Skip to content

What is your content worth to AI companies? Follow the money

We tracked 28 AI licensing deals, programs and a settlement to separate reported prices, company revenue and what creators receive.

Veri Research14 min read

On this page
  1. A thousand hours for $33,000
  2. How much do AI companies pay for creator content?
  3. Why do quoted rates and real payouts differ?
  4. Who keeps the money from AI licensing deals?
  5. Why would AI companies pay if they can train on public video?
  6. Is human-made content getting more valuable?
  7. How much is one creator’s data worth?
  8. Should you sell your videos for AI training?
  9. What does YouTube’s AI training setting do?
  10. Know what your content is
  11. What we don’t know yet.
  12. More Veri Research Reports
  13. Sources
  14. Frequently asked questions

AI companies pay for creator content, but a licensing price is not always a creator payout. Press reports put some unused video at $1 to $4 per minute,¹ a three-year book license at $5,000 per title split equally with the author,² and Reuters, citing sources, put Google’s Reddit agreement at about $60 million a year.³ Our 28-row dataset covers deals, programs and one settlement. It separates reported prices from disclosed terms and individual earnings from company totals. Those distinctions matter more than any headline rate when you decide whether to license your content.

A thousand hours for $33,000

Peter Hollens makes a cappella videos on YouTube. Semafor reported in October 2025 that licensing about 1,000 hours of his old videos for AI training had earned him $33,000 over the preceding year.⁴ Our calculation is about $33 per hour of content, or 55 cents a minute. That is earnings over a year, not necessarily the price of one license.

Bloomberg reported in January 2025 that AI companies including OpenAI, Google and Moonvalley were paying for unpublished footage at $1 to $4 per minute.¹ Applying that range to 1,000 hours gives $60,000 to $240,000. It does not establish what a buyer paid for Hollens’s footage, or how much an intermediary retained.

The same Bloomberg report said most footage, such as unused video made for YouTube, Instagram and TikTok, was selling for $1 to $2 per minute, with 4K, drone and 3D-animation footage priced higher.⁴⁴ At that typical range, 1,000 hours would come to $60,000 to $120,000.

The available figures leave an important gap: they rarely show the full path from buyer payment to creator earnings. The question for a creator is how much reaches you, and what rights you give away to get it.

To answer it, we assembled 28 public-source rows covering licensing deals, programs, marketplace updates and one copyright settlement. This is a selected sample, not a census or 28 independent transactions: several rows concern the same platform. For each we recorded the buyer, seller or intermediary, licensed material, public price or creator share, and consent terms where available. We distinguished figures disclosed in filings or company terms from press reports and company claims, and checked the article’s links on November 5, 2026.

How much do AI companies pay for creator content?

Prices vary by orders of magnitude, mostly because buyers are paying for different things. Here is a selection from the 28 rows.

Deal Date What was licensed Price or rate Creator share Basis
Reddit to unnamed licensees Jan 2024 Platform content and data $203M aggregate contract value over 2 to 3 years⁵ None disclosed for these licenses Disclosed
Reddit to Google Feb 2024 Posts for model training About $60M a year³ None disclosed Reported
News Corp to OpenAI May 2024 News archive and current content About $250M reported deal value⁶ None disclosed Reported
Defined.ai marketplace Apr 2024 Images, video, film $100 to $300 per hour of film⁷ Split, undisclosed Reported
Adobe contributor offer Apr 2024 Short everyday-action clips About $2.62 per submitted minute on average, up to about $7.25⁸ Submission payment offered Reported offer
HarperCollins to Microsoft Nov 2024 Nonfiction titles, 3-year license $5,000 per title² 50% Reported
AI labs from creators Jan 2025 Unused video footage $1 to $4 per minute¹ Not stated Reported
Troveo, first year Mar 2025 Creator footage, exclusive About $5M for 1M hours⁹ Percentage not stated Reported company totals
Shutterstock data licensing 2023 revenue Content datasets $104M AI-licensing revenue in 2023, reported via a law-review article¹⁰ 20% average corporate royalty rate in current terms¹¹ Revenue reported, rate disclosed
UMG to Udio Oct 2025 Recorded music and publishing Not disclosed¹² Not disclosed Disclosed deal
Believe to Spotify Oct 2026 Repertoire for AI remixes Not disclosed¹³ Royalties, rate undisclosed Disclosed deal
Fox to Meta Undisclosed Multimedia content Price withheld from discovery¹⁴ Not stated Court record
YouTube setting Current policy Permission to train YouTube does not facilitate payments¹⁵ No payment mechanism in setting Disclosed policy

The table mixes contract values, company revenue, offers and creator earnings because each answers a different question. It cannot establish an industry-wide intermediary margin. Nor should its rows be added together: the Google agreement may overlap with Reddit’s aggregate licensing figure.

Why do quoted rates and real payouts differ?

A buyer-side rate and a creator-side payout are different numbers. Even figures quoted per minute may cover different rights, content or payment periods.

Defined.ai’s chief executive told Reuters in April 2024 that companies were generally willing to pay $1 to $2 per image, $2 to $4 per short video and $100 to $300 per hour of longer film, with rates varying by buyer and content.⁷ The same article reported that one Brazilian supplier paid owners of photos, podcasts and medical data about 20% to 30% of total deal amounts.⁷ That is one supplier’s split, not evidence of a standard video royalty.

Adobe’s offer illustrates another arrangement. Bloomberg reported in April 2024 that Adobe was offering its photographer-and-artist network $120 to submit videos and requesting more than 100 short clips of actions, emotions and anatomy. According to documents seen by Bloomberg, the submission payment worked out to about $2.62 per minute on average and up to about $7.25 per minute.⁸ Those are reported offer terms, not verified completed payouts or proof that the footage had to be newly filmed.

The New York Post reported in March 2025 that Troveo had paid about $5 million for 1 million hours of content since launching a year earlier, while footage typically sold for 75 cents to $3 per minute.⁹ Dividing those approximate totals gives $5 per hour, or about 8 cents per minute. That is our aggregate calculation, not a verified per-license rate: the report does not reconcile its hour count with the footage priced at the quoted range.

Troveo’s creator page now advertises $50 million-plus paid and 7 million-plus hours licensed.¹⁶ A separate article dated October 19, 2026 advertises $20 million-plus paid.¹⁷ These are company claims checked November 5, 2026. They are lower bounds, so they neither establish a reliable dollar-per-hour average nor contradict one another.

Averages also leave out the payout distribution. Troveo’s chief executive told Semafor that the company had paid a handful of creators more than $1 million each,⁴ but that does not tell us what a typical creator earned. The writer of one pseudonymous Reddit post from March 2026 alleges being promised $36 to $72 an hour and receiving under $8 in a year.¹⁸ We have not verified the contract or earnings behind that complaint. It is not evidence of a typical payout. Troveo’s support page says “Payouts only occur when your content is licensed” and that there is no fixed payout schedule.¹⁹ The company says there’s no public rate card.¹⁷

One small survey provides a view of stock-photography payouts. In 2023, photographer Robert Kneschke asked 58 other photographers about their Shutterstock Contributor Fund payments and compared them with portfolio sizes. The median was $0.0069 per image.²⁰ This self-selected sample is not a representative rate card or a lifetime payment per image.

Who keeps the money from AI licensing deals?

Follow the largest sums and they lead to platforms. Reddit’s IPO filing disclosed that it had signed data licensing deals in January 2024 with a total contract value of $203 million over two to three years.⁵ Reddit’s “Other revenue” line, mostly data licensing, grew from $15.2 million in 2023 to $114.7 million in 2024 and $140.0 million in 2025.²¹ Neither filing discloses a licensing-revenue share for the people who wrote the posts. That is separate from Reddit’s contributor earnings programs.⁵

Spectrum News attributed an approximately $250 million valuation of News Corp’s OpenAI deal to the Wall Street Journal.⁶ A 2026 law-review article, citing Bloomberg, reported $104 million in Shutterstock AI-licensing revenue for 2023.¹⁰ Neither figure is a creator payout. Shutterstock’s current terms state a 20% average corporate royalty rate on revenue received for data licenses.¹¹ We cannot use that current rate to reconstruct the actual distribution of its 2023 revenue.

Adobe’s 2026 Firefly contributor bonus is discretionary and varies by eligible contributor. Its formula uses eligible assets considered for training and the licenses those assets generated during the specified 12-month period. The public FAQ gives no universal per-asset rate and says Adobe Stock content currently has no training opt-out.²²

The reported HarperCollins agreement offered authors a defined share and a choice. Microsoft’s three-year license paid $5,000 per nonfiction title, split equally with the author, who had to opt in.² The reported author share is $2,500 per title. It does not prove that opt-in alone produces a better price.

The music deals examined here leave artist payment amounts undisclosed. Universal Music Group settled with Udio in October 2025 and agreed to license a new AI platform.¹² Warner Music Group settled with Suno in November 2025.²³ Believe, which owns the distributor TuneCore, announced an agreement for Spotify’s upcoming AI remix tool on October 20, 2026, with a two-step artist opt-in and royalties on remix streams.¹³ The Los Angeles Times reported in September 2026 that none of the label deals had disclosed what artists would be paid.²⁴

Likeness is following the same path. ElevenLabs says voice creators have earned more than $22 million from its Voice Library, where other customers pay to generate audio with their cloned voices.²⁵ That is a likeness market, not a training market. Its published cumulative earnings figure is company-reported, not an independently audited payout total. Our separate piece on getting an AI clone without giving up ownership covers those terms.

A court order in Kadrey v. Meta confirmed that Fox licensed multimedia content to Meta for AI training. The court granted a protective order withholding unredacted fee and scope terms from discovery and required the plaintiffs to destroy inadvertently produced unredacted documents.¹⁴ The public order gives no license price.

Why would AI companies pay if they can train on public video?

Licensing can secure permission and content that a buyer cannot simply assume it may use. The Atlantic reported in September 2025 that it found more than 15.8 million YouTube videos from more than 2 million channels in at least 13 AI-related datasets, while explicitly warning that dataset inclusion does not establish that a particular video was used for training.²⁶

The book cases do not settle the legal position for video. In June 2025, Judge William Alsup ruled that Anthropic’s use of the plaintiffs’ books for training was fair use, calling it “exceedingly transformative”.²⁷ He separately found that copying and storing pirated books in its central library was not fair use and ordered a damages trial. On June 25, Judge Vince Chhabria ruled for Meta in Kadrey v. Meta. He found the authors had not shown enough market harm, pointed to market dilution as the stronger theory they had failed to argue, and wrote: “If using copyrighted works to train the models is as necessary as the companies say, they will figure out a way to compensate copyright holders for it.”²⁸

Anthropic then settled the piracy-related claims. The $1.5 billion settlement received final approval on July 20, 2026, with an approximately $3,000 per-work award divided among valid copyright-owner claimants under applicable splits.²⁹ That is not a guaranteed $3,000 payment to each author or a market price for a voluntary training license.

A license can specify permitted uses, payment and exclusivity. Payment alone does not establish that footage is unique or that all necessary rights have been cleared. The US Copyright Office’s May 2025 report on AI training found that voluntary licensing markets were developing, and that harm to traditional, reasonable or likely licensing markets can weigh against fair use. An unpaid hypothetical license fee does not automatically establish market harm.³⁰ In Europe, the AI Act requires general-purpose model providers to publish a summary of their training content, with obligations applying from August 2, 2025 and a transition until August 2, 2027 for models placed on the EU market before that date.³¹ A training-content summary is a transparency requirement, not proof that each use was licensed.

Is human-made content getting more valuable?

There is a real case that it is. In a 2024 paper in Nature, Ilia Shumailov and colleagues showed that models trained on the output of other models degrade over generations, losing the rare cases first. They called this model collapse.³² The authors argued that genuine human-interaction data would become more valuable. Their recursive-training experiments do not prove that all synthetic-data methods fail or forecast a price for creator footage.

Epoch AI’s 2024 projection estimated an effective, quality- and repetition-adjusted stock of public human text at about 300 trillion tokens, with a 90% confidence interval of 100 trillion to 1,000 trillion. Under its trend assumptions, full utilization could occur between 2026 and 2032, an 80% confidence interval.³³ This is a conditional text-scaling projection, not a forecast of video licensing revenue.

That helps explain why the quoted prices favor raw, unpublished footage. Bloomberg reported that the footage buyers want has never been posted online, which makes it unique.¹ That gives creators a reason to ask about unpublished footage rather than assume only finished uploads have value.

How much is one creator’s data worth?

Researchers have tried to measure what each piece of data contributes to a model. In 2019, Amirata Ghorbani and James Zou proposed Data Shapley, a way to assign each training example a share of the model’s performance.³⁴ In their supervised-learning experiments, values varied with the dataset, algorithm and evaluation metric, and some mislabeled examples had negative values. The study did not measure the value of creator catalogues in frontier generative models.

At the scale of a frontier model, one creator’s catalogue is a small slice. Meta made that argument to the Copyright Office: “Any fair royalty due would be incredibly small in light of the insignificance of any one work among an AI training set.”³⁰ That is an interested party’s argument, not a measured royalty entitlement or a conclusion of the Copyright Office.

An alternative economic framework treats data production as labor. In a 2018 conceptual paper, Imanol Arrieta-Ibarra and colleagues argued that treating data as free capital neglects the contribution of the people producing it and distributes gains unequally.³⁵ This framework offers a bargaining-power explanation, not an empirical price estimate for a creator’s catalogue.

Neither argument tells you what your archive should earn. A proposed deal still needs a buyer, a defined use and a price you can evaluate against the rights you would give up.

Should you sell your videos for AI training?

It can be worth it, for the right footage on the right terms. Before you sign with any marketplace, get written answers to these questions.

What exactly are you licensing?

Ask the contract to distinguish training, likeness and reference uses. Protege says its media-company licenses do not permit name, image or likeness duplication, including voice.³⁶ Created by Humans lets authors choose training and reference rights separately.³⁷ These are the providers’ published descriptions, checked November 5, 2026, rather than a review of a contract offered to you.

Is it exclusive, and for how long?

The New York Post described exclusive AI-training rights granted to Troveo in March 2025.⁹ Troveo’s own October 2026 guide describes non-exclusive licenses to multiple buyers, which does not settle whether a creator’s intermediary agreement is exclusive.¹⁷ Ask whether your contract reserves your ability to make other deals or train your own model.

How is the payout calculated?

Ask for the per-minute or per-hour rate, the intermediary’s share and the payment trigger. A quoted range is not a payout until a buyer licenses your footage.

Can you leave?

Ask what happens to licenses granted before you exit. ElevenLabs, for example, lets voice creators set a withdrawal notice period of up to two years.³⁸

Who are the buyers?

Wirestock’s creator page describes payment when approved content is licensed as part of a collection, without naming the buyers on that page.³⁹ Ask who can receive your content and whether you approve each license. Wired described Calliope Networks as a founding member of a data-provider alliance that requires creators and rights holders to opt in to scraping.⁴⁰

If the answers are vague, ask for clarification before signing. A promised price range does not resolve an unclear grant of rights.

What does YouTube’s AI training setting do?

YouTube added a setting in December 2024 that lets creators allow third-party companies to train on their videos.⁴¹ The setting is off by default. The initial list of eligible companies ran to 17 names, including OpenAI, Anthropic, Meta, Apple, Amazon, Microsoft, Nvidia and xAI.⁴²

The setting does not arrange a payment. YouTube’s help page says it is not currently facilitating payments between creators and the third parties.¹⁵ The setting does not govern Google’s own training use, according to reporting on its launch.⁴¹ We cover the details in our guide to YouTube’s third-party AI training setting.

The setting supplies permission without itself arranging compensation. A licensing deal sells similar permission for money. If you plan to license footage, check whether the setting overlaps with an exclusivity clause.

Know what your content is

The licensing market prices creator content as raw material. Your footage goes in, and someone else’s model comes out. The large platform contracts in our table show the bargaining position that comes with controlling a collection of rights. They do not tell an individual creator whether a particular offer is fair.

Veri doesn’t run a licensing marketplace and doesn’t pay creators for content. Its agents start from your own transcripts and published videos, and your saved preferences and corrections shape how they write. Veri says your data never trains a shared model, and that data received through platform APIs is not used to train AI models. A personal model of your own is a separate opt-in choice, with training by request and requests currently joining a waitlist, explained in your data and model training.

In practice, that means knowing what you have, including unpublished footage, deciding on YouTube’s third-party training setting on purpose, and keeping the option to build a model you own before signing exclusive rights to someone else. If you use AI writing tools, check whether they train on your content too.

Before exchanging rights for a check, decide which uses of your content you want to preserve. Keep that decision separate from whether you use an AI writing tool or request a personal model.

What we don’t know yet.

Realized creator payouts. We did not find representative per-creator payout data. Our historical Troveo calculation divides approximate reported dollars by reported hours, whose treatment of repeat licensing is unclear. Current minimum totals cannot establish an average.

Music and likeness terms. The label deals with Udio, Suno and Spotify have not disclosed artist rates, and the LA Times found the same.²⁴ Loti AI’s Interchange, launched April 30, 2026, promises revenue sharing without a published percentage.⁴³

Where the law lands. The June 2025 fair use rulings covered books, not video. The Copyright Office report is a pre-publication version, and appeals and new cases could shift the market either way.

Whether rates rise. If model collapse makes human footage scarcer, prices could climb. If synthetic data works well enough, they could fall. We’ll update this piece as new deals disclose numbers.

The full dataset of 28 deals and programs, with sources and check dates, is on the data page.

Connect a channel at joinveri.co to use agents that start from your connected content and saved preferences. Personal-model training is a separate opt-in request and currently has a waitlist.

More Veri Research Reports

Every report, with its data, is on Veri Research.

The data and the full source list for this report are on the data page.


Sources

  1. Counts, A. (2025, January 10). YouTubers Are Selling Their Unused Video Footage to AI Companies. Bloomberg News, via Bloomberg Law. news.bloomberglaw.com
  2. Harmon, G. (2024, November 25). Microsoft and HarperCollins sign AI licensing deal, but author opt-in still required. EMARKETER. emarketer.com
  3. Tong, A., Wang, E., & Coulter, M. (2024, February 22). Exclusive: Reddit in AI content licensing deal with Google. Reuters. reuters.com
  4. Jones, R. (2025, October 31). The AI boom is turning old content into cash cows for creators. Semafor. semafor.com
  5. Reddit, Inc. (2024, March 11). Amendment No. 1 to Form S-1 registration statement. U.S. Securities and Exchange Commission. sec.gov
  6. Carpenter, S. (2024, May 23). News Corp signs deal with OpenAI. Spectrum News. spectrumlocalnews.com
  7. Paul, K., & Tong, A. (2024, April 5). Inside Big Tech’s underground race to buy AI training data. Reuters. reuters.com
  8. Bloomberg. (2024, April 11). Adobe is buying videos for US$3 per minute to build AI model. The Business Times. businesstimes.com.sg
  9. Schlott, R., & Moynihan, L. (2025, March 13). Influencers are making big money selling leftover videos - ones not yet posted online - to train AI. New York Post, via Yahoo Tech. tech.yahoo.com
  10. Joralemon, V. (2026, July). The OnlyFans Economy: Intellectual Property’s Pivot from Scarcity to Authenticity. Virginia Law Review Online, 112, 135–161. The $104 million figure cites Bloomberg reporting. virginialawreview.org
  11. Shutterstock. (2026, September 18). Shutterstock Data Licensing and the Contributor Fund. Shutterstock Contributor Help Center. Checked November 5, 2026. submit.shutterstock.com
  12. Universal Music Group. (2025, October 29). Universal Music Group and Udio announce Udio’s first strategic agreements for new licensed AI music creation platform. Press release. universalmusic.com
  13. Believe. (2026, October 20). Believe and Spotify partner to unlock new opportunities for artists through licensed fan remixes and covers. Press release. believe.com
  14. Hixson, T. S. (2026, October 19). Discovery Order, Kadrey v. Meta Platforms, Inc., No. 3:23-cv-03417. U.S. District Court for the Northern District of California, via GovInfo. govinfo.gov
  15. YouTube. (n.d.). Your content & third-party training. YouTube Help. Checked November 5, 2026. support.google.com
  16. Troveo. (n.d.). License Your Content to Top AI Companies. Troveo. Checked November 5, 2026. troveo.ai
  17. Troveo Team. (2026, October 19). How to Sell Your Videos for AI Training. Troveo. Checked November 5, 2026. troveo.ai
  18. FearlessBarracuda195. (2026, March 15). Do NOT sign a Troveo.ai Content License. They altered my contract, trapped my 1,700+ videos in a warehouse, and paid me Less Than 8$. Reddit, r/Reviews. Unverified personal allegations. reddit.com
  19. Troveo. (2026, February 19). How do payouts work on Troveo? Troveo Support. Checked November 5, 2026. support.troveo.com
  20. Growcoot, M. (2023, July 12). Shutterstock May Have Paid Out Over $4 Million From its AI Contributor Fund. PetaPixel. petapixel.com
  21. Reddit, Inc. (2026, February 5). Annual report on Form 10-K for fiscal year 2025. U.S. Securities and Exchange Commission. sec.gov
  22. Adobe. (2026, September 16). Adobe Firefly for Contributors FAQ. Adobe Stock Contributor. Checked November 5, 2026. helpx.adobe.com
  23. Reuters. (2025, November 25). Warner Music Group settles copyright case with Suno for licensed AI music. Reuters. reuters.com
  24. Davies, C. (2026, September 21). Why some musicians aren’t happy about labels signing AI deals with platforms. Los Angeles Times. latimes.com
  25. Amin, A., & Barrull, R. (2026, May 22, updated November 3). $22 million earned by voice creators on ElevenLabs. ElevenLabs Blog. Checked November 5, 2026. elevenlabs.io
  26. Reisner, A. (2025, September 10). AI Is Coming for YouTube Creators. The Atlantic. theatlantic.com
  27. Brittain, B. (2025, June 24). Anthropic wins key US ruling on AI training in authors’ copyright lawsuit. Reuters. reuters.com
  28. Authors Alliance. (2025, June 26). Meta Wins on Fair Use for Now, but Court Leaves Door Open for ‘Market Dilution’. Authors Alliance. authorsalliance.org
  29. Authors Guild. (2026, July 21). Court Grants Final Approval of $1.5 Billion Anthropic Copyright Settlement. Authors Guild. authorsguild.org
  30. U.S. Copyright Office. (2025, May). Copyright and Artificial Intelligence, Part 3: Generative AI Training (pre-publication version). copyright.gov
  31. European Commission, Directorate-General for Communications Networks, Content and Technology. (2026, March 26). Template for general-purpose AI model providers to summarise their training content. Checked November 5, 2026. digital-strategy.ec.europa.eu
  32. Shumailov, I., Shumaylov, Z., Zhao, Y., Papernot, N., Anderson, R., & Gal, Y. (2024). AI models collapse when trained on recursively generated data. Nature, 631, 755-759. nature.com
  33. Villalobos, P., Ho, A., Sevilla, J., Besiroglu, T., Heim, L., & Hobbhahn, M. (2024). Will we run out of data? Limits of LLM scaling based on human-generated data. Epoch AI. epoch.ai
  34. Ghorbani, A., & Zou, J. (2019). Data Shapley: Equitable valuation of data for machine learning. Proceedings of the 36th International Conference on Machine Learning, PMLR 97. proceedings.mlr.press
  35. Arrieta-Ibarra, I., Goff, L., Jiménez-Hernández, D., Lanier, J., & Weyl, E. G. (2018). Should we treat data as labor? Moving beyond “free.” AEA Papers and Proceedings, 108, 38-42. aeaweb.org
  36. Protege. (n.d.). Video. Protege. Checked November 5, 2026. withprotege.ai
  37. Created by Humans. (n.d.). Author overview. Created by Humans Help Center. Checked November 5, 2026. support.createdbyhumans.ai
  38. ElevenLabs. (n.d.). What are voice actor payouts? ElevenLabs Help Center. Checked November 5, 2026. elevenlabs.io
  39. Wirestock. (n.d.). Video Content Categories. Wirestock. Checked November 5, 2026. wirestock.io
  40. Knibbs, K. (2024, September 30, updated October 2). This Startup Wants YouTube Creators to Get Paid for AI Training Data. Wired. wired.com
  41. Perez, S. (2024, December 16). YouTube will now let creators opt in to third-party AI training. TechCrunch. techcrunch.com
  42. Peters, J. (2024, December 16). YouTube is letting creators opt in to allowing third-party AI training. The Verge. theverge.com
  43. Loti AI. (2026, April 30). Loti AI Launches Interchange: An Industry-First Platform for Generative AI Consent, Control, and Compensation. PR Newswire. prnewswire.com
  44. Counts, A. (2025, January 13). YouTubers are selling their unused video footage to AI companies. Bloomberg, via The Japan Times. japantimes.co.jp

Frequently asked questions

How much do AI companies pay for video footage?

Reported licensing prices include $1 to $4 per minute for some unused footage and $100 to $300 per hour for longer film. One creator reported earnings equivalent to about $0.55 per minute of licensed content. Adobe separately offered payments averaging about $2.62 per minute of requested submissions. These are different arrangements, not a market-wide rate card.

Can I sell my YouTube videos for AI training?

Licensing intermediaries offer routes to AI buyers, but eligibility, exclusivity and payment terms vary. Unpublished footage appears in reported deals, while some creators have also licensed older published videos. Get the permitted uses, revenue share and payment trigger in writing.

Does YouTube pay creators for AI training?

YouTube says it is not currently facilitating payments between creators and third-party AI companies. Its third-party training setting is off by default and does not govern Google's own training use. Turning it on is not a paid licensing deal.

Who keeps the money from AI data licensing deals?

Large aggregate deal values and revenue figures appear at Reddit, News Corp and Shutterstock. Examples of public creator shares include Shutterstock's 20% average corporate royalty rate and a reported 50% author share in the HarperCollins deal. The music agreements examined here do not publish artist royalty rates.

Is it legal for AI companies to train on my videos without paying?

Two US federal judges found book-training uses fair in particular cases in June 2025. One separately allowed piracy claims to proceed, and the other stressed the weakness of the plaintiffs' market-harm evidence. Those rulings do not establish a general right to train on anyone's videos without payment.

How much is one creator's content worth to an AI model?

There is no verified rate for one creator's contribution to a frontier model. Data-valuation research shows that training examples can contribute unevenly, but it does not price creator catalogues. Reported deals suggest that unpublished footage and clear licensing rights can attract buyers.

Does Veri pay creators or license their content to AI companies?

No. Veri does not run a licensing marketplace or pay creators for content. Veri says your data never trains a shared model. Personal-model training is a separate opt-in request and currently has a waitlist.

All articles

Try it on your own channel.

Veri, the creator co-pilot, has a full suite of creator tools that work from your own channel: Ideas, Validate, Research, Script, Titles, Thumbnails with Layers, Clips and Signals.

Get started

Free early access, no card required.