Sell Data to AI
Home Log in AI Data Value Score Pricing API documentation MCP server for AI assistants 26 US company lists Niche lists
For brokers
The Sell Data to AI Rules 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? Anonymize before selling Score, then make it safe to sell FAQ and glossary
Score any company Check your own company
Website check for companies

Check your company: does it qualify to sell data to AI?

Last updated: 9 October 2026

For companies, not individuals. Enter your company website. You get its AI Data Value Score from 0 to 100 and the data types we identify, then the check reads a few of your public pages, finds the facts AI data buyers screen on, shows each one with the sentence and the page it came from, and compares them with the rules micro1, Mode and Grepped publish. It does not price your data. Nobody can do that without seeing it.

0 to 100your AI Data Value Score, with the data types we identify
8 signalseach shown with the quote it came from
3 programsmicro1, Mode and Grepped, as published
Freeabout 20 seconds, no login, no email, a few checks a day

Rather answer a few questions? Use the eligibility checker. It runs in your browser.

The check

Enter your company website

Use the main domain of the company that owns the records, such as example.com. The result lists what we found, quoted from your pages, then a likely, possible or unlikely fit for each program, with an apply button for each.

We score your company with our AI Data Value Score, detect the data types you hold, and compare your public pages with the rules the data buyers publish. About 20 seconds. Free, no email, 5 checks a day.

Checking another company, not your own? Use the AI Data Value Score demo.
Reading your public pages
Finding team size, location, history and systems, with quotes
Scoring your data and detecting your data types
Comparing with each buyer's published rules

Independent site. Some links are referral links: if your company signs with a buyer through them, the buyer may pay us a fee. You are not charged, and we never see your data. Disclosure.

Main domain only

Enter the operating company's own site. A product microsite or a holding group's page rarely says how big the team is or how the business runs.

Kept for 7 days

A site checked in the last week returns the stored result. Checking a site you already checked today does not use up one of your daily checks.

Site says little?

If your pages do not state your size or systems, the eligibility checker asks seven questions and sends nothing anywhere.

New: anonymization

Before you sell: remove the personal data

A high score means your company holds data AI labs want. Before any of it leaves, our anonymization service removes names, contact details, ID numbers and secrets, across 150+ data types.

Use it on its own, or with the score and lists: score, then clean the data.

The data AI labs pay most for, identified for every company

Not all company data is worth the same. Our research shows AI labs pay most for the records of real work, held in the systems below. We identify which of these systems each company runs, and it is one of the most important factors in the AI Data Value Score.

Work tickets and internal wikiJira, Confluence
Customer support ticketsZendesk, Intercom
CRM recordsSalesforce, HubSpot
Calls and meetingsZoom, Gong
Contracts and documentsDocuSign, Box
Design filesFigma, Miro
Finance and ERP recordsNetSuite, QuickBooks
Code with historyGitHub, GitLab
Work email and documentsMicrosoft 365, Google Workspace
HR and recruiting recordsWorkday, Greenhouse
Training and course contentLearning platforms, course libraries
Fleet and sensor dataFleet and machine data
Clinical and lab systemsPatient and lab records
Human-era archivesYears of records written before 2023

Every company comes with the systems we identified, shown as tags. Company level only: we show which systems a company runs, never what is inside them.

How the check works

Four steps, about 20 seconds

The check does one narrow job well: it turns what your own website says into the facts a buyer program screens on, and shows you the evidence for each one.

You enter a domain

We reduce it to the bare domain and confirm it looks like a real website address. That is the only thing we ask for. No account, no name, no email.

We read a few public pages

We read a few of your public pages, once per check. Only text that any visitor can open is read: nothing behind a login, and nothing from your internal systems.

Facts come back with quotes

Every fact is shown with the sentence and the page it came from. If your pages do not state a fact, the card says so in plain words rather than filling the gap with a guess.

Facts meet published rules

Each program's published rules decide a likely, possible or unlikely fit. The reasons are written out in plain words, next to the program's published range and an apply button.

Why every fact carries a quote. A result you cannot check for yourself is worse than no result. Each signal on the card shows the sentence it came from and the page it came from, so you can see at a glance whether the check read your site correctly. When your pages do not state a team size, no number is shown at all: the card says "not stated on the pages we read" instead of guessing.
What the check reads

Eight signals, and the published rule each one meets

Each signal maps to something a buyer publishes, or to something that decides what you could offer at all. Rules are quoted from each program's own pages, as published, checked 7 October 2026.

Team size

What we look for
A stated headcount.
Published rule
micro1 lists 30+ employees (its referral posting says 30 to 200). Mode lists 20+ full-time US office employees, 10+ for accounting firms and 6+ for law firms.
Why buyers care
Headcount is the simplest proxy a buyer has for how much connected work a company produces. It is also the most common reason a program is out of reach.

Location and US presence

What we look for
Where your pages say the company is based, and whether it has a presence in the US. The check also notes the main language of your pages.
Published rule
micro1 prioritizes the US, then other Western markets, and lists primarily English. Mode counts US office employees and names US-based teams as the strongest fit.
Why buyers care
Location decides which programs look at you first, and which privacy laws cover your records. An EU or UK team brings GDPR into scope.

Operating history

What we look for
The year your pages say the company was founded.
Published rule
Mode asks for several years of records the company owns. micro1 asks for mature operations.
Why buyers care
Buyers want to see how work played out over time: a project from kickoff to delivery, a client across several renewals. A company founded less than two years ago shows as unlikely for Mode.

Industry

What we look for
What your pages say the company does, summed up as an industry.
Published rule
Mode's minimum drops to 10+ for accounting firms and 6+ for law firms. Grepped lists any vertical.
Why buyers care
Industry changes the headcount bar, and it changes how much of your record is off limits for confidentiality reasons.

Systems you say you use

What we look for
Business software your pages say the company uses for its own work.
Published rule
micro1 asks for modern software tools. Buyers name the systems they handle, from Slack and Microsoft Teams to Salesforce, QuickBooks, Zendesk and GitHub.
Why buyers care
A named system tells a buyer how the records would be exported, and connected systems show more of a workflow than any single tool.

Documentation

What we look for
Signs on your pages that the company keeps written procedures, such as a help center, a knowledge base or documented processes.
Published rule
micro1 lists documented processes, and asks for SOPs, knowledge bases and internal documentation.
Why buyers care
Written procedures are among the records buyers name first, and they are usually the easiest category to scope and de-identify.

Ownership

What we look for
Whether the company presents itself as independent, investor or PE backed, part of a larger group, or a public company.
Published rule
Mode asks for records the company owns.
Why buyers care
Ownership decides who can sign. If a parent company owns the systems or sets the data policy, the decision likely sits with the parent, and the result says so.

Client-confidential data

What we look for
Whether your work involves client secrets or regulated records.
Published rule
No program publishes a rule on it, so it never changes a fit result.
Why buyers care
It changes what you can offer. Client files under NDA, privileged matters and patient records are often excluded, which can shrink the sellable part of your records more than any other factor.
Reading your result

Likely, possible, unlikely, and "not stated"

Every program card on the result carries one of three labels. Every signal can also read "not stated on the pages we read". Here is what each one means, and what it does not mean.

Likely fit

Your pages meet the rules

The facts on your pages meet every published rule the check can test, such as a stated team of 45 with a US office against micro1's 30+ and its US priority. A reason to apply, not an acceptance.

Possible fit

A deciding fact is open

A deciding fact is not stated, sits at the edge of a rule, or the program publishes no minimum at all. Grepped is always possible for that reason. Read the reasons, then confirm in the program's own application.

Unlikely

A stated fact misses a rule

Your pages state something that misses a published rule: a team of 12 against micro1's 30+, a team with no US presence for Mode, or a company founded less than two years ago.

Not stated

Your pages are silent

We found no quote for this fact. It is not a negative. It means your public pages do not say, which is common, and the program's result stays at possible on that point.

What decides each program's label

The check applies the same published rules as the eligibility checker. The difference is the input: here, a rule can only be met or missed by a fact your pages state. Where they say nothing, the rule stays open.

ProgramPublished ruleLikely when your pages showUnlikely when your pages showPublished range
micro130+ employees (referral posting: 30 to 200); mature operations, documented processes, modern software tools; primarily English; US prioritized, then other Western marketsA stated team of 30 or more, on an English site. No visible US presence turns likely into possible.A stated team under 30, or a site that is not in English and does not state 30+. Over 200 people is noted, not penalized."$100K-$2M+ for approved data packages"
Mode20+ full-time US office employees; accounting firms 10+; law firms 6+; several years of records the company owns; US-based teams strongest fitA stated team at or above your industry's minimum, and a visible US presenceA stated team below your minimum, pages that show no US presence, or a founding year less than two years ago"$100K-$5M"
GreppedAny verticalNever: no minimum is published, so there is nothing to meetNever, for the same reason"$20K-$5M"

Last updated: 7 October 2026, on each program's own pages. Ranges are what each program publishes, not offers, and the top of a range is not the typical result. Full side-by-side details: buyer programs compared.

When your site says little

Your website does not show your size or systems. That is normal.

Most real companies do not publish their headcount. A company website is written for customers and future hires, not for a data buyer. In our own test runs, a large and well-known US software company came back with team size not stated, because none of the pages we read gave a number.

Systems are even rarer. Few companies say in public which software their own teams work in every day, so this signal is often empty, even for companies that run on a dozen well-known tools. An empty systems line says nothing about how modern your business is.

What to do instead

Answer the questions yourself. The eligibility checker asks seven questions: country, full-time office staff, industry, years of records, systems, language and client-confidential data. It applies the same published rules and runs entirely in your browser, so nothing you enter is sent anywhere.

Keep what the website check did find. The quotes are still useful. They show what a buyer's first look at your site will see, and they can save you typing in the questionnaire.

Do not edit your website to pass. Buyers verify headcount, ownership and records in their own review. A number added to your site for a checker changes nothing in that review, and it can confuse your customers and your future hires.

Facts your site may not show, and where they live

  • Full-time office headcount: your payroll or HR system, counted the way Mode counts, without contractors.
  • Years of records: the oldest data still in your email, chat, ticketing and accounting systems. Not your founding year.
  • Systems in use: your software subscriptions or the admin console your IT lead manages.
  • Who owns the data: your client contracts, and any agreement with a parent company or investors.
  • Client-confidential content: your NDAs and engagement letters.
Answer seven questions

Website check or questionnaire?

Use both. The website check shows what your public pages prove on their own. The questionnaire shows where you stand with the facts only you know.

 Website check (this page)Eligibility checker
What it usesYour public website text, quotedYour own answers to seven questions
HeadcountOnly when your pages state itThe number you enter
Years of recordsFounding year, as a rough proxyYears of records you still hold
SystemsOnly those your pages say you useEvery system you tick from the buyers' list
Programsmicro1, Mode, GreppedThe same three, plus Miro Advisory and direct lab intake for information
What is sentYour domain, to our serverNothing; it runs in your browser
LimitA few new websites a day per visitorNone
Not a buyer's review

How this check differs from a buyer's own review

The check is a first filter on public text. A buyer's review is the real decision, and it looks at things no website shows.

This check

What it reads
A few public pages of your website
How long it takes
About 20 seconds
What it decides
A fit label against published rules, nothing more
What it sees
What your company says about itself in public
Price
None. It shows each program's published range.
Contracts and consent
Not checked

A buyer's review

What it reads
A manifest and samples of your actual records, usually under an NDA
How long it takes
Weeks to months. Practitioners cite 60 to 90 days from inquiry to close.
What it decides
Acceptance, scope and price
What it sees
What your records contain: depth, structure, uniqueness, and how much must be removed for privacy
Price
An offer for your scope, if there is one
Contracts and consent
Exclusivity, scope of use, indemnities, deletion, and what your staff and clients agreed to

The website check sits before the first step. Practitioners' advice for everything after it: share a manifest and samples first, never a full dataset before a price, and get more than one offer.

Worked examples

Three fictional companies, and what the check returns

These companies and their quotes are invented to show how the rules play out. The labels are what the check returns for those pages. No prices appear, because no program publishes one for a given company.

Fictional example

Example A: a 14-person accounting firm in Ohio

What its pages say (invented)

  • "Our team of 14 has served local businesses since 2009."About page
  • "Experience with QuickBooks Online and Xero is required."Careers page

Signals

Team size
At least 14 people
Location
United States
History
Founded 2009
Industry
Accounting
Systems
QuickBooks, Xero
Confidential
High

Result

  • Unlikelymicro1: the site suggests about 14 people; micro1 publishes 30+.
  • LikelyMode: 14 people against 10+ for accounting firms, a US office, about 17 years of history.
  • PossibleGrepped: no published minimums.

Next: Mode is the program to look at before the others. The high client-confidential flag comes before any application, though. Client tax and bookkeeping files sit under engagement letters, so the firm checks what it may include. What remains is often its own operations: close checklists, review procedures and templates.

Fictional example

Example B: a software company of about 120 people in Texas

What its pages say (invented)

  • "We're 120+ people across Austin and remote."Careers page
  • "You'll work in Jira, Slack and GitHub every day."Careers page
  • "Founded in 2015 and backed by growth investors."About page

Signals

Team size
At least 120 people
Location
United States
History
Founded 2015
Industry
Software
Systems
Jira, Slack, GitHub
Docs
Help center
Ownership
Investor backed

Result

  • Likelymicro1: at least 120 people against 30+, US presence visible.
  • LikelyMode: at least 120 against 20+, US presence, about 11 years of history.
  • PossibleGrepped: no published minimums.

Next: two likely fits and one possible, so it can apply to all three and compare offers on terms, not only on price. Before sending anything, it scans its repositories for secrets, checks open-source licenses, asks whether its investors must approve, and plans how to tell staff whose Slack messages are in scope.

Fictional example

Example C: a UK engineering consultancy whose site gives no headcount

What its pages say (invented)

  • "Structural engineering from our Manchester studio since 2001."Home page
  • "We are an independent practice with documented quality procedures."About page

Signals

Team size
Not stated on the pages we read
Location
United Kingdom
History
Founded 2001
Industry
Consulting / professional services
Systems
None named
Docs
Documented procedures

Result

  • Possiblemicro1: team size not stated; micro1 lists other Western markets after the US.
  • UnlikelyMode: no US presence; Mode says US-based teams are the strongest fit.
  • PossibleGrepped: no published minimums.

Next: the check could not see the firm's 38 staff, so the owner runs the eligibility checker with the real number. micro1 stays possible as a Western market outside the US, and Mode stays unlikely. GDPR covers staff and client contacts in the records, so lawful basis comes first.

Fictional companies with invented quotes, for illustration only. Not real firms, not offers, and not a prediction of any buyer's decision. A real result card shows each program's published range, never a price for your company.

Privacy

What the check keeps, and what we never do

The check reads public pages only. It shows you nothing a visitor to your website could not read for themselves, and it keeps almost nothing. Full details are in our privacy policy.

No login, no email

You type a domain. That is the only input. There is no account to create and no contact form behind the result.

No access to your data

We never connect to your email, chat, files or any internal system. We never receive, view or store your records, before or after you apply.

Three things stored

Nothing is stored except the domain, the result and a salted hash of your IP address, used only to enforce the daily limit.

Results reused for 7 days

A result is kept for 7 days so the same website is not read again and again. After that, a new check reads the pages fresh.

No outreach

We never contact the company that was checked, and we never tell a buyer that a check was run or what it found.

Never used for lists

Checked websites are never used for anything else. They are not added to a list, not used for marketing and not shared. A buyer hears from you only if you apply on its own form.

Limits

What the check cannot do

The check is deliberately cautious. It would rather say "not stated" than guess. These are the limits to keep in mind when you read a result.

A few pages, read once

The check reads a few of your public pages, not your whole site. A fact your site states only somewhere else will read as not stated, so add it in the questionnaire.

Sites that cannot be read

Some sites cannot be read automatically; the check then says so and suggests the questionnaire rather than inventing a result.

Team size can read low

A website rarely shows the whole team. If the size on your result looks low, use the questionnaire with your real number.

Founding year is not records

A firm founded in 2005 may hold only three years of email after a system migration or a retention policy. Buyers care about the records you still hold.

Three programs only

The check compares micro1, Mode and Grepped, the programs with apply links here. Miro Advisory and direct lab intake appear in the eligibility checker, for information.

Daily limit

A few checks per visitor per day. When the limit is reached, the questionnaire still works.

7-day cache

Edits to your website show only after the stored result expires. The date and time of the reading are printed under every result.

Rules as of a date

Rules are as published on 7 October 2026. When a program changes its terms, we update the rules and the date, here and on the comparison page.

Nothing about your records

The check cannot see what your records contain, how clean they are, or what your contracts allow. Those decide the outcome, and only a buyer's review sees them.

Common mistakes

Six ways owners misread a result

Each one costs time with a buyer later. All of them are easy to avoid once you know them.

Checking the wrong website

Entering a product site, a brand site or the parent group's domain.

Check the site of the company that owns the records and would sign the deal.

Reading "likely" as accepted

Treating a likely fit as a promise of a deal or a price.

Read it as "your public facts meet the published rules". The buyer still reviews the records and decides.

Reading "not stated" as "no"

Giving up because the check could not find your team size.

Silence is not a miss. Enter the real number in the eligibility checker.

Stretching the numbers

Counting contractors, or listing tools you stopped using, to clear a minimum.

Mode counts full-time US office employees. Give honest numbers in the application; the review checks them anyway.

Skipping the confidentiality line

Ignoring a high client-confidential flag because it does not change the fit label.

Read your client contracts first. See is it legal to sell company data. General information, not legal advice. Talk to your own lawyer before you sign.

Applying before anyone inside agrees

Clicking apply before deciding who signs, what is in scope and how staff will hear about it.

Settle those first. Staff wrote the messages; see employees and selling company data.

Next steps

From a result to an application, in six steps

The check is step one. Here is the order that saves the most time for a company of 10 to 1,000 people.

  1. Run the website check and read every quote. Confirm the check read your site correctly. If a quote is about a product you sell rather than your own company, discount that signal.
  2. Fill the gaps with the questionnaire. Anything marked "not stated" goes into the eligibility checker with your real numbers.
  3. Read the rules side by side. Buyer programs compared shows each program's published payouts, eligibility and process on one page.
  4. Settle the legal and people questions. Ownership, client contracts, employee notice and personal data. Start with is it legal to sell company data. General information, not legal advice. Talk to your own lawyer before you sign.
  5. Prepare a manifest, not an export. List the systems, years and record types you would offer, and what you would never include. Prepare your data for sale walks through it, and what data AI labs want shows which records matter most.
  6. Apply to the likely and possible programs, then compare. Use the apply buttons on your result. Applying to more than one lets you compare offers on terms, not only on price. How to sell data to AI companies covers the full path from review to payment.
FAQ

Questions about the website check

Does the check tell me what my company’s data is worth?

No. Nobody can price your data without seeing it, and this check never sees it. It reads your public website, finds the facts buyers screen on, and compares them with each program’s published rules. The result shows each program’s published payout range, as published, which is a range and not a price for your company.

Why does my result say “not stated on the pages we read”?

Because none of the pages we read stated that fact in a sentence we could show you as its source. That is common: most companies do not publish their headcount or the software they use inside the business. It is not a negative. Use the eligibility checker to enter the real numbers yourself.

What do you read, and what do you keep?

A few of your public pages, only text any visitor can open. Nothing is stored except the domain, the result and a salted hash of your IP address, used only for the daily limit. There is no login, no email and no access to any of your systems.

Will a buyer, or anyone else, be told that I checked my company?

No. We never contact the company that was checked and never pass a result to a buyer. Checked websites are never used for anything else, including lists. A program hears about your company only if you click an apply button and fill in its own form.

I updated my website. Can I run the check again?

Results are cached for 7 days, so a site checked in the last week returns the stored result. Your changes show once that week is over. In the meantime, the eligibility checker gives an answer from your own inputs right away.

Why is Grepped always a “possible fit”?

Grepped lists any vertical and publishes no employee, country or years minimum. With no published rule to meet or miss, the honest result is possible fit: only Grepped’s own review can decide.

The result says “likely fit”. Does that mean we will be accepted?

No. Likely fit means the facts on your public pages meet every published rule the check can test. Each buyer then reviews what your records actually contain, and decides acceptance, scope and price. No buyer promises acceptance, and neither do we.

Our team is outside the US, or our site is not in English. Is the check still useful?

Yes, as a reality check. micro1 prioritizes the US, then other Western markets, and lists primarily English. Mode’s rule counts full-time US office employees. Grepped publishes no country rule. The result names which rule your location or language touches, so you can decide where to apply.

See what your website says about you

Your AI Data Value Score, eight signals with quotes, three programs. Free, no login, and nothing kept beyond the domain, the result and a salted IP hash.