95% of Playtesters Enjoyed AI NPCs in Meaning Machine's Murder Mystery Game, Study Finds

Meaning Machine, the studio that promotes a "Game Conscious" Battle Banter tool on their website, partnered with the University of Bristol to study how players respond to generative AI-powered NPCs. The results landed in their favor. Of 68 participants who played the studio's noir murder mystery Dead Meat, 95% found the experience enjoyable and 97% described it as rewarding.

The participant group comprised 31 male, 31 female, and seven non-binary players. Each played Dead Meat for 20 minutes, interrogating AI-powered suspects by typing or speaking their own questions. The NPCs responded with generated dialogue delivered in voices that sit just close enough to human to register the distance. Seventy-five percent of participants said the game let them express themselves or make meaningful choices, and Dead Meat's team research noted that those figures held consistent across the different game types included in the study.

Players described the freedom to write their own questions as a source of satisfaction.

"I did find it really rewarding, just like making my own questions up for once."

— Study participant

That same openness produced friction for others.

"The freedom can be very exhilarating at first, and it feels immersive, and it feels cool. And it can also feel overwhelming."

— Study participant

"This research helps to ground what is otherwise quite an emotionally charged debate about AI in games. It does this by putting the player at the heart of the debate — asking what they feel about AI-powered experiences in practice, not just in theory."

— Dr Richard Cole, lead researcher

Meaning Machine co-founder Thomas Keane drew a distinction between AI that displaces creative work and AI that creates interactions no traditional system could support.

"Players kick back at AI that is taking away from creativity. But when AI is used to power totally new types of interactive experience, then it's a very different story."

— Thomas Keane

The backlash Keane references is concrete. Eurogamer docked points from the extraction shooter Arc Raiders specifically for using AI-generated voice lines. BAFTA-winning actor Jane Perry has spoken publicly about the threat generative AI poses to performers working in games.

I think the study design does most of the argumentative work for Meaning Machine here — 68 players, 20 minutes each, self-selected into a research context where they already knew AI NPCs were involved before reporting their enjoyment. That is a narrower claim than the headline figures suggest.

The 20-minute session length is acknowledged as a limitation in the study itself. Researchers noted they do not yet know how players respond across longer play sessions, though they recorded that participants wanted to continue. A full paper covering both Dead Meat and Meaning Machine's follow-up title Blood Will Out is due at the end of the year. The Blood Will Out data was collected at the same volume and involves a more advanced version of the studio's technology, according to researchers.

Whether positive short-session responses translate into sustained player engagement with AI-driven NPCs remains untested. The study confirms players did not reject the format outright in a controlled setting. It does not confirm they would choose it over a game built around writing from actual authors. I see the distinction as significant, particularly as studios weigh AI implementation against the cost of the writer and voice talent it tends to replace.