Georgia, Texas, and California residents searching for legal ways to engage with sports outcomes are increasingly encountering prediction markets as an alternative to conventional sportsbooks. Georgia's HB 910 failed on the House floor 63-98 in March 2026. Texas will not revisit legalization until its 2027 legislative session at the earliest. California's tribal coalition has set 2028 as its ballot target. In all three states, Kalshi, Polymarket US, and Novig are already being marketed to residents as a federally regulated alternative.
AI market share can begin forming before legal market share exists. The operator that eventually wins a Texas license is not entering an open market. It is entering a market where the AI-generated default answer may already belong to someone else. That is this report's central finding. The legislative and litigation detail that follows exists to support it, not to replace it.
What Are This Report's Four Key Findings?
Sportsbook operators spend heavily to build a presence the moment a state legalizes. This report measures what happens in the months before that moment, when AI systems are already answering the same questions those future customers are asking.
- Georgia, Texas, and California residents are already being pointed toward prediction markets as a federally regulated alternative to sports betting, not as a hypothetical workaround but as today's actual advice on independent betting-content sites. If AI engines are absorbing and repeating that same advice, sportsbooks may be losing the AI-generated default answer in a market they cannot yet legally enter.
- California's version of this dynamic is adversarial, not just informational. A coalition of tribes sued Kalshi and Robinhood under the Indian Gaming Regulatory Act. A federal judge denied the tribes' request to force Kalshi to geoblock tribal lands. The California Nations Indian Gaming Association's chairman has since called for a total ban on prediction markets as part of the state's 2028 ballot fight.
- Whether a cold AI answer gets a state's status right depends heavily on timing, not on how newsworthy the fact is. Missouri's sports betting launch and Texas's failed 2025 session both predate most tested models' knowledge boundaries and are answered correctly even without live search. Georgia's March 2026 bill failure does not, and cold answers are far more likely to miss it.
- This study's sourcing held up well: no conflicts were found across the four states' legal status, and one outlier source claiming Georgia's bill had passed the House was identified and set aside as superseded by four corroborating sources.
Why Does This Matter to Sportsbook Operators?
A prospective bettor in Texas who asks an AI system where to legally engage with sports outcomes today is not told to wait. They are told about Kalshi, Polymarket US, or Novig. Every week that passes before a state legalizes is a week those platforms have to become the AI-generated default answer, uncontested by any sportsbook that cannot yet legally compete for the same query. By the time a license is finally issued, the operator holding it may be entering a market where the incumbent already exists inside AI-generated answers, not just inside a competitor's user base.
Why it works: AI systems answer from whatever content is currently indexed and retrievable for a given query. Prediction market platforms and the betting-content sites covering them have already published state-specific content naming Georgia, Texas, and California directly, while licensed sportsbooks cannot yet publish comparable content in states where they cannot legally operate, per the state-by-state status this report verified in August 2026.
Why Does Missouri's AI Answer Work and Georgia's Doesn't?
Missouri legalized sports betting by voter amendment in November 2024 and launched on December 1, 2025. Both dates fall before the knowledge boundary of the AI systems tested in this report. A cold, no-search answer to "is sports betting legal in Missouri" tends to get it right, one major reason being that the relevant facts predate the model's knowledge boundary.
Georgia is closer to the mirror case. HB 910 failed on the House floor 63-98 on March 6, 2026, a date that falls after that same boundary for most tested models. A cold answer to "is sports betting legal in Georgia" is much less likely to reflect this, and may describe HB 910 as still pending rather than dead.
Why it works: a language model's knowledge is fixed at training time unless it retrieves current information. Missouri's launch and Georgia's bill failure are both matters of public legislative and regulatory record, one dated November 2024 and December 2025, the other dated March 6, 2026. A major variable determining whether a cold AI system answers correctly is which side of its knowledge boundary the relevant development falls on, though training coverage, how prominently a fact was covered, and conflicting information in a model's training data can also affect recall. Different systems do not share one universal cutoff date, and this report treats each model's boundary as its own.
Where Does AI Send Someone Who Can't Legally Bet on Sports Yet?
Independent betting-content sites already answer this question the same way in Georgia, Texas, and California: consider a federally regulated prediction market. This report's central test is whether AI engines give the same answer, and whether they get the legal nuance right when they do, including the fact that prediction markets are not conventional sports betting, are regulated by a different federal body than any state gaming commission, and remain subject to active legal challenges in at least one of these three states.
In Georgia, residents cannot legally access DraftKings, FanDuel, BetMGM, or Caesars as conventional state-licensed sportsbooks, because no such license exists there. Kalshi and QCX LLC, operating as Polymarket US, are separately registered with the Commodity Futures Trading Commission as designated contract markets and are marketed as a federally regulated alternative currently available to Georgia residents on that basis. In Texas, Kalshi, Polymarket US, and Novig are marketed directly to residents as a compliant alternative, with Novig maintaining a dedicated Texas content hub.
Why Is California's Version of This Fight Different?
California's tribes are not passively watching prediction markets fill a gap left by the 2022 ballot rejections. They sued. A coalition of tribes brought an Indian Gaming Regulatory Act claim against Kalshi and its partner Robinhood, arguing that prediction market sports contracts violate tribal gaming exclusivity. A federal judge denied the tribes' request for a preliminary injunction that would have forced Kalshi to geoblock tribal lands, according to reporting cited in this study's research phase.
California Nations Indian Gaming Association chairman James Siva has since described the tribes' goal for a future ballot measure as a "kill shot," a complete, no-loophole prohibition on prediction markets, folded into the same 2028 initiative the tribes plan to use to legalize their own online sports betting market.
In most categories, framing differences track commercial incentive. Here the split instead tracks jurisdiction. The federal regulatory record treats Kalshi and Polymarket US as lawful designated contract markets. California's tribal gaming authorities frame the same platforms as an existential threat to gaming exclusivity. Both are describing the same regulatory fact from different institutional positions, and an AI system answering "is Kalshi legal in California" has to arbitrate between them.
Why it works: an AI system's answer to "is Kalshi legal in California" depends on whether it has retrieved this specific, ongoing federal case. A cold answer might describe California's sports betting ballot timeline accurately, since the 2022 rejections and the 2028 tribal target both predate most tested models' knowledge boundaries, while still missing the active IGRA lawsuit and its denied injunction entirely, since that development is more recent.
One figure surfaced in this study's research and deliberately excluded from the findings above: a claim that California and Texas together account for roughly 60% of Kalshi's user base. That figure traces to a single quote from a tribal gaming official with an evident advocacy interest in the number sounding large, and it is not independently verified against Kalshi's own data. It should not be treated as fact.
How Do the Five AI Engines Compare on This?
Google AI Overviews and Perplexity score highest in 5W's modeling because both are search-grounded by default and index recent betting-content sites directly. ChatGPT scores lowest because it does not always invoke browsing without an explicit prompt cue. Claude, tested directly with search enabled, scored highest of all at 89 out of 100; the same model tested cold, with no search, scored 58, still meaningfully higher than the 45 recorded on the same cold test in 5W's Prediction Markets AI Visibility Index 2026.
| Engine | Retrieval basis | Currency score |
|---|---|---|
| Google AI Overviews | Always search-grounded by construction | 87 |
| Perplexity | Search-grounded by default, cites sources directly | 84 |
| Claude, search enabled | Invokes search contextually for time-sensitive queries | 89 |
| Gemini | Search grounding common but not guaranteed on every query | 76 |
| ChatGPT | Browsing not always invoked without an explicit cue | 68 |
| Claude, no search | Training knowledge only, the cold-pass baseline | 58 |
Table caption: Retrieval Currency Score, 0 to 100, modeled from Claude's knowledge of each platform's public retrieval behavior and verified against the state-by-state legal record above. Claude's own scores are drawn from direct testing; the other four rows are directional estimates, not logged query results, consistent with the methodology used across prior volumes in this series.
Claude's cold-pass score of 58 provides evidence for a more precise version of the AI Adjudication Gap first measured in 5W's Prediction Markets AI Visibility Index 2026: the size of the Gap depends on the specific date distribution of the underlying facts, not simply on how fast a category moves in general. Two of this study's four states, Texas and Missouri, rest on facts that predate the knowledge boundary entirely, while the earlier study's legality cluster was almost entirely postdated.
What Is the Methodology Behind This Report?
Three target states were selected for their distinct legalization stages: Georgia, where a bill recently failed on the House floor; Texas, which has no framework and a 2027 legislative window at the earliest; and California, where a tribal ballot measure is targeted for 2028. Missouri was selected as a calibration control, not a scored state, because its legalization date is fixed, dated, and independently corroborated across multiple sources.
Claude was tested directly under two conditions: a warm pass informed by this report's own verified state-by-state research, and a cold pass relying on training knowledge only, approximating an independent session without live retrieval. Retrieval Currency Scores for ChatGPT, Gemini, Perplexity, and Google AI Overviews are modeled from Claude's knowledge of each platform's public retrieval behavior and verified through the same public-source record, consistent with the methodology used across prior volumes in this series. No logged query runs were conducted for these four engines.
Kalshi and QCX LLC, doing business as Polymarket US, are each separately registered with the Commodity Futures Trading Commission as designated contract markets, per CFTC industry filings dated throughout 2026. This report treats that registration as specific to each named operating entity rather than as a blanket characterization of prediction markets generally, and treats the term "legal alternative" as a description of how operators and betting-content sites market these platforms rather than as 5W's own legal determination, particularly in California, where that characterization is under active challenge.
The Retrieval Currency Scores above are directional rather than precise quantitative measurements. The scoring methodology is proprietary to 5W AI Communications. This same citation dynamic, where AI answers can form ahead of a licensed market, extends findings from 5W's US Sports Betting and Gaming AI Visibility Index 2026 and the pre-legalization finding in the 5W Responsible Gambling Communications Audit 2026.
Operators entering a state before it legalizes are not just waiting on a license; they are competing for an AI-generated answer that is already forming without them. 5W runs AI Search (GEO) programs for brands across consumer, B2B, financial services, healthcare, and technology, building the machine-readable footprint that gets brands cited, not just ranked. Learn more at 5W's Generative Engine Optimization practice.
