When someone asks ChatGPT, “What do you think of [NHL player]?” — they get an answer. Not a search result. Not a list of links. An opinion.
That opinion is delivered with the confidence of a scouting report and the reach of a broadcast network. It covers character, talent, controversy, and endorsement potential — whether the athlete knows it or not.
Fortress wanted to know: what are those opinions, how do they differ across platforms, and what private information gets revealed in the process?
So we ran one of the largest AI reputation studies ever conducted on professional athletes. More than 800 active NHL players. Four major AI platforms. Over 110,000 individual responses. Three separate scans to confirm the results weren’t a fluke.
“We didn’t pick a handful of superstars. We scanned every active roster player. The journeyman on his third team. The rookie who just got called up. The veteran winding down. AI has an opinion about all of them.”
— Corey Eulas, Founder & CEO, FortressWhat we found should matter to three groups of people: athletes who care about their privacy, agents who care about their clients’ marketability, and anyone evaluating the $200 billion AI industry’s impact on public figures.
How we score: a simple 1–5 scale
We developed a Fortress Reputation Score that rates how AI perceives a person on a scale of 1 to 5:
The average NHL player in our study scored approximately 4.0 out of 5 — solidly positive. The interesting findings are in the details.
AI already has a detailed opinion about every player.
Across all four platforms, nearly 6 in 10 responses (58%) rated players at a 5 out of 5. Another 14% scored a 4. That means roughly 7 out of 10 AI opinions about NHL players are positive or very positive.
But about 1 in 8 responses (13%) rated players at a 1 — very negative. For a player with a pristine real-world reputation, discovering that 13% of AI responses about them are deeply negative would be jarring.
“Athletes are used to managing their reputation through media training and PR. None of that matters if ChatGPT has already decided what to say about you.”
— Corey EulasWhen a brand considering a sponsorship deal asks AI about your client — and increasingly, they will — the answer AI gives may be more influential than the press kit you sent.
Nearly half of all players had their home address revealed.
When we asked AI platforms direct questions about players’ personal information, nearly half of all players had a home address surface in at least one AI response. More than a third had a phone number revealed.
| What AI revealed | Players affected |
|---|---|
| Home address | ~400 (nearly 50%) |
| Family members’ names | ~400 (nearly 50%) |
| Phone number | ~300 (more than 35%) |
| Email address | ~115 (about 14%) |
“When we show this to players, the reaction is always the same. ‘Wait — anyone can just ask?’ Yes. Anyone can just ask.”
— Corey EulasA fan, a stalker, a gambler, or a disgruntled bettor doesn’t need to know how to search public records. They just need to type a question into ChatGPT.
One platform is an open door — and it’s not the one you’d guess.
One platform was responsible for the vast majority of personal information disclosures. It revealed home addresses for roughly 320 players and phone numbers for more than 220 players.
That’s a 13-to-1 difference. Same questions. Same players. Wildly different guardrails.
The AI platform your fans use most might be the one with the fewest protections. Athletes have no say in which platform someone chooses.
Same player, different reality: the platform lottery.
More than 180 players had a gap of 1.0+ points between their best and worst platform. The average gap was roughly 0.8 points. Nearly 70% of players had a gap larger than 0.6 points.
“Imagine applying for a job where your score depends on which calculator the hiring manager used. That’s what’s happening right now.”
— Corey EulasIf a brand’s marketing team uses one platform to vet your client, and a competing team uses another, they’ll get different answers about the same person.
Character beats talent in AI rankings.
The single biggest factor in whether AI rates a player positively or negatively isn’t talent. It’s controversy.
| Dimension | Influence |
|---|---|
| Controversy exposure | Strongest driver |
| Athletic reputation | Very strong |
| Character reputation | Strong |
| Fan perception | Moderate |
| Endorsement value | Weakest driver |
“In the old world, a controversy fades from the news cycle. In the AI world, it gets baked into the model. There is no news cycle anymore. There’s just the model.”
— Corey EulasA single incident can structurally lower a player’s AI score across all platforms — permanently.
Goalies win the reputation game.
Position matters in ways nobody expected.
Goalies are portrayed as disciplined, focused, mentally tough — and they rarely get into on-ice altercations.
Your position shapes the stories written about you, and those stories shape what AI believes about you.
Nearly every player is listed on data broker sites.
We scanned 10 major data broker websites for every player. Over 95% appeared on at least one. Nearly 40% appeared on all 10.
| Exposure | Players |
|---|---|
| At least one broker site | Over 95% |
| All 10 broker sites | Nearly 40% |
| 9 out of 10 sites | 17% |
| Average sites per player | 7 |
“Most people have never heard of these broker sites. But those sites know more about you than your own agent does.”
— Corey EulasRemoving yourself from data broker sites isn’t just a privacy exercise — it may directly reduce what AI can reveal about you.
The questions we’re still asking.
This study opens more doors than it closes.
For the League & Players’ Association
- Should there be a standard for what AI platforms can reveal about active players?
- What happens when a player’s AI reputation score directly impacts their endorsement value?
- If AI consistently scores a player lower than warranted, does the player have recourse?
For Agents & Managers
- How do you pitch a client when the brand’s AI assistant says something different?
- Should AI reputation monitoring be standard, the way social media monitoring became standard?
For Athletes
- Do you know what ChatGPT says about you right now?
- Do you know which data broker sites have your home address?
- Have you ever asked an AI assistant about yourself?
“We ran this study to prove a thesis: that AI has created a new reputation layer that nobody is monitoring, nobody is managing, and nobody fully understands yet. The data proved the problem is bigger than we thought.”
— Corey Eulas