Does using AI to judge a contest make it an illegal lottery?

It can, depending on how the judging works. A skill contest avoids lottery classification because winners are selected on merit against stated criteria rather than by chance. If a model scores entries in a way the operator cannot explain, cannot reproduce, and cannot tie to published criteria, a regulator or plaintiff can argue the selection was effectively random. Combine that with a prize and any consideration and you have the three elements of a lottery. The fix is not to avoid AI, but to keep judgement anchored to published criteria with human review of the outcome.

Can we prohibit AI-generated entries?

Yes, and if you intend to you must say so in the official rules before entries open, because rules generally cannot be changed mid-promotion to a participant's detriment. The harder problem is enforcement: detection tools are unreliable enough that disqualifying an entrant on a detector score invites a dispute you may not win. Operators who care about this usually pair the prohibition with a warranty from the entrant and a right to disqualify for breach, which shifts the question from detection to representation.

Who owns an AI-generated contest entry?

Purely machine-generated material may not carry copyright at all, which means the entrant cannot assign what they do not hold and the brand cannot obtain exclusivity through the usual rights grant in the rules. If the promotion exists to acquire usable creative assets, that undermines the point of running it. Rules should require the entrant to warrant they hold the rights they are granting and disclose material AI assistance, and the marketing plan should not assume exclusivity in an entry that may be uncopyrightable.

Do registration and bonding requirements change if AI is involved?

The triggers do not change: certain states require registration and bonding for promotions above a prize-value threshold, on the ordinary timeline before the promotion opens. What AI changes is the classification analysis feeding into them. A promotion the operator considers a skill contest, and therefore outside the sweepstakes registration regime, is exposed if the judging method makes the skill characterisation hard to sustain. Decide the classification before you build the judging pipeline, not after.

Can AI screen entries for eligibility?

Automated screening for rule violations and eligibility is common and generally fine, but it is an automated decision affecting an individual, and it sits inside the same disclosure and appeal expectations as other automated decisions. Practical safeguards: log why each entry was rejected, provide a route to human review, and test the screen for systematic exclusion of a group. A screen that quietly rejects entries in a particular language or format creates a discrimination problem that has nothing to do with gambling law.

AI-Judged Contests and Sweepstakes 2026: When a Skill Contest Becomes an Illegal Lottery

The Structure Marketing Teams Inherited

Prize promotions run on a narrow legal structure. A lottery requires prize, chance, and consideration; a private operator running one without authority is running an illegal lottery. Promotions escape by removing an element. A sweepstakes removes consideration — free entry, no purchase necessary. A contest removes chance — winners are chosen on skill against published criteria. Those are the two shapes, and each carries its own obligations around rules, disclosure, registration, and bonding.

AI reached this area through the judging step, which is precisely where the contest structure is load-bearing. A team that would never have replaced its judging panel with a coin flip will comfortably replace it with a model score, without noticing that the legal defensibility of the promotion depends on being able to explain the difference.

Where AI Touches a Promotion

HIGH RISK

Model selects the winner with no reviewable rationale

A score with no mapping to published criteria and no human confirmation. Weakest position on the skill element.

HIGH RISK

AI-generated entries with an exclusivity plan attached

The brand intends to reuse winning creative, but machine-generated material may not carry the copyright the rights grant assumes.

DEFENSIBLE

Model shortlists, named judges decide

AI narrows the field on stated criteria; humans apply the criteria to select winners and can explain the result. Defensible.

DEFENSIBLE

Automated eligibility screening with human appeal

Rule-violation and duplicate detection, logged, with a documented route to human review of a rejection.

Running an AI-Assisted Promotion Without Breaking the Structure

1. Fix the Classification First

☐Decide sweepstakes or contest before designing the judging pipeline, not afterwards
☐If it is a contest, write judging criteria specific enough that a third party could apply them
☐If it is a sweepstakes, confirm the free-entry route works without the AI step in the path
☐Check registration and bonding thresholds for the states you are open to, on their pre-open timeline

2. Keep Judgement Explainable

☐Map every model score to a published criterion; discard scores that do not map
☐Keep a named human judge who confirms winners and can state the reason for the selection
☐Retain the scoring record for each finalist through the dispute window
☐Disclose in the rules that automated tools assist judging, and how

3. Handle AI-Generated Entries Explicitly

☐State whether AI-assisted entries are permitted — before entries open, not mid-promotion
☐Require the entrant to warrant they hold the rights they are granting you
☐Require disclosure of material AI assistance rather than relying on a detector score
☐Do not build a downstream campaign that assumes exclusivity in an entry that may be uncopyrightable

4. Treat Screening as an Automated Decision

☐Log the reason for every automated rejection in a form you could show an entrant
☐Provide a human-review route and staff it for the duration of the promotion
☐Test the screen for systematic exclusion by language, format, or accessibility need
☐Keep the entry route usable for entrants using assistive technology

The Rules Document Is the Contract

Official rules are enforceable terms between the operator and every entrant, and they are generally fixed once entries open. That timing is what makes AI decisions expensive to defer: a prohibition on AI-generated entries, a disclosure that automated tools assist judging, and a rights grant calibrated to uncertain copyright all have to be in the document at launch. Discovering the issue at judging time leaves the operator choosing between a rule it never published and a promotion it cannot cleanly complete.

There is also a straightforward advertising-law overlay. Prize value, odds, and eligibility statements are marketing claims subject to the same substantiation and deception standards as any other claim, and generated promotional copy that overstates any of them creates exposure independent of how the winner was chosen.

Frequently Asked Questions

Is a purely internal, employees-only contest exempt?

Employee promotions sit outside consumer promotion law in some respects, but they pick up employment-law issues instead. An AI-judged internal competition tied to recognition, bonuses, or advancement is an automated decision about employees, with the discrimination-testing and record-keeping expectations that attach to those. The analysis changes; it does not disappear.

Does an AI chatbot running the entry flow create additional exposure?

It creates the usual chatbot problems inside a promotion: a bot that misstates odds, eligibility, deadlines, or prize value has made a marketing claim the operator has to stand behind. Restrict the bot to retrieved facts from the official rules rather than generated summaries of them, and make sure the rules themselves are one click away at every step.

Can we use AI to pick winners in a pure sweepstakes?

A sweepstakes needs a genuinely random selection, so the question is whether the mechanism is auditable randomness. A conventional random draw is easy to document and easy to defend. A model-mediated selection is neither, and it introduces the possibility of systematic bias in something that is legally required to be chance. Use a documented random selection method and keep the model out of the winner draw.

How long should we keep the judging records?

Through the dispute window stated in your rules and any applicable state limitations period, whichever is longer, and longer still if a challenge is raised. The records that matter are the ones that let you show a winner was selected against published criteria: entry, score, criterion mapping, and the human confirmation. Keep them together rather than scattered across the tooling that produced them.

Why machine-generated material may carry no copyright to assign.

AI-Generated Reviews and the FTC Testimonial Rule 2026

The advertising-law standard that applies to generated promotional content.

Rights clearance for campaign assets, including user-submitted material.

Reusing entrant and creator content across owned channels.

Decide the Judging Method Before You Publish the Rules

Every AI decision in a promotion has to be made before entries open, because the rules document freezes at that moment. Classification, judging method, AI-entry policy, and rights grant are all launch-day decisions with no clean retrofit.

The entry form itself is worth checking too: a promotion open to the public that cannot be completed with a keyboard or a screen reader excludes entrants for reasons that have nothing to do with your judging criteria.

A Contest Is Legal Because Someone Can Explain Why That Entry Won