Replika's marketing leans on outside studies, not its own trial

The design

Replika, the companion app made by Luka, Inc., publishes a page at replika.com/about/story that, as retrieved on 16 September 2026, tells the company's own account of its history and states: 'When the world's leading researchers wanted to understand whether AI could actually change someone's life, they studied Replika... at Stanford, at Harvard, across years of research... Replika doesn't replace human connection. It deepens it.' The page then names two specific citations to support that claim: Guingrich & Graziano, described on the page as arXiv 2311.10599, 2023, and Ta et al., JMIR 2020.

What the evidence says

Neither paper Replika names is company-funded or company-authored. The Guingrich & Graziano paper, 'Chatbots as social companions', by Rose E. Guingrich and Michael S. A. Graziano, was later published as a chapter in Oxford Intersections: AI in Society (Oxford Academic, March 2025). Its own abstract describes a study comparing people who regularly used companion chatbots with people who did not, finding that users rated the relationships as beneficial to their social health while non-users viewed the same relationships as harmful – a correlational comparison of self-reported opinion, not a randomised trial, and not a study of Replika alone. The company's page compresses this into 'researchers... found... it deepens' connection, without the comparison, the correlational caveat, or the fact that the researchers studied companion-chatbot users broadly rather than assigning new users to try Replika.

What it asks of people

A visitor reading the story page is asked to accept a company's own selection and compression of independent research as evidence for a specific benefit – deepened human connection – that the cited paper itself frames far more narrowly, as a self-reported and comparative finding among people who already chose to use these apps.

Privacy and safeguards

The story page is a living marketing document, not a filing reviewed by any regulator for accuracy, and it can change or add citations at any time; the description here reflects the page as retrieved on 16 September 2026. Nothing on the page discloses what data Replika itself collects from the conversations that supposedly produce this benefit, a separate question the company's privacy policy, not its story page, is meant to answer.

• Does citing independent research that studied companion-chatbot users in general support a specific claim about Replika?
• What would change in the company's framing if the non-users' more negative view, also reported in the cited paper, were included?
• Should a reader expect a company's own summary of outside research to carry the same caveats the original authors stated?

Real, independently authored research does exist and does report broadly positive self-assessments among people who already use companion chatbots; that is a narrower and more conditional finding than the company's own page presents it to be.