AI Content Copyright and Data Licensing: Investment Research

Second, the AI-search and RAG profit pool will more likely land with a few high-authority content libraries and the interface layer. The reason is that search/Q&A needs fresh, traceable, citable, and correctable content; enterprise RAG further requires permissions and audit. So productized databases such as Reuters, Factiva, LexisNexis, Westlaw, Wolters Kluwer, and Pearson are easier to charge for over the long term than ordinary news pages and general web text.

Third, the profit pool for music, voice, characters, and likeness will more likely tilt toward "rights management + filtering + revenue splits" rather than the raw model. Because the commercial-use risk on the user side is higher, and personality-rights/consent mechanisms cannot be simply substituted. Large-scale financial landing is still hard to see in public markets, but the industry direction has already shifted from "whether to license" to "who licenses, who filters, who clears."

Fourth, the visual-content profit pool will most likely concentrate on platforms with "clear rights + complete metadata + indemnification capability." Getty directly commoditizes "uncapped indemnification" and contributor compensation; Shutterstock sells training data on one hand and operates generation tools and a contributor fund on the other. This business model is closer to long-term high-margin SaaS/subscription than to one-off data sales.

On the question of whether AI copyright licensing is a long-term cost for model companies, my judgment is scenario-dependent: general-pretraining licensing looks more like a transitional, strategic cost; high-value professional data, RAG permissioned data, AI-search real-time citation, and commercial-use music/visual/likeness licensing look more like long-term structural cost. The EU AI Act's requirements on training-content summaries and copyright policy, and the rise of access-control layers such as Cloudflare/RSL/TollBit, are both pushing "transparent provenance + conditional payment" to become the norm.

Among these three scenarios, the model most worth investing in over the long term is not the one-off big deal but upgrading traditional subscription/copyright revenue into AI-native workflow revenue. This is also why companies like RELX, Thomson Reuters, Wolters Kluwer, and Pearson—though their "AI licensing narrative" runs less hot than media's—often offer higher investment quality.

Track Depth and Competitive Landscape

Below, the thirty sub-tracks the user listed are compressed by investable logic into fifteen-plus "profit-pool units." Scores are research priorities, not buy/sell recommendations.

On the competitive landscape, it can be summarized into four main threads. First, media groups' game paths differ: News Corp takes a hybrid strategy of "sign + keep negotiating + sue when necessary"; NYT sued first then signed, choosing Amazon rather than OpenAI for its first deal; AP entered licensing cooperation earlier; Reuters chose to license trusted news content to tech platforms such as Meta; Perplexity tries to win over publishers through rev-share. Top news brands have bargaining power, but that power is highly concentrated.

Second, professional-information companies prefer to turn content assets into AI workflow products rather than sell raw corpora to general models. Thomson Reuters has explicitly stated that third-party model partners may not use customer data to train models; its news business once recorded "generative AI related content licensing revenue," but its overall strategic focus is on professional products such as CoCounsel. RELX, Wolters Kluwer, and Pearson likewise embed AI into existing workflows and emphasize in their filings the value of trust, verification, evaluation, and embedded data.

Third, music and visual are two different copyright economics. The music rights chain is more complex but more concentrated, making it easy to form a "license–filter–revenue split" loop; the visual content rights chain is relatively clear, and as long as there are releases, metadata, and indemnification capability, it is easier to form enterprise safe-generation products. The former's core is catalog control and royalty systems; the latter's core is commercial safety and metadata.

Fourth, AI companies' content strategies also clearly diverge. OpenAI is more active in signing top licensing deals and announcing them loudly; Google buys both content and community data; Meta is later and more selective on news and social content; Anthropic faces relatively heavy pressure in public copyright litigation; Perplexity, ProRata, and TollBit represent the new path that "AI search/AI agents must pay content owners directly."

Investment Targets and Company Tiers

The table below prioritizes coverage of high-credibility companies with existing public evidence; for projects that do not separately disclose AI licensing revenue, it clearly notes "not separately disclosed" or "mainly defensive."

Based on public evidence, companies can be split into five tiers:

• Tier A: core direct beneficiaries of AI copyright/data licensing Thomson Reuters, RELX, Wolters Kluwer, News Corp, Reddit, Wiley, Stack Overflow. The common thread: either they already have clear licensing revenue, or they upgrade high-barrier content directly into AI workflow subscriptions.

• Tier B: clear beneficiaries, but with valuation, litigation, regulatory, or sustainability risk NYT, Getty, Shutterstock, Pearson, Adobe, WMG, UMG, Visual China, Cloudflare, ProRata, TollBit. The common thread: the commercialization direction is clear, but either revenue share is still small, or the market has partly reflected it, or it depends on new-standard adoption.

• Tier C: AI licensing is mainly a defensive tool, with weak near-term profit elasticity Integrated education platforms, some large content platforms, Created by Humans, Vermillio, and Loti—companies leaning more toward infrastructure and rights management. Their direction is right, but scale validation is early.

• Tier D: strong narrative, but lacking verifiable licensing revenue in this public dataset Most AI music/video generation startups, some "AI copyright concept" A-share software stocks, and companies that emphasize "AI content cooperation" but do not separately disclose contract amounts, revenue contribution, or customer expansion should stay on a watch list rather than be treated directly as beneficiaries. The most common problem in public materials is disclosing only "cooperation," "exploration," or "integration" without disclosing financial contribution.

• Tier E: may be hit by AI-generated content, AI search, or unlicensed substitution Free-traffic-dependent mid-to-long-tail media, low-differentiation footage libraries, visual-content platforms lacking releases and metadata, and general data brokers without clear proof of provenance will face stronger price pressure and a compliance discount in the AI era. This risk is directly flagged in the risk disclosures of companies such as News Corp, Getty, and Stack Overflow.

Risk, Valuation, and Final Conclusions

First, valuation and market expectations. As of May 19, 2026, NYT's P/E was about 32.4x, Reddit about 44.7x, Wiley about 14.6x, Warner Music about 20.5x, and S&P Global about 26.5x; while Shutterstock's market cap was only about US$584 million, Getty about US$404 million, News Corp about US$15.1 billion, and Reddit about US$31.7 billion. Purely from market pricing, the "AI content option" on Reddit/NYT is already not cheap; the revaluation room for Wiley, News Corp, and Getty/Shutterstock depends more on subsequent contract renewals and rising revenue share; while the professional-database leaders look more like high-quality compounders—their valuations are not cheap, but their logic is the most stable.

Using the user's weighting framework, my suggested positive scoring model is as follows: direct exposure to AI licensing revenue 20%; content-asset and copyright clarity 25%; customer quality and bargaining power 15%; data governance/metadata/API capability 10%; litigation and regulatory risk management 10%; financial quality and margins 10%; valuation reasonableness 10%. Under this model, the group with the highest current priority is usually not the "hottest" media stocks but companies that combine all three of content barriers + workflow + compliance.

The corresponding reverse risk-scoring model can be set as: insufficient licensing-revenue sustainability 20%; copyright-litigation and regulatory uncertainty 20%; high content substitutability 20%; AI companies' bargaining power too strong 15%; generated content depressing original-content value 15%; valuation too high 10%. Under this model, the highest risk is usually not professional databases but the high-narrative news, long-tail books, and the music/film/likeness tracks not yet covered by standardized licensing, that have not formed stable revenue-share mechanisms.

There are mainly six categories of systemic risk. First, U.S. courts continuing to expand fair-use space would suppress the long-term pricing of general-training licensing; second, AI companies cutting procurement and shifting to user data, internal data, and synthetic data would make one-off corpus contracts hard to renew; third, AI search diverting traffic faster than licensing compensation would squeeze small and mid-size publishers on both sides; fourth, the complex rights chain for music/likeness/voice means that even willingness to pay may not enable efficient rights clearance; fifth, provenance, watermarking, and detection technologies are not perfect, so infrastructure companies also carry technology-delivery risk; sixth, at high valuations the market will focus more on "revenue share" than on the "story."

The final conclusions can be condensed into the following ten points:

First, AI content copyright and data licensing is the "high-quality data supply layer" of the AI value chain, but not all content can be repriced; what can truly be priced is content that is rights-clear, rich in metadata, timely, industry-scarce, and verifiable in provenance.

Second, the five sub-tracks most worth watching are: professional-database AI workflows, UGC/API data licensing, rights-cleared visual data, academic/educational TDM licensing, and enterprise RAG data governance. If "selling shovels" is included, then Cloudflare/RSL/TollBit/compliance audit also merit high-priority tracking.

Third, the ten listed companies most worth deep research are: RELX, Thomson Reuters, Wolters Kluwer, News Corp, Reddit, Wiley, Informa, Pearson, Getty Images, and Adobe. For those who prefer music and the Chinese market, add Warner Music, UMG, and Visual China as elasticity supplements.

Among the unlisted, the ten most worth tracking are: Stack Overflow, TollBit, ProRata, Created by Humans, Vermillio, Loti, Scale AI, Databricks, Rightsify, and RSL Collective. The first five bet more directly on the "copyright and licensing layer," while the latter five lean toward the "data and infrastructure layer."

The five points the market most easily misreads are: First, assuming all copyright will be charged for universally; in reality it is more likely that "high-value data gets charged first." Second, assuming all AI licensing revenue will become large incremental gains; in reality much of it is merely defensive compensation. Third, assuming media is the highest-quality beneficiary; in reality professional databases are often stronger. Fourth, assuming music and likeness will immediately form a large market; in reality rights clearance and revenue splitting are the most complex. Fifth, assuming copyright-tech platforms have already scaled; in reality most are still in early validation.

The metrics most worth tracking over the next six to twelve months are: the share of AI licensing revenue, the licensing/other revenue split, the number of licensing contracts and renewal rate, the number of enterprise RAG customers, API call volume, AI-search citation traffic and revenue share, AI platforms' rev-share disclosure to publishers/creators, EU training-data summary enforcement, and the next round of rulings in key U.S. AI copyright cases.

For so-called "AI copyright platform-type companies," I lean toward placing RELX, Thomson Reuters, Wolters Kluwer, Reddit, Stack Overflow, Getty, and Cloudflare at the core; "AI-native data-licensing challengers" are TollBit, ProRata, Created by Humans, Vermillio, Loti, and RSL Collective; "AI content-licensing shovel sellers" include the layer of Adobe, Cloudflare, Snowflake, Databricks, Elastic, and MongoDB.

Companies at higher risk of being hit by AI-generated content, AI search, or unlicensed training are not all content companies but those that lack originality, lack real-time data, have unclear rights chains, have no data-interface capability, and depend heavily on search-traffic redistribution. This risk is most evident in news, general internet content, and low-differentiation footage libraries.

For narrower follow-up research directions worth digging into further, my suggested priority order is: professional-database licensing and AI workflows, UGC data licensing, news-content licensing and AI-search revenue sharing, image-library training data and safe generation, music AI licensing and likeness rights, enterprise RAG data governance, and content provenance and model compliance audit. These directions are closer to real profit pools than a vague "AI copyright."

Open questions and limitations also need to be made clear: a large number of licensing contracts still do not disclose amounts; many companies blend AI licensing revenue into "other revenue/licensing/subscription" without breaking it out; public financial evidence for the music, film/TV, and personality-rights tracks is clearly less than for news and professional databases; and the rules in jurisdictions such as China, South Korea, Japan, and Australia are still evolving. So at this stage the most important thing is not "who told an AI copyright story" but who has already proven: contracts, customers, revenue, renewals, and a workflow position.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.