College sports spent five years turning name, image, and likeness (NIL) into a regulated, reportable, licensable market. AI doesn’t need to invent the model. It needs to copy the infrastructure.
Name, Image, and Likeness (NIL) is not just a college-sports story. It is the first scaled proof that a person’s identity can become a licensable economic asset once the market has clear rights, intermediaries, pricing, reporting and enforcement.
The most important comp for licensed AI likeness is not a chatbot. It is a quarterback signing an endorsement deal.
Since 2021, college athletes have been able to earn money from their name, image and likeness. What started as a messy patchwork of collectives and sponsorships has evolved into a market with national compliance rails, formal reporting, pricing review and direct institutional compensation.
That matters because the hard question behind AI likeness is not “will people pay?” It is “can identity be turned into a permissioned, trackable commercial right at scale?” NIL already answered yes.
The athlete controls commercial use of their name, image and likeness rather than surrendering identity outright.
Collectives, agencies, schools and marketplaces connect the rights-holder to brands and counterparties.
Division I third-party deals at the reporting threshold run through NIL Go, creating a compliance record instead of an invisible handshake.
Deals can be reviewed for valid business purpose and whether compensation sits within a reasonable range.
The College Sports Commission oversees settlement-related financial rules and third-party NIL compliance.
Brands, schools and other counterparties pay for licensed access to identity — exactly the economic behavior AI likeness needs.
The breakthrough was not letting athletes “be influencers.” The breakthrough was building a system where identity itself became a governed commercial asset.
| Layer | College NIL | Licensed AI Likeness |
|---|---|---|
| Asset | Name, image, likeness | Face, voice, persona, style |
| Rights holder | Athlete | Creator / public figure / individual |
| Buyer | Brand, school, sponsor | AI platform, brand, fan product |
| Intermediary | Collective, agency, marketplace | Rights registry / licensing platform |
| Compliance | NIL Go + CSC rules | Consent ledger + federal/state likeness rules |
| Pricing logic | Comparable market value + deliverables | Scope + duration + channels + synthetic usage |
| Enforcement | Deal review / eligibility consequences | Revocation / takedown / access termination |
Delphi proves people may want access to an AI version of a person. NIL proves a regulated likeness market can actually clear transactions at scale. One is product-market evidence. The other is market-structure evidence.
One of the most interesting 2026 developments is that formalization did not shrink NIL. It expanded it. Opendorse revised its 2026–27 NIL estimate from roughly $2.8B to $4.5B, arguing that direct school payments created a base layer while commercial NIL became an additional channel above it.
That is the lesson for AI likeness: regulation does not necessarily kill monetization. Clear rules can unlock more counterparties because buyers know what they are allowed to purchase, rights-holders know what they are giving up, and intermediaries can underwrite the transaction.
The obvious company to build is a better AI clone. The more durable company may be the one that sits underneath every clone: a registry that can prove consent, define usage rights, price a license, route payouts, log derivative uses and revoke access.
NIL shows why. Once transactions become large enough, everyone needs the boring layer: compliance, records, valuation and enforcement. That layer becomes more valuable as the number of apps on top increases.
An athlete endorsement is bounded. A synthetic likeness can generate thousands of outputs after one license.
A creator’s AI can say something the creator never said. NIL deals usually do not create autonomous speech.
AI raises questions around training, remixing, fine-tuning and downstream outputs that classic endorsements do not.
Taking down one ad is easy. Pulling a likeness from models, caches and third-party products is not.
The business model has already been tested in college stadiums, brand campaigns and compliance portals. What AI changes is the frequency, scale and programmability of the licensed identity.