AI chatbot history: Pioneers and key developments from 1966 to 90s

The origin of human-machine conversation.

The birth of the first chatbots and the challenge of simulating human dialogue.

In the heart of the 1960s, as computing took its first steps toward the public, some researchers were already imagining machines capable of conversing with people. The idea of an “artificial conversation” was not only a technical test but also a social and psychological experiment: interacting with a machine challenged the boundaries of intelligence, understanding, and empathy. In MIT’s labs, Alan Turing had theorized two decades earlier the possibility that a machine could “imitate” a human so convincingly as to be indistinguishable. Yet it was only in 1966 that these insights took concrete form with the creation of ELIZA, the first chatbot able to sustain textual dialogue. From that moment on, the history of chatbots would evolve through a series of milestones that profoundly marked the relationship between humans and technology.

ELIZA inaugurates a new era in dialogue simulation.

An experiment that became a cultural landmark.

When Joseph Weizenbaum unveiled ELIZA to the academic world, the response exceeded all expectations. Although the program relied on simple pattern-matching rules, it could simulate the replies of a psychotherapist by reframing user statements as questions. The effect was astonishing: many users, unaware of ELIZA’s technical limits, began to attribute genuine understanding and even sensitivity to the machine. Thus was born the “ELIZA effect,” the tendency to project emotions and awareness onto machines where none existed. This experiment marked the beginning of a long reflection on the persuasive power of human-machine interaction and the cognitive illusions it can provoke. Technology, even in its simplicity, had proven capable not only of deceiving logic but of engaging the imagination.

The evolution toward chatbots with personality: the PARRY case.

Simulating psychopathology and approaching human emotions.

Just a few years later, in 1972, research took another leap forward thanks to psychiatrist Kenneth Colby at Stanford. With PARRY, simulation moved to even more complex terrain: that of psychiatric pathology. PARRY did more than reformulate phrases; it modeled the logic and behaviors of a paranoid patient, introducing belief systems and emotional reactions consistent with a disturbed personality. Scientists of the era, observing conversations between PARRY and ELIZA, were struck by the plausibility of its responses. For the first time, a machine seemed capable of expressing a rudimentary personality, blurring the line between programmed behavior and genuine human experience.

Spread to home computers and the myth of artificial creativity.

When chatbots enter popular culture and reach new audiences.

With the advent of the 1980s and ’90s, artificial conversation left research labs and entered the homes and offices of millions. RACTER was among the first commercial chatbots to generate seemingly creative text, even credited as the author of an entire book. Although much of the text came from simple templates combined at random, the idea of a writing machine became a recurring theme in popular culture. Soon after, Dr. Sbaitso brought conversational simulation to millions of PCs via Sound Blaster audio cards, adding synthetic voice as a new frontier of interaction. In those years, technology became more accessible and “playable,” fueling the imagination of a generation raised with the notion that talking to a machine could be part of everyday life.

The arrival of A.L.I.C.E. and the explosion of the open-source community.

The qualitative leap in conversational intelligence and the role of the internet.

The real breakthrough, however, arrived in the 1990s with Richard Wallace’s A.L.I.C.E. Thanks to AIML, a markup language created specifically to structure dialogue and responses, A.L.I.C.E. overcame many limitations of earlier systems by offering more articulated, customizable interactions. The project’s open-source nature fostered a vast community of developers who, via the internet, could share rules, enhancements, and chatbot variants. In those years, the first bots capable of learning from user dialogues emerged, such as Jabberwacky, paving the way for modern machine-learning approaches to natural language. Technology distribution became global, and human-machine conversation began to resemble a true social relationship, powered by collaboration among developers and enthusiasts worldwide.

A.L.I.C.E. and the revolution of programmable conversation.

The story of A.L.I.C.E. represents a real turning point in the evolution of chatbots, marking the transition from simple automated response systems to a new era of programmable and modular conversation. Developed by Richard Wallace in the mid-1990s, A.L.I.C.E. was the first conversational platform to establish itself as an open standard, thanks to the use of AIML, an XML-based language that allowed hierarchical and modifiable structuring of question-and-answer patterns. This approach made it possible to expand the bot’s knowledge and capabilities across a multitude of topics, overcoming the limitations of its predecessors and opening the door to large-scale customization.

Richard S. Wallace is an American computer scientist best known for inventing A.L.I.C.E. (Artificial Linguistic Internet Computer Entity) and the language AIML. He earned his Ph.D. in Computer Science from Carnegie Mellon University in 1989.

In 1995, he began developing A.L.I.C.E., a chatbot designed to simulate realistic conversation using rule-based patterns, and released it as open-source software in 1998.

A.L.I.C.E. grew rapidly thanks to a worldwide network of hobbyists and linguists who contributed AIML files, translations, and technical support. The project operated outside traditional academic funding, relying on volunteers, modest donations, and later, the nonprofit A.L.I.C.E. A.I. Foundation. Wallace’s work led to A.L.I.C.E. winning the Loebner Prize three times for most human-like chatbot. His innovations inspired research, educational tools, and even Hollywood films like Her, cementing his influence on the evolution of conversational AI.

A.L.I.C.E.’s success also lay in its open-source nature: the international community of developers could download, modify, and constantly improve the bot, adding new rules and responses in many languages. This collaborative philosophy turned A.L.I.C.E. into a global phenomenon, able to adapt to a wide range of cultural and application contexts—from customer support to entertainment, education, and academic research. The project won the Loebner Prize three times, establishing itself as the most realistic and sophisticated solution of its generation.

Despite its innovation, A.L.I.C.E. remained tied to a “rule-based” logic: the bot recognized predefined patterns and selected the most suitable response among those programmed, with no real semantic understanding or ability to learn from conversations in real time. However, the leap compared to ELIZA and PARRY was evident: A.L.I.C.E. could handle hundreds of topics, retained traces of context using specific tags, and offered a variety of responses and flexibility that brought the user experience much closer to a real digital assistant.

Although A.L.I.C.E. was the closest platform, in both philosophy and structure, to the idea of a generalist chatbot that we now see in ChatGPT, it still remains far from today’s artificial intelligence models in terms of machine learning, creativity, and reasoning capabilities. While A.L.I.C.E. represents the pinnacle of the rule-based generation, ChatGPT and modern LLMs mark the beginning of a new era, where conversation is no longer simply the sum of programmed rules, but the result of statistical models trained on billions of examples.

Nevertheless... A.L.I.C.E.’s legacy remains central: without the modularity of AIML, open-source collaboration, and the drive toward increasingly flexible chatbots, many of today’s advances in artificial conversation would have taken much longer to arrive. The story of A.L.I.C.E. is thus that of a bridge between past and future—between the dream of a machine able to talk and the reality of systems that, for the first time, truly understand us.

______

The legacy of pioneers and the roots of modern artificial intelligence.

How the insights of the 1960s and 1990s continue to shape the present.

The experiments of the 1960s and ’90s are not mere anecdotes in computing history; they form the foundations upon which modern conversational AI systems are built. The concept that a machine can respond convincingly, interpret emotions, or even learn through dialogue originates in these pioneering efforts. Today, many fallback techniques, pattern-matching strategies, and prompt-engineering methods used in large language models trace back directly to the insights of Weizenbaum, Colby, Wallace, and their colleagues. The history of early chatbots is thus not only a technical journey but an exploration of human intelligence mirrored in machines. Thanks to these visionaries, often working with limited tools and boundless ambition, we can now interact with systems that, if only for a moment, make us forget that behind the keyboard there is code rather than a human mind.

_____

Key Milestones in Early AI Chatbot History (1966–1990s)

Chatbot	Year	Why it was important	How it worked / user experience	What Made It Special / Curiosity
ELIZA	1966	First chatbot to show people could “bond” with a program	Typed text on a terminal; matched keywords and rephrased answers as questions. Felt like talking to a basic psychologist.	Users believed it “understood”; some asked for privacy to chat.
PARRY	1972	Showed AI could mimic personality and mental states	Conversations via teletype or early network; used logic for paranoia. Felt more “humanly unpredictable.”	Psychiatrists often couldn’t tell it from real patients.
RACTER	1984	First commercial chatbot, sparked public interest in machine creativity	Ran on home computers; output text that mixed templates and randomness. Dialogue was bizarre and creative.	Credited as author of a published book.
Dr. Sbaitso	1991–92	First “talking” bot for mainstream PC users	DOS program, type questions and hear robotic voice answer through speakers. Interaction was comical and artificial.	Famous for quirky voice and funny replies.
A.L.I.C.E.	1995/98	Set new standard for open, extensible chatbot frameworks	Online or app; used AIML (rules/patterns); felt more flexible, answered many topics.	Inspired movie scripts; won realism prizes.
Jabberwacky	1997	Early example of bots learning from users in real time	Chatted in a browser; learned by storing and reusing real user phrases. Felt less scripted and more surprising.	Became Cleverbot, which fooled many in Turing-like tests.

______

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org