RexBot: a conversational doorway to sensitive questions

Read the campaign story

Durex introduced RexBot in 2017 as a Facebook Messenger conversation experience about sex. Social Samosa’s reporter tried the bot and described a menu-led interaction with character-based sections and topic choices. The brand sought to make a socially uncomfortable subject easier to approach through a conversational format rather than a conventional product advertisement.

Economic Times Brand Equity credited Ansible and reported the company’s claims about the experience. Its account described an animated character and content organised for male and female audiences. Those reports establish the interface concept and launch period. They do not verify clinical accuracy, confidentiality, inclusive coverage or the quality of every answer. A private-looking chat window should not be confused with a proven privacy guarantee.

The editorial mechanism is to lower the social cost of asking an initial question. A conversational entry point may feel less formal than searching a long information page, and a menu can help people who do not know how to phrase a query. Yet the same friendliness can make a system seem more knowledgeable or trustworthy than it is. Clear boundaries are especially important when the subject involves health, relationships or sensitive personal disclosures.

For a small team, the safe transferable experiment is a low-risk FAQ guide, such as explaining returns or event access, rather than an unreviewed sexual-health adviser. Use a short menu, show the source of each answer, allow users to leave easily and provide a human support route. Collect no personal information that is unnecessary for the task. Test whether people know they are interacting with a scripted or automated system and can recognise when its scope ends.

Any real health-information adaptation would require qualified review, appropriate age handling, inclusive language, data protection and routes to professional support. Engagement alone would be an inadequate success measure. This case studies the design of an approachable conversation interface, not a recommendation to rely on the historical bot for advice. Its most useful lesson is the tension between reducing embarrassment and increasing perceived authority: making a question easier to ask must not make unsupported answers easier to believe.

Use clear topic choices

Help people navigate without forcing unnecessary personal disclosure.

Show limits and sources

Make automation, answer provenance and escalation routes visible.

Start with low-risk topics

Prototype on routine FAQs before considering regulated or sensitive information.

A small test for another brief

Prototype a no-login returns-policy FAQ menu.

Correct answers found and recognition of the system’s limits.

Stop if users disclose sensitive information or mistake the guide for professional advice.

• Write verified answers to five low-risk questions.
• Label automation, sources and a human escalation route.
• Test comprehension without collecting personal data.