Sell Data to AI
Home Data Asset Score Pricing API documentation US labs and CROs list
For brokers
How to become an AI data broker Data broker business model Buyer programs compared Qualify a company AI training data companies
For data companies
Firmographic data providers Company data API
Seller guides
How to sell data to AI companies Is it legal? FAQ and glossary About
Check domain/company
For would-be brokers and referrers

How to become an AI data broker: the honest economics

Last checked: 7 October 2026

Buyers of company data now publish referral fees in the tens of thousands of dollars, and many people are asking how to start a data broker business around them. Here is how the work is actually paid: only after a deal closes, at the buyer's discretion, months later, and with real legal exposure if you ever touch the data.

$10K to $50KPublished per-referral amounts
60 to 90Days to close, as practitioners cite
$0Paid for introductions that do not close
ClawbacksAllowed under published terms
First, which kind of broker?

"Data broker" means two different businesses

In everyday use, a data broker is a company that collects personal information about consumers and sells profiles or lists. That business is regulated in a growing number of places, and it is not what this page is about.

The newer meaning is an intermediary in AI training data deals: someone who connects a company that has records of real work with a buyer that wants them. Buyer programs such as micro1, Mode and Grepped publish what they pay companies and, separately, what they pay people who introduce those companies. The rest of this page covers that second business, and the full distinction is on what is an AI data broker.

Three models

Three ways to work in the middle, and how exposed each is

The further you move from introducing toward handling data, the more you can earn per deal in theory, and the more legal and financial risk you carry in practice.

Lowest exposure

Referrer

You introduce a company through a buyer program's referral process and step out. The buyer handles review, contract, export and payment. You are paid a published fee only if the company signs and meets the program's conditions. You never see the data.

Medium exposure

Seller-side representative

You advise a company through a sale and negotiate for it, usually paid by the seller under your own agreement. No standard rate is published. You take on duties to your client, conflicts if you also take buyer fees, and often access to samples.

Highest exposure

Buyer and reseller

You license or buy data and resell it. You hold the data, give warranties to your buyers, carry privacy-law and security duties, and sit between two sets of contracts. This is a regulated data business, not a side project.

ModelWho pays youWhenWhat you touch
ReferrerThe buyer program, under its referral termsAfter the company signs and meets milestones; holdbacks may applyAn introduction and the company's consent to it
RepresentativeThe selling company, under your own agreementAs your agreement says; often tied to the closeNegotiations, often samples under NDA
ResellerYour own buyersAfter you have paid for, prepared and delivered dataThe data itself, end to end

What the two higher-exposure models involve in practice

A seller-side representative works for the company. That sounds simple until you list the duties it brings. You are advising on a contract that touches employee messages, customer records and client confidentiality, so the company will rely on what you say about scope, exclusivity and liability. If you also accept a referral fee from a buyer, you are paid by both sides of the same deal. At a minimum that must be disclosed to the company in writing, and some programs' terms require you to disclose relationships to them too. Representatives also tend to see samples under NDA, which means your own security and confidentiality practices are now part of the deal.

A buyer and reseller is in a different business altogether. You sign a licensing agreement with each company as the buyer, then a second agreement with each of your customers as the seller. Every warranty you receive from a company has to be at least as strong as the warranty you give onward, or the gap is your risk. You need somewhere secure to hold data, a way to de-identify it that you can defend, retention and deletion schedules, and the ability to honor deletion requests across copies. Buyer programs that do this work at scale publish detailed privacy statements for a reason. Before considering this model, talk to a lawyer about privacy registration, contracts and insurance where you operate.

Most people asking how to start a data broker business are really asking about the first model. The rest of this page focuses on it, because it is the only one where the terms are published and the exposure is limited to your own conduct.

What programs publish

The published referral amounts, and the conditions behind them

Amounts as published, checked 7 October 2026. Each is paid per qualifying closed deal, not per introduction.

ProgramPublished referral amountPublished condition
micro1"Earn $50,000" per company; 7.5% of the license or purchase price when no amount is listedAfter onboarding plus a minimum revenue threshold (workflow partnerships) or after data is accepted (corpus partnerships)
Mode"Earn $50K per referral" on its site; an X post says "up to $55k or 6% of the transaction, whichever is greater"Read Mode's referral terms
Grepped"Refer for another $10K"Read Grepped's referral terms

Sources: micro1.ai/company-referral and micro1.ai/legal/referral-terms (updated September 24, 2026); data.mode.inc; grepped.ai. Full comparison on data referral programs.

micro1's terms are the most detailed, and they show what any referrer should expect to find in the fine print. A referral only qualifies if the company was not already in micro1's system and goes on to a paid engagement that meets requirements such as a minimum contract value and timeframes. micro1 may reject a referral, or withhold, delay, reduce or set off a payout, in its sole discretion. Payouts may sit in a holdback or clearance period. If the company's arrangement is later cancelled, refunded or unwound, micro1 may ask for the money back.

The same terms set conduct rules. You need the consent of an authorized representative of the company before you refer it. You may not present yourself as micro1's agent, representative or partner, promise outcomes, or share any part of a payout with the company or its people. Payouts are gross, taxes are yours, and a W-9 or W-8 form may be required.

A realistic path

If you still want to start: six steps, in order

This is the referral route, because it is the one with published terms. None of it guarantees a payout, and every step depends on decisions other people make.

  1. Read every program's terms before you introduce anyone. Check who may take part, which countries are accepted, how payouts are made, which tax forms apply and what conduct is forbidden. micro1's terms, for example, exclude current and former micro1 staff and their households and require participants to be at least 18.
  2. Learn the published criteria, so you only introduce companies that fit. As published on 7 October 2026: micro1 lists 30+ employees, primarily English, US first then other Western markets; Mode lists 20+ full-time US office employees, 10+ for accounting firms and 6+ for law firms; Grepped accepts any vertical. Introducing companies that clearly do not fit wastes their time and yours.
  3. Introduce only companies you already know, with consent. The person who agrees must be authorized to speak for the company. Tell them plainly that you may be paid if they sign, that they are not charged, and that the buyer decides everything. Then step out of the conversation.
  4. Disclose any interest. If you own part of the company, work for it or advise it, say so to the program in writing before you refer it. micro1’s terms, for example, require that disclosure in writing and allow it to invalidate pending payouts when its conduct rules are broken.
  5. Keep a dated record and wait. Note when, how and through which link each introduction was made. Then expect months: buyer review, the company's own legal checks, export, de-identification and acceptance all come before any referral milestone.
  6. Plan for tax and for nothing. Payouts are usually gross, so set money aside for tax when one arrives. And plan your finances as if none will arrive, because the timing and the outcome are outside your control.
When money moves

From introduction to payout: why it takes months

Each step below happens between the company and the buyer. A referrer waits through all of them.

1

Introduction

You submit the company, with its consent, through the program's referral process.

2

Buyer review

The buyer checks fit against its criteria. Many companies stop here.

3

NDA and agreement

Scope, price, exclusivity and liability are negotiated. The company's lawyers get involved.

4

Export, de-identification, acceptance

Data is exported, cleaned and accepted, or onboarding starts for ongoing work.

5

Referral milestone and holdback

The program checks its own conditions, such as a revenue threshold, and any clearance period runs.

6

Payout, subject to clawback

Practitioners cite 60 to 90 days just to close a deal. Add the milestone and holdback, and payment takes months, if it comes at all. More on how long an AI data deal takes.

Illustrative, not an offer or a forecast

A year as a referrer, with invented numbers

Every number in this box is hypothetical. We have no data on real conversion rates, and nobody should plan income on these. The point is the mechanics.

Suppose you make 20 introductions in a year to a program that publishes a flat $50,000 per company.

20Introductions made (hypothetical)
14Do not meet size, language or location criteria
4Reviewed, but the company or buyer walks away
2Sign, one of them later unwound

One payout of $50,000 arrives about five months after the introduction. The other deal is unwound after payment and the fee is clawed back under the terms. The hypothetical result is $50,000 gross, before tax, for a year of work, with timing you never controlled. Change two assumptions and the result is zero.

Costs and risks

What the work costs, and where the legal exposure starts

General information, not legal advice. Talk to your own lawyer before you take on any role in a data deal.

Costs of the work

  • Unpaid time on every introduction that does not close.
  • Months of waiting before any payout, with no control over the timeline.
  • Tax and admin: gross payouts, tax forms, payment processor accounts.
  • Reputation: a company that feels pushed, or promised too much, remembers who sent it.

Exposure if you touch data

  • Privacy law: holding or forwarding records about people can bring duties under laws such as GDPR, CCPA/CPRA, HIPAA or GLBA.
  • Security: a lost laptop or misdirected file becomes your incident.
  • Contracts: warranties and indemnities you give buyers can outlast any fee. See indemnities and warranties.
  • Client secrets: companies owe confidentiality to their own clients; you inherit that problem if you carry their files.
The simplest rule: introduce, do not carry. Never collect a company's data "to show the buyer." The buyer runs discovery, export and de-identification under its own agreement with the company. Whether a company can sell at all is covered in is it legal to sell company data.

What a responsible referrer does

  • Reads each program's terms in full before registering.
  • Gets consent from someone authorized at the company.
  • Discloses any relationship with the company.
  • Points the company to the buyer's own published criteria.
  • Keeps dated records of each introduction.

What breaks terms or trust

  • Calling yourself a buyer's partner, agent or representative.
  • Promising amounts, acceptance or timing.
  • Offering the company a share of your fee.
  • Handling, storing or forwarding company data.
  • Referring a company you have an interest in without saying so.

Where to start. Read the programs' own terms before anything else: micro1.ai/legal/referral-terms, the referral section of data.mode.inc, and grepped.ai. Our data referral programs page lists what each publishes, side by side. If you also want to know which companies fit which program, the buyer programs comparison and the eligibility checker use the same published criteria a buyer applies.

For transparency: this site is itself a referrer. It is independent, is not any buyer's partner or agent, never sees company data, and does not share fees.

Common mistakes

Six mistakes new referrers make

Each of these can cost a payout, a relationship or both. Most come from treating a referral like a sale you can close yourself.

Quoting the top of a range

Telling a company it could get "$1M+" or "$5M" because a program publishes that tier sets an expectation nobody can meet. Point to the published range and say that only the buyer can price the data.

Counting introductions as income

An introduction is not a deal. Programs pay per qualifying closed deal, and many introductions will not meet the criteria, will be turned down or will stall on the company's side.

Offering to share the fee

A rebate or cash-back to the company feels generous and can break the program's terms. micro1's terms forbid sharing any part of a payout with the referred company or its people.

Collecting samples "to help"

Asking a company to send you examples of its data so you can pitch it moves you from referrer to someone who holds data, with all the duties that brings. Let the buyer run discovery.

Using a buyer's name as your own

Describing yourself as a buyer's partner, agent or representative is forbidden under published referral terms and misleads the company. Say you are an independent referrer.

Cold outreach at scale

Mass emails to companies you do not know rarely produce the consent and trust a data deal needs. Published referral terms also require consent from an authorized representative before you refer a company. Introductions work best where a relationship already exists.

FAQ

Questions from would-be data brokers

How do I become an AI data broker?

There are three models: a referrer who only introduces companies through a buyer program, a seller-side representative who advises a company through a sale, and a broker who acquires data and resells it. The referral route is the one with published terms and the least legal exposure, because you never handle the data. Start by reading each program's terms in full.

How much do AI data referral programs pay?

As published on 7 October 2026: micro1 states "Earn $50,000" per company, with 7.5% of the license price as the default when no amount is listed; Mode states "Earn $50K per referral" on its site; Grepped states "Refer for another $10K". These are amounts per closed deal, not per introduction.

When does a referrer get paid?

Only after the referred company signs and meets the program milestones, such as onboarding plus a minimum revenue threshold, or acceptance of the data. Practitioners cite 60 to 90 days for a deal to close, and programs may add holdback periods, so payment usually takes months.

Can a referral fee be taken back?

Yes, under some terms. micro1's terms, for example, allow it to withhold or reduce payouts in its discretion and to require repayment if the company's arrangement is cancelled, refunded or unwound.

Do I need a license to be a data broker?

It depends on where you operate and what you do. Some places regulate businesses that sell personal information, and anyone who handles data takes on privacy and contract duties. Ask a lawyer in your jurisdiction. This is general information, not legal advice.

Can I share my referral fee with the company I refer?

Not under micro1's terms, which forbid sharing any part of a payout with the referred company or its people. Read each program's terms; offering cash-back to a company is a common way to lose a payout.