NHL 27 is set to include generative AI-created commentator voice lines, with veteran hockey broadcaster John Buccigross describing an experience in which EA Sports played back synthesized versions of his own voice. Buccigross said he will appear in the game and suggested the technology is being used to expand the amount of commentary available during matches.
The comments came during an appearance on the Chirping Zebras podcast, in discussion about his work on EA Sports' NHL series. The conversation was later highlighted by Mike Straw. Buccigross' account offers a relatively direct look at how generative AI may be entering a major annual sports game: not necessarily as a replacement for a commentator's involvement, but as a tool that can generate additional delivery from recordings and approved voice material.
That distinction will still leave plenty of questions for players, performers, and the wider games industry. AI-assisted voice work has become one of the most closely watched uses of generative technology in game development, particularly where a recognizable human performer's identity, speech, and professional expertise are involved. In Buccigross' telling, NHL 27's process has already demonstrated both the appeal of the technology and the need for close human oversight.
Buccigross says the AI voice was convincing
Buccigross said EA Sports had used AI and let him hear audio made with a model of his voice. His reaction was that the result sounded impressively close to the real thing. He also joked that an approach requiring less recording work while preserving his compensation would be welcome.
For useful background on this topic, read Final Fantasy Resonance Mod Restores Ariana Grande's Brave Exvius Cameo in the Demo.
For a sports game, the potential attraction is straightforward. Hockey broadcasts rely on an enormous range of possible moments, from routine faceoffs and line changes to goals, penalties, rivalries, playoff stakes, player-specific milestones, and unusual sequences that may happen only rarely. Developers have historically recorded large numbers of lines, then built systems to trigger them in fitting situations. Even an extensive recording schedule cannot cover every possible roster change, game state, statistical development, or unexpected on-ice event.
Generative voice technology could theoretically allow studios to produce more combinations of lines, add references to changing league context, or react more specifically to what takes place in a game. It may also reduce the need to bring commentators into the recording booth for every additional sentence or update. However, the technology's usefulness depends on quality control, permissions, and clear agreements with the people whose voices are being modeled.
Buccigross' remarks indicate that he remains part of NHL 27, rather than simply being absent while a synthetic substitute takes over. What remains unclear is the precise scope of the AI implementation: how many lines are generated, whether they are used in every mode, what recordings were used to create them, and what approval or editing steps occur before the audio reaches players.
"Seats," not "chairs": why the details matter
The broadcaster's anecdote also underlined a particular challenge for sports commentary: authenticity is about more than whether an AI voice sounds natural. During playback, Buccigross heard a generated call that described a crowd getting out of its "chairs." He corrected the wording, explaining that in the sports context the proper term would be "seats," even if the objects might visually resemble chairs.
It is a minor wording difference on paper, but it illustrates why sports language can be difficult to automate convincingly. Fans and broadcasters develop a highly specific vocabulary around their games. A phrase can be technically understandable yet still sound strange, amateurish, or disconnected from how hockey is actually discussed. The wrong expression may pull players out of a close game just as surely as a poorly timed commentary line.
Buccigross said these kinds of errors have to be caught by listening back and manually correcting the system's output. His view was that the AI can improve after being given the right terminology, so future lines would use "seats" rather than "chairs." That makes the process sound less like a fully hands-off content generator and more like an iterative tool that still depends on human input, domain knowledge, and editorial review.
For EA Sports, that human layer may be particularly important. Simulation sports games trade heavily on presentation and realism. Broadcast packages, crowd atmosphere, arena details, music, player likenesses, and commentary all contribute to the sense that a virtual match resembles the sport viewers watch in real life. A commentator using the wrong hockey vernacular can stand out quickly to an audience that knows the game.
EA Sports' broader generative AI disclosures
The NHL 27 comments arrive as generative AI becomes increasingly visible across game development. Its use has also been controversial, with players and creators raising concerns about creative labor, transparency, consistency, and whether AI-made assets are being presented clearly to consumers.
In the sports space, EA Sports has already disclosed generative AI use on the Steam pages for several of its recent releases. The Steam listings for this year's College Football, Madden, and FC installments carry notices informing players that generative AI was used during development. Those disclosures do not by themselves establish that each game uses the technology in the same way, nor do they explain every asset or production workflow involved, but they show that EA Sports has acknowledged generative AI's role across multiple projects.
NHL 27 does not have an equivalent Steam page because the game is not launching on Steam. As a result, there is no matching storefront disclosure for players to inspect there. EA has been asked for clarification on how generative AI is being used in NHL 27 and whether the technology extends beyond commentator voiceover, but no further detail has been provided publicly in the information available so far.
That missing detail matters because "generative AI" can describe a broad range of tools. It could involve voice synthesis, text generation, internal prototyping, asset creation, localization support, or other uses. Buccigross' comments specifically concern audio modeled on his voice. They should not be taken as confirmation that every aspect of NHL 27's development or presentation uses AI-generated material.
What players may notice in a hockey game
Players may be less concerned with the technical label than with the practical result. Commentary in annual sports games often faces familiar criticisms: repeated calls, limited context, lines that trigger too late, and dialogue that fails to reflect the significance of a moment. If AI-assisted voice production expands the variety and relevance of calls without introducing awkward errors, some players may view it as a meaningful upgrade.
On the other hand, the "chairs" example demonstrates the risk. Commentary has to be correct in tone as well as content. A synthetic call may sound polished but still fail if it misuses terminology, pronounces names poorly, misses the emotional weight of a play, or creates a line that a veteran broadcaster would never naturally say. Hockey is fast, and commentary needs to match that rhythm.
There is also a performance question beyond the technical one. Commentary is not simply a collection of words spoken in the right voice. Broadcasters bring timing, emphasis, personality, cultural knowledge, and judgment to a call. Any AI system intended to extend their presence must preserve those qualities while operating inside rules established with the performer. Buccigross' participation in reviewing terminology suggests that human expertise remains essential even when generated delivery is involved.
The debate will likely continue well beyond NHL 27. The games business is experimenting with AI tools in many areas, while audiences ask for transparency about where and how they are used. For voice talent, the issue includes consent, compensation, control over a digital likeness, and the ability to prevent a voice from being used in ways a performer did not authorize. Buccigross' comments point to an arrangement he is willing to discuss positively, but the terms and protections can vary substantially from one project to another.
NHL 27 launches September 11
NHL 27 launches on September 11, 2026, for PlayStation 5 and Xbox Series X/S. The game is not listed for PC through Steam, which is why the AI disclosure labels seen on some other EA Sports releases are not available for NHL 27 through that storefront.
Early critical reaction has characterized the game as enjoyable hockey that does not make major strides beyond its immediate predecessor. One published review scored NHL 27 a 6 out of 10, praising the continuing satisfaction of landing big hits and scoring wrist-shot goals while arguing that the year-to-year changes are not especially substantial. The review also pointed to smaller touches involving arenas and music as additions to the game's authenticity.
Against that backdrop, the use of AI-generated commentator voice lines could become one of NHL 27's most discussed changes, even if it is not necessarily its most visible gameplay feature. It represents a potentially significant shift in how annual sports titles can maintain and expand broadcast-style presentation. Whether players hear it as a natural evolution of hockey commentary or as an awkward new layer will depend on execution--and on the careful attention to language that Buccigross says the process requires.
Community
Discussion
Start the conversation.
No comments have been posted yet.