Search once required a person to locate, compare, and interpret scattered sources. An AI assistant can now assemble an answer about a player, their reputation, and potentially sensitive personal information in seconds. That changes the practical risk, not because AI creates every underlying record, but because it makes dispersed information easier to ask for and use.
Executive summary
Fortress evaluated the information and opinions returned by four leading AI platforms when asked standardized questions about more than 800 active NHL players. Across three scan periods, the study collected more than 110,000 responses addressing reputation and personal-information exposure.
The central finding is straightforward: AI systems are now highly accessible interpreters of an athlete's public footprint. They do not simply point to source material. They formulate an answer, often confidently and conversationally, that can shape a viewer's perception of a player's character, marketability, and personal life.
For a player and family, this is a privacy and safety issue. For a wealth manager, agent, or family office, it belongs in the wider client-risk picture. For a team security department, it is another way public information can become operationally useful to someone with unwanted intent.
AI provides substantive opinions about virtually every player examined.
Responses commonly went beyond basic biographical information. They characterized athletic ability, professionalism, personality, fan perception, controversy, and endorsement potential. On the Fortress Reputation Score, a standardized 1-to-5 assessment of those dimensions, the average player score was approximately 4.0. Fifty-eight percent of responses received the highest rating and another 14% were rated positive.
This favorable average should not be mistaken for low risk. Thirteen percent of responses rated a player 1 out of 5, the study's very-negative category. Even in a generally positive dataset, roughly one in eight AI assessments was deeply unfavorable. The same individual could also receive materially different assessments depending on which platform a user consulted.
AI is not just cataloging a player's statistics. It is retelling the story around that player, and its retelling can vary by platform.
Personal information surfaced at a level that warrants attention.
When prompted about personal information, AI systems returned information associated with players that could affect family privacy and security planning.
| Information surfaced in at least one response | Players affected |
|---|---|
| Home address | ~400 (nearly 50%) |
| Family members' names | ~400 (nearly 50%) |
| Phone number | ~300 (more than 35%) |
| Email address | ~115 (about 14%) |
The exposure is not necessarily created by AI. In many cases, the underlying information was already available in public records, data-broker listings, or other sources. What AI changes is the access model: it reduces the effort required to discover and assemble information that previously sat across multiple places.
Platform safeguards vary substantially.
Identical prompts produced materially different privacy outcomes across platforms. One platform returned home addresses for roughly 320 players; the platform with the lowest address exposure returned approximately 25. That is a difference of roughly 13 to 1.
This is why there is no single AI environment to manage. A player cannot choose which platform a fan, bettor, journalist, prospective partner, or bad actor will use. Durable privacy work must reduce exposure at underlying sources and continue to observe what the AI layer returns.
Reputation is inconsistent across AI systems.
More than 180 players showed a gap of at least 1.0 point between their highest- and lowest-scoring platform results. The average spread was roughly 0.8 points, and nearly 70% of players showed a spread larger than 0.6.
That inconsistency matters as third parties use AI to accelerate ordinary research. A brand evaluating a potential partnership, a concierge team screening a request, or a business contact conducting quick diligence can receive a different characterization of the same player depending on the system used.
Controversy has an outsized effect on AI-generated reputation.
Among the factors evaluated, controversy exposure was the strongest driver of unfavorable AI assessments. Athletic reputation and character were also meaningful inputs; endorsement value was comparatively weak. In practice, this suggests that AI systems can preserve and reframe past narratives long after the original coverage has faded from everyday attention.
Position is associated with materially different AI reputation profiles.
The study found a notable position-level pattern: goalies averaged 4.5, compared with 4.1 for forwards and 4.0 for defensemen.
This is not evidence that one position produces better people. It is evidence that AI absorbs durable public narratives: goalies were more often framed as disciplined, focused, and mentally tough, while skaters more often carried narratives linked to physicality, conflict, or controversy.
Data brokers are a persistent upstream source of exposure.
| Data-broker exposure across 10 sites | Players affected |
|---|---|
| Listed on at least one site | Over 95% |
| Listed on all 10 sites | Nearly 40% |
| Listed on 9 of 10 sites | 17% |
| Average sites per player | 7 |
These listings are not only a privacy concern in their own right. They are part of the broader information environment from which searchable and AI-accessible narratives are built.
What these findings mean in practice
The operating question
The appropriate response is not alarm. It is ownership. High-performing individuals already use specialists to manage investments, taxes, insurance, health, and physical security. Digital privacy needs the same basic discipline: identify exposure, establish priorities, take action at the source, verify results, and monitor for recurrence.
A practical path from observation to protection
- Establish the current picture. Review what search engines, AI assistants, data brokers, and relevant public sources actually reveal.
- Prioritize by consequence. Separate items that create a genuine family-privacy or security concern from ordinary public information.
- Address the source. Where removal or correction is appropriate, work on the underlying records and publishers, not just the surface result.
- Verify and continue monitoring. Recheck whether the exposure has changed and maintain visibility as AI models, sources, and circumstances evolve.
Where Fortress fits
Fortress is built for the operating discipline behind that work: helping high-value individuals and the people who advise them understand digital exposure, coordinate remediation, and keep a clear record of what has been addressed and what still requires attention.
The goal is not to make the internet disappear or promise control over every AI response. It is to replace uncertainty with a practical, ongoing program; one that respects the difference between public visibility and unnecessary exposure, and gives players, families, and their trusted teams a clearer basis for action.
Methodology and limitations
Scope. The study included more than 800 active NHL roster players across all 32 teams. Four major AI platforms (ChatGPT, Gemini, Grok, and Perplexity) were evaluated using an identical set of prompts. Fortress conducted three separate scans over a multi-week period and collected more than 110,000 individual responses. Data-broker exposure was assessed across 10 major broker websites.
Scoring. The Fortress Reputation Score is a composite assessment of athletic reputation, character, fan perception, endorsement value, and controversy exposure, normalized to a 1–5 scale.
Limitations. AI outputs can change with prompt wording, model updates, source availability, geographic context, and time. Standardized prompts and repeated scans were used to reduce variability, not eliminate it. Data-broker findings reflect point-in-time observations. This study describes observed outputs; it does not establish that every disclosed item was current, accurate, or sourced directly from a particular publisher.
Conclusion
AI is becoming a routine interface between the public and a professional athlete's information. The consequences are not confined to online reputation. They extend to family privacy, personal security, commercial diligence, and the stewardship of a high-value personal profile.
The question is no longer simply whether information can be found online. It is what an AI assistant can say today, what it may reveal, and whether someone is accountable for understanding and reducing unnecessary exposure.