Truace
Subscribe
68/100

Moderate evidence

Cite
More
Export CSVExport JSONEmailLinkedInX

Chatbots that remember customers raise satisfaction, but privacy worries offset some of the gain

Study title: When AI Remembers: Investigating the Dual Role of Long‐Term Memory in Chatbots

Evidence type
Controlled or validated comparison
Sources
1 sourcePeer-reviewed
Published
2 min read
Truvace record
TRV-2026-2560cite or check it

Study title: When AI Remembers: Investigating the Dual Role of Long‐Term Memory in Chatbots

A peer-reviewed paper in Psychology & Marketing reports three controlled, scenario-based online experiments in e-commerce customer service. Researchers varied whether an AI chatbot had long-term memory and found that memory significantly increased satisfaction and continuance intention.

Read the study abstract

ABSTRACT This research investigates the dual role of long‐term memory (LTM) in AI chatbots and its impact on customer outcomes. Drawing on Mind Perception Theory and Privacy Calculus Theory, we conceptualize LTM as a continuity‐based design capability that may be interpreted as a functional service feature, a privacy‐relevant cue, or a basis for anthropomorphic inference. Across three controlled scenario‐based online experiments in e‐commerce customer service, we manipulated the presence of LTM and, in Study 3,…

55Gain side

How the two sides scoreClose call

58Problem side

The gain

In three online experiments in e-commerce customer service, chatbots with long-term memory significantly increased customer satisfaction and continuance intention, mainly by making the bot seem more competent.

The problem

In the same experiments, the benefits of chatbot memory were partially offset by perceived anthropomorphism and privacy concerns, and baseline perceived privacy empowerment did not reduce those concerns.

A fixed formula sets both scores from the evidence and its sources. Reader votes don’t move them.

What the scores mean

G is the gain side, out of 100. It goes up when the gain is specific, when it measurably helped people and when the sources behind it are strong.

P is the problem side, out of 100. It goes up when the problem is specific, when it measurably hurt people and when the sources behind it are strong.

In brief

The effect on word-of-mouth intention ran mainly through other factors rather than directly. Memory made the chatbot seem more competent, but the gain was partly cancelled by customers perceiving the bot as human-like and by privacy concerns.

In the third experiment, giving users an explicit data-control mechanism within the memory-enabled condition raised perceived privacy empowerment, which lowered privacy concerns and improved outcomes. The authors say this offers guidance for designing memory-enabled chatbots with user controls.

Main points

  1. The research used three controlled, scenario-based online experiments set in e-commerce customer service.
  2. Researchers manipulated whether the chatbot had long-term memory; Study 3 also manipulated user data control.
  3. Memory significantly increased satisfaction and continuance intention.
  4. The effect on word-of-mouth intention was primarily indirect.
  5. Memory improved outcomes by making the chatbot seem more competent.
  6. Perceived anthropomorphism and privacy concerns partially offset those benefits.
  7. In Study 2, baseline perceived privacy empowerment did not reduce privacy concerns tied to memory.
  8. In Study 3, an explicit data-control mechanism increased perceived privacy empowerment, reduced privacy concerns and improved outcomes.

The full story

The study draws on Mind Perception Theory and Privacy Calculus Theory and treats long-term memory as a continuity-based design feature. Customers may read it as a useful service function, a privacy-relevant signal or a sign of human-like qualities.

Across the experiments, memory increased satisfaction and continuance intention. Its link to word-of-mouth intention was mostly indirect, running through competence, anthropomorphism and privacy concerns.

Study 2 found that customers' baseline sense of privacy empowerment was not enough to ease memory-related privacy concerns. Study 3 added an explicit user data-control option to the memory-enabled chatbot.

That control increased perceived privacy empowerment, which reduced privacy concerns and improved outcomes through a serial mediation path. The authors conclude that explicit data control can help manage privacy concerns in memory-enabled chatbots.

What this doesn’t fix

The findings come from controlled, scenario-based online experiments in an e-commerce customer service setting, not from observed behavior with real deployed chatbots.

Sources

1 on file
  1. Psychology & Marketingdoi:10.1002/mar.70267
    Peer-reviewed2026-09-11

Talk it over

Four questions to start onThis record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

Start with a question

Chatbots that remember customers raise satisfaction, but privacy worries offset some of the gainhttps://truvace.com/item/when-ai-remembers-investigating-the-dual-role-of-long-term-memory-in-chatbots

  1. This record found a gain and a problem in the same place. Which matters more to you, and who might answer differently?

    The gain In three online experiments in e-commerce customer service, chatbots with long-term memory significantly increased customer satisfaction and continuance intention, mainly by making the bot seem more competent.

    The problem In the same experiments, the benefits of chatbot memory were partially offset by perceived anthropomorphism and privacy concerns, and baseline perceived privacy empowerment did not reduce those concerns.

    Answer in the thread

  2. How do we know? What can this kind of evidence show, and what can’t it?

    The evidence This rests on a controlled comparison, from Psychology & Marketing.

    Answer in the thread

  3. What does it leave open? What else would you want to know before acting on it?

    The record’s own caveat The findings come from controlled, scenario-based online experiments in an e-commerce customer service setting, not from observed behavior with real deployed chatbots.

    Answer in the thread

  4. Have you seen AI used this way where you study, work or live? Did it go the same way?

    Answer in the thread

The questions are the same on every record. The details under them are taken from this record as written.

What readers say

More from Business