free and pay nothing until a sale clears.

Looking for something else? Check our Partner Program, read The Protocol, or browse the Blog.