How to Sell Videos to AI Companies: Prices, Buyers and Rights

AI companies buy video by the minute or by the hour, and most of it reaches them through marketplaces and brokers that license footage on to the labs. The widely reported rate for a creator's unused footage is $1 to $4 a minute, but the one creator who published what he actually earned made $33,000 from licensing a thousand hours of old videos, about $0.55 a minute. Across every public price on record, the median sale is $42 per hour of video, against $95 per hour of video for what buyers offer. Whether you can sell at all is a question of rights: you need to own every frame, everyone recognisable must have agreed, and music, logos and clips you licensed in from others stay theirs.

This guide is for anyone who holds video: a creator with a back catalogue or a drive of unused footage, a production company with a decade of shoots, a broadcaster, a stock or archive library. It covers who buys video for AI training and how each buyer pays, what they pay per minute and per hour (and why realised prices sit below quoted ones), which footage sells, the rights you need for every frame, the platform settings that matter, the choice between a flat fee and a revenue share, and how to approach a sale. Every number links to its source.

AIData.Broker keeps the price board, the buyer board and the video page this guide draws on, and it also acts as the agent of holders who license or sell their data, so we are a party in this market. Check any figure you plan to rely on against its source.

This guide is research from public sources, not legal, tax or investment advice.

Contents

1. Who buys video for AI training
2. What buyers pay per minute and per hour
3. What sells
4. The rights you need in every frame
5. Platform terms that matter
6. Payment models: flat fee or revenue share
7. How to approach a sale
8. Questions people ask
9. What this means for you

Who buys video for AI training

Video models (the systems that turn a text prompt into a clip) and "world models" (systems that learn how scenes move and objects behave) learn from footage. The open web holds plenty of video, but the large platforms forbid taking it: YouTube's own help page says its Terms of Service prohibit "unauthorized downloads and scraping". Footage with clear rights is what a careful lab can buy, and since late 2024 a market has formed to supply it. In early 2025 Bloomberg reported that OpenAI and Google were buying unpublished, unused creator video footage through intermediaries, in deals often worth thousands of dollars.

Most holders never deal with a lab directly. The labs prefer to buy large, cleared, consistently described volumes, so the market runs through intermediaries that gather footage from thousands of holders, check the rights, describe it, and license it on. A few very large libraries, studios and stock agencies, negotiate with labs themselves. The table below lists the buyers and intermediaries that publish how they take video and how they pay; the buyer board and the vendor register keep their current pages.

Protege grew partly by buying Calliope Networks, a leader in aggregating media content for licensing by generative AI companies, in December 2024; Calliope was one of the two intermediaries named in the early reports of labs buying creators' footage. Troveo's more than $20 million paid to the holders of the content it licenses by April 2026 had risen to $50M+ on its own page by October. These are companies' own statements, but they are consistent with the press reports and with each other.

Labs and libraries that deal directly

At the top of the market, the deals are with whole libraries, and they look different. OpenAI licensed Shutterstock's images, videos, music and metadata over the next six years. Shutterstock's chief financial officer told Reuters its deals with large technology companies initially ranged from $25 million to $50 million each, most later expanded. Lionsgate agreed with Runway to train a custom video model on its own catalogue of more than 20,000 titles, and Disney signed a three-year partnership bringing its characters to OpenAI's Sora, with Disney making a $1 billion equity investment in OpenAI: money flowing to the lab, not a licence fee to the studio. These deals show what large libraries can negotiate; they are not a benchmark for a creator's hard drive. The deal ledger keeps every disclosed video licence with its terms.

A separate market: video recorded to order

Robot makers need a different kind of footage: first-person ("egocentric") video of hands doing real tasks, recorded with head-mounted cameras to a specification. That is commissioned work, not an archive sale, and it is paid by the approved hour: micro1 pays $20 per hour of approved video for one recording project, and HUMXN says workers earn $20 per hour of approved footage. Across the collectors' side of the market, the median paid to the people who record is $2.97 per hour of video per hour of video. If you are thinking of recording new footage rather than licensing what you have, the physical-world data page covers that market.

What buyers pay per minute and per hour

Video is priced per minute when creators talk about it and per hour when buyers and archives do. The conversion is simple (multiply a per-minute rate by the minutes in an hour), and it matters, because a rate that sounds small per minute is a large number per hour, and a catalogue total that sounds large can be small per hour. The table separates what was quoted (published rates and reported ranges, set before anyone looked at a particular library) from what was realised (amounts a holder actually received).

The chart puts the main ranges on one axis, per hour of footage. The point it makes is that the realised rates sit at the bottom of the quoted ones: the creator who published his earnings is below every quoted range, and the highest quotes are for premium, described, high-resolution professional material, not for a typical creator's drive.

Sources: footage used, per hour, creator catalogue, realised, footage via Troveo, unused creator footage, creator footage, quoted, premium archive, longer films, Adobe's average, HD content through Calliope.

The market rates, from every record

Individual quotes are anchors. The rates page computes the market's own figures from every public price per hour of video at once. Today the median sale is $42 per hour of video, the median offer to holders $95 per hour of video, and the median asking price for a ready-made dataset off the shelf $37 per hour of video; the middle half of all public prices per hour of video runs $15 to $99 per hour of video. For comparison, the people paid to record footage to order receive a median of $2.97 per hour of video per hour of video.

Ready-made datasets listed for sale give a sense of what finished, described collections ask. A talking-head collection of 500+ hours is listed at $35,800, and a full library of 4K driving footage of 1,750+ hours at $150,000. Those are asking prices, not sales, and both are specialised, described collections, which is why they sit above general creator footage per hour. The lot index lists 32 lots of video on offer in the public catalogues it reads, each with its stated terms.

Why realised sits below quoted

The gap has the same causes everywhere in this market. Volume of general footage: generic footage is plentiful, with one marketplace alone holding licences to 7M+ hours, so a buyer choosing between catalogues pays the general rate and keeps the premium for what is scarce. Premiums are for specific clips: the top of the quoted range is for 4K, drone and 3D work and for well-described archives, not for a whole drive. Buyers pay for what they accept: footage that fails a quality or rights check is not paid for. The one creator who published his earnings came in at 55% of the bottom of the quoted range and 14% of its top.

Large totals come from large catalogues, not high rates. A creator manager told Semafor he had negotiated about five deals, each worth tens of thousands of dollars, for terabytes of video, and Troveo's chief executive said it had paid a handful of creators more than $1 million each. Libraries with professional material report larger sums: Protege says a regional sports rights distributor made $250,000+ in year one from archival footage and a broadcaster over $1M in 6 months from its scripted and cinematic library (its own case studies).

What the valuation uses

The valuation on this site prices general footage at $33 to $120 an hour of footage and high-specification footage (4K, drone, cinematic) at $60 to $240 an hour of footage, from the realised creator rate up to the standard quoted rate, and discounts footage where the people in it have not signed releases. On its starting answers, 400 hours of general footage in which nobody appears or everyone has signed a release comes out at $13,000 to $48,000. Like every range here, it is an indication: only a buyer who has seen a sample can price your footage.

What sells

Buyers do not pay for video in general; they pay for footage that teaches a model something it cannot get elsewhere, that they are sure they are allowed to use, and that comes described well enough to find what they need inside it. The holders who earn the most per hour are the ones who can say yes to all three. The rest of this section goes through what the evidence says buyers look for, and what they pass over.

The single clearest signal is in what the labs went after first. When OpenAI and Google began paying creators in late 2024, what they bought was unpublished, unused creator video footage: the raw takes, the b-roll, the cut scenes, not the finished videos already sitting on public platforms. Footage nobody else has is worth more to a buyer than footage its competitors can also reach, which is the logic behind every premium below.

Five things that raise the price of footage

Resolution and craft. The early reports noted premium rates for high-quality 4K footage, drone video, and 3D animation. Troveo's page for footage holders asks for high resolution ("1080p or 4K preferred"), clear audio and well-framed, clear subjects, in its own words on that page.

Description. An hour of video a buyer can search is worth more than an hour it has to watch. Veritone's chief executive said the broker can charge anywhere from $1 to $1,000 per hour of video depending on the application and the available tagging and metadata, and a MIPCOM 2025 panel cited AI platforms paying $60-$300 per hour for premium, metadata-rich content. Tagging and metadata, not the footage alone, decide where an hour lands in that range: what is in each shot, where and when it was filmed, which camera and settings, who appears and whether they signed.

Specific subjects. Buyers commission or pick what their models lack. Adobe paid contributors for videos depicting everyday actions and emotions. Troveo's categories include scripted, documentary and reality, wildlife and aerial or drone footage. Finished collections that list for the most per hour are narrow and specific: talking heads for lip-sync models, 4K driving footage, first-person task video for robots.

Length and coherence. Short clips are priced per clip and longer films per hour: Defined.ai quoted $2 to $4 for short videos against $100 to $300 an hour for longer films. A long continuous take, or a full scripted programme, shows how a scene develops over time, which is what video and world models need to learn.

Clean rights. Footage with releases from everyone in it, no third-party music and no clips licensed in from others can go into a dataset as it is. Everything else needs cutting, muting or clearing first, and buyers price that work in, or decline. The next section covers it frame by frame.

What buyers pass over

The other side of the list is shorter but just as useful. Generic footage in bulk is the most plentiful thing in the market: with one marketplace holding licences to 7M+ hours of video, a buyer can always find more vlogs and ordinary b-roll. Duplicates are worth nothing: a broker's founder described a library offered by the terabyte that contained " roughly four thousand copies of Frozen 2". And footage meant to copy a person is excluded by buyers' own terms: Protege says, " We do not permit Name-Image-Likeness (NIL) duplication / copying / use from media companies, including voice." What sells is footage that teaches a model how the world looks and moves, not how to imitate a particular person.

The what labs buy page tracks which kinds of data are rising and cooling across the market, and the video page keeps the current prices, buyers and lots for footage specifically.

The rights you need in every frame

A clip is a bundle of rights, not a single thing. The picture belongs to whoever shot it (or whoever they work for); the people in it have rights in their own faces and voices; the soundtrack may belong to a record label and a music publisher; a logo, a painting or a film playing on a screen in the background belongs to someone else again. A buyer licensing your footage for AI training is licensing all of it at once, so it will ask you to warrant that you hold every one of those rights for this use, and a single frame you cannot clear is a frame you have to cut.

This is where most video sales slow down. The rights check is not a formality at the end; it decides which part of a library can be offered at all, and therefore what the library is worth. The checks below are the ones that come up for almost every holder, in the order to do them. The guide on training data provenance explains why buyers now ask for this evidence up front.

1. You must own the footage

Copyright in a video starts with the person who made it: under US law, " Copyright in a work protected under this title vests initially in the author or authors of the work." For a solo creator who shot everything personally, that is simple. For a production company it is the first thing to check. Footage your employees shot in the course of their work belongs to the company. Footage a freelancer shot for you counts as a work made for hire only if it was commissioned for use as part of a motion picture or other audiovisual work and both parties signed a written agreement saying so; otherwise the copyright stays with the freelancer unless they assigned it to you, and an assignment must be in a signed written instrument to be valid. Footage shot for a client may belong to the client, depending on the contract.

The practical step is a chain of title for each part of the library: who shot it, under which contract, and whether that contract made it yours. Gaps are common in older archives, and a gap does not mean the footage is lost to you: it means the freelancer's or client's agreement is needed before that footage goes into a sale.

2. Music stays with its rights holders

A licence to put a track in your edit is not a licence to hand that track to an AI company for training. Music rights holders are the most active litigants in AI: record labels sued two AI music companies seeking up to $150,000 per infringed work, a Munich court ruled for the German collecting society GEMA against Suno on AI music training, and Warner Music Group settled with Suno on terms that bring in new licensed AI models. A buyer will not take a soundtrack it has no licence for. The fix is usually easy: deliver the raw files, or picture with the original location sound only, and leave licensed music out entirely. The cases page follows the music suits.

3. People: releases, likeness and biometrics

Anyone recognisable in your footage has rights in their own image and voice, and three things follow for a sale.

Releases need to cover this use. A release signed for a shoot usually describes that production. The standard a cautious buyer looks for is written releases naming AI training and sale, from contributors who consented. Footage without releases is not worthless, but the valuation discounts it, and buyers either discount it too or take only the shots where nobody is identifiable.

Faces can be biometric data. Illinois's biometric privacy law states that no private entity holding a biometric identifier " may sell, lease, trade, or otherwise profit from" it, and suits under it are active: in October 2026 two proposed class actions accused delivery-robot companies of scanning customers' faces and hands and retaining the data for AI training. The usual protection is not to extract face templates yourself, to exclude or separately release people who may be in Illinois, and to have the buyer warrant it will not extract identifiers without its own consent basis.

Digital replicas are their own question. In California, a services contract clause that allows a digital replica of the individual's voice or likeness in place of work the individual would otherwise have performed in person is unenforceable unless it describes the intended uses with reasonable specificity or the performer was represented by counsel or a union. Training a general video model is not the same as making a replica, but if your talent agreements are old or vague, assume they do not cover replica uses, and expect buyers' terms to exclude them in any case, as Protege's do.

4. Other people's work inside your frame

Logos, artworks, a film playing on a screen, another creator's clip cut into yours, a stock shot you licensed for an edit: each is someone else's work. The standard a buyer applies is that third-party content is identified and removed or licensed; what remains is the company's own original work. Incidental background is rarely what a buyer worries about; whole shots of someone else's footage are. Stock clips are the trap production companies fall into most: a stock licence was written for use inside your production, so unless it says you may license the clip onward for AI training, it comes out of the dataset.

The rights file

Put together, these checks become a rights file: for each part of the library, who owns it and by what document, which shots have releases and what they cover, what was removed (music, stock, third-party shots) and why. It is what lets a buyer say yes quickly, and it is the same document a datasheet asks for. The guide on what a data licensing agreement contains explains the warranties a buyer will ask you to give on top of it, and the rights hub covers the law behind each check.

Platform terms that matter

Uploading a video to a platform does not give the platform your copyright, but it does put the video under the platform's terms, and those terms now speak to AI training directly. The one that matters most to creators is YouTube's.

YouTube's third-party training setting

YouTube lets creators and rights holders decide whether outside AI companies may train on their videos. By its own help pages, the facts are these. It is off unless you turn it on: " By default, the third-party training setting is turned off", and if you do not want your content used for third-party training you need do nothing. You can choose who: you select companies from a list, or allow all third-party companies, which YouTube says means "any third-party company, even if they're not listed". Only public videos qualify, and only where all the applicable rights holders on the video allow it. Turning it on lets YouTube share your eligible videos with the companies you allowed, for training the models they build. And, most important for anyone hoping to earn from it, YouTube does not pay you for it: " YouTube isn't facilitating payments between third-party companies and creators or other rights holders at this time". YouTube also says that it cannot control what a third-party company does once it has the videos. Changes can take up to a week to show in YouTube's public data interface, and rights holders using YouTube's Content Manager control the setting there ( YouTube Help).

What this means in practice: the setting is a permission, not a sale. Turning it on, especially for all companies, gives away the thing a buyer would otherwise pay for, with no price and no say over terms. A creator who wants to be paid licenses footage through a buyer or marketplace under a written agreement, and that agreement usually covers material the creator controls outside YouTube anyway: the raw files, the unpublished takes and the higher-resolution masters, which are what buyers want most. Before you turn the setting on, check that it does not conflict with an exclusive licence you have signed, or plan to sign, for the same footage.

Marketplace and app terms

The intermediaries' own terms differ more than their headlines, and the differences are the price of the deal as much as any per-minute rate. Read the grant (what rights the buyer gets), the exclusivity, and what happens if you change your mind.

Two lines in that table deserve a second read. A perpetual, irrevocable, sublicensable licence is in effect a sale of the training rights: the intermediary can pass your footage on through several tiers of buyers for as long as it likes. And an exclusive licence stops you selling the same footage anywhere else, which is worth agreeing to only for a price that reflects it.

Payment models: flat fee or revenue share

There are three ways footage gets paid for, and most holders meet all three in their first round of conversations. The choice between them is a choice about risk and about how many times the same footage can earn, so it is worth deciding on purpose rather than taking whichever model the first buyer offers.

A flat fee per unit pays a fixed amount per clip, minute, hour or submission, once. Adobe's $120 per submission, Defined.ai's $100 to $300 an hour for longer films and Stockfilm's from $0.95 per clip are all flat. You know what you will get, and you are paid when the footage is accepted. The cost is that the buyer, not you, captures whatever the footage earns later.

A revenue share pays you a share of whatever the intermediary earns when it licenses your footage, on every sale. Troveo pays a share of the revenue on every sale, the first and every one after, Protege according to the agreed partnership terms, and Veritone passes a revenue share to the content rights holders. Non-exclusive data can be sold again: Troveo advertises an average of 4 sales per dataset on its business page, labelled illustrative. The costs are that you do not know the total in advance, the shares are mostly unpublished, and the platform keeps a large part: Wirestock set its dataset-deal fee at 50 percent, and Newsflare's worked example pays footage holders half of the hourly rate the AI buyer pays.

A lump-sum licence pays one negotiated amount for a whole library over a term. Shutterstock's large-technology deals initially ranged from $25 million to $50 million each, and Kled reported a $12M non-exclusive data licensing deal over two years. This model is open to libraries large and distinctive enough to negotiate, and it is where terms matter most.

Exclusivity is a price, not a default

Whichever model you take, exclusivity is the term that changes the price most. An exclusive licence fetches about 4 to 5 times a non-exclusive one, by what founders who sell training environments told Epoch AI; the advice that follows is to price exclusivity as a multiple, and never give it by default. Troveo says 95 percent of its licensors signed exclusively, so many creators have given it, often for a general rate. If you do give exclusivity, keep it narrow: for a defined field and a defined window, and for a multiple. The term of the licence matters too: for a living library, a two to three year term with renewal and refresh rights keeps the holder's leverage.

The licence terms page sets out each clause a footage licence should contain, and the negotiation page covers how to hold a position on exclusivity, term and price.

How to approach a sale

The right approach depends mostly on how much footage you hold and how clean its rights are. A creator with a drive of general footage and a production company with a decade of professional shoots are in the same market, but not in the same conversation. What follows is the sequence for each, with a worked example of each. The examples are illustrations, not valuations of any real holder, and their only numbers come from records.

If you are a creator

1. Separate published from unpublished. List the hours of raw takes, cut material and b-roll that never went online, and note resolution and subject. That is the part buyers pay most for.
2. Strip what is not yours. Remove or mute licensed music, cut clips from other creators and stock shots, and note any brand or artwork that fills a shot.
3. Check the people. Mark which footage has nobody identifiable, which has people with releases, and which has people without. Offer the first two; the third needs releases or cutting.
4. Describe it. A spreadsheet with what is in each file, where and when it was shot and on what camera is enough to start, and it moves footage up the price range.
5. Decide the YouTube setting deliberately. It gives permission without payment; leave it off unless you have decided to allow training without being paid.
6. Compare intermediaries on terms, not headlines. Read the grant, the exclusivity, the share and the withdrawal terms side by side before you sign anything.

Example: a travel creator with a drive of unused 4K footage. Take a creator with a large drive of unused 4K b-roll from a decade of travel videos, about as many hours as the valuation's starting answer: landscapes, cities, aerial shots, very few identifiable people, licensed music only in the published edits, the raw files clean. At the valuation's general rate of $33 to $120 an hour of footage an hour, 400 hours of footage comes out at $13,000 to $48,000; at the high-specification rate for 4K, drone and cinematic work, $60 to $240 an hour of footage an hour, the same hours price higher. The one public realised figure, about $0.55 a minute, sits at the bottom of that, and tells the creator what a large general catalogue has actually fetched. For a library of this size, the practical route is a marketplace that pays per hour or a revenue share: the range straddles the $26,000 from which we take on a sale, and the marketplace's share is the price of not running a sale yourself. What raises this creator's result is the aerial and 4K material, good descriptions, and keeping exclusivity off the table unless it is paid for.

If you are a production company or a library

1. Build the chain of title. For each production: who shot it, under which crew or freelance agreement, whether the footage was a work made for hire or assigned in writing, and whether a client owns it under the commissioning contract.
2. Map the releases. Which talent, contributor and location releases exist, and whether they mention AI training or onward licensing. Split the library into footage that is clear, footage that needs new releases, and footage to leave out.
3. Remove music and licensed-in material. Deliver picture with original sound, and pull stock and third-party shots unless their licences allow onward AI licensing.
4. Describe at scale. Shot logs, production notes and existing archive metadata are worth converting, because description sets where an hour prices.
5. Decide the terms before the first call. Non-exclusive or exclusive by field and window, term length, refresh rights, and whether the buyer may sublicense.
6. Let buyers compete. Large professional libraries are where competing offers make the most difference, because the buyers' needs and prices differ widely.

Example: a documentary and commercial production house. Take a company with a decade of 4K documentary and commercial footage, a large archive shot partly by staff and partly by freelance crews, talent releases on its commercial shoots, music licensed into its finished pieces, and good production logs. Its first job is the chain of title: the staff footage is the company's, but each freelance shoot needs either a written work-made-for-hire clause or a written assignment, and commissioned commercial footage may belong to the client who paid for it. Its second is the split: wildlife, landscape, aerial and city footage with nobody identifiable can go to market at once; commercial footage with actors goes only where releases cover it, and California's rule on digital replica clauses is a reason not to stretch old talent agreements. Its third is pricing: professional, described, high-resolution material is what the higher quotes describe, from the $60-$300 per hour cited for premium archive content to $100 to $300 an hour for longer films, and Protege's own case studies report libraries earning $250,000+ in year one and over $1M in 6 months. One structural point from those case studies is worth noting: Protege said that holding a broadcaster's exclusive AI training rights let it maximise the price because buyers could not negotiate with anyone else. Exclusivity concentrates the competition in whoever holds it, which is good for the holder only if that party then runs a real competition among buyers.

Questions people ask

How much do AI companies pay for videos? The widely reported rate for a creator's unused footage is $1 to $4 a minute, with most footage at between $1 and $2 per minute. Realised amounts are lower: one creator received about $0.55 a minute on a large catalogue. Across every public record, the median sale is $42 per hour of video per hour of video; premium, described archive content is quoted at $60-$300 per hour.

Can I sell my YouTube videos to AI companies? Yes, if you own them and hold the rights to everything in them. Buyers want unpublished footage and high-resolution masters most, and they license through marketplaces such as Troveo or brokers such as Protege under written agreements. YouTube's terms prohibit unauthorized downloads and scraping, which is part of why labs pay to license.

Does YouTube pay me if I allow third-party training? No. YouTube's help page says it " isn't facilitating payments between third-party companies and creators or other rights holders at this time". The setting is off by default, and turning it on gives permission to the companies you choose, or to all of them, with no payment attached.

Do I need releases from people in my videos? For anyone identifiable, a cautious buyer wants written releases naming AI training and sale, and buyers exclude footage meant to copy a person's likeness, as Protege's terms do. Footage without releases sells at a discount or only where nobody can be identified, and faces can count as biometric data under Illinois law.

Can I sell footage that has music in it? Not with the music. A licence to use a track in your edit does not cover training, and music rights holders are suing AI companies over training, seeking up to $150,000 per infringed work. Deliver raw files or picture with original sound only.

Is a flat fee or a revenue share better? A flat fee is certain and paid once; a revenue share can pay many times, since non-exclusive data can be sold again, but the share is usually unpublished and platforms keep up to about half. For a large, distinctive library, a negotiated licence with a fixed minimum usually beats both.

Should I sign an exclusive deal for my footage? Only for a price that reflects it: exclusive licences fetch about 4 to 5 times non-exclusive ones. Keep any exclusivity to a defined field and a defined window, and watch for perpetual, irrevocable, sublicensable grants such as Kled's, which amount to selling the training rights outright.

What kind of footage sells best? Unpublished, high-resolution, well-described footage of specific subjects, with clean rights: the early buyers paid premium rates for high-quality 4K footage, drone video, and 3D animation, and description alone can move an hour across a thousand-fold range. Generic vlogs and duplicates are the most plentiful and the least valuable.

What this means for you

Selling video for AI training is mostly a rights exercise with a price attached. The holders who earn the most per hour have footage nobody else has, described well, with every right cleared, and they choose their terms rather than accepting the first ones offered. The steps below turn that into a plan.

1. Know your number before anyone quotes one. The price board and the rates page hold every public price per minute and per hour, and the valuation puts a range on your own footage from them. The broader guide to AI training data prices shows where video sits against other kinds of data.
2. Build the rights file first. Chain of title, releases, music and licensed-in material, as set out above; the datasheet tool structures it. The guides on fair use and AI training and scraping for AI explain why buyers now pay for footage they are sure they may use.
3. Treat the YouTube setting as a decision, not a formality. Off by default means nothing is given away; on means permission without payment.
4. Compare buyers on terms. The video page, the buyer board and the guide to AI training data companies cover who takes footage and how each pays; the guide to data licensing agreements explains the grant, exclusivity and warranties to read for. Check any written offer with the offer check.
5. Match the route to the size. A small or general library is usually best listed directly with a marketplace; a large, professional or specialist library is where competing offers make a difference. The page on going direct or through us sets out the trade-off.

AIData.Broker acts as the holder's agent: the holder itself licenses or sells its data to the buyer, and we do not take ownership of it. Only the holder pays us, a fee fixed in the mandate and paid only if a sale closes, and we run sales only for holders based in the United States.