Why chatbots on services like Messenger and Kik have struggled to live up to the hype, and how developers are trying to reverse their downward trend
Chatbots in apps like Facebook's Messenger and Kik have struggled to live up to the hype, but that might actually help them succeed. Tweets: @mgsiegler , @joshconstine , and @caseynewton Tweets: M.G. Siegler / @mgsiegler : No surprise here. (From a year ago: http://500ish.com/...) http://twitter.com/... Josh Constine / @joshconstine : The definitive article on chatbots. Why they're built, why they fail, and how utility + discoverability can improve http://marketingland.com/... Casey Newton / @caseynewton : @Techmeme my take: because the chatbots are bad
Context & Ripple Effects
A year into the experiment, the verdict is in on the bot gold rush Facebook started when it opened Messenger to chatbots in April 2016: early builds were slow, frustrating, and largely useless, and the same pattern held on Kik. This piece is the post-mortem — Constine, Newton, and Siegler all publicly agree the bots are bad — but it also argues the failure is diagnostic rather than fatal.
The developers' proposed fixes target the two gaps the coverage keeps naming: utility (bots that do something worth typing at) and discoverability (users who can find them). The stakes extend past 2017 — Meta would later scrap its celebrity AI chatbots within a year of launch, suggesting the same utility-and-discoverability problem outlived the rule-based bot era.
First-order effects
- Developers who built shopping and travel bots on the chat SDK Facebook seeded in early 2016 now face collapsing engagement inside Messenger and Kik, forcing them to rework products around narrow, high-utility tasks instead of open-ended conversation.
- Facebook and Kik must answer for platform-level failures — poor discovery surfaces and weak tooling — or watch the developer base they recruited at launch drift away.
Second-order effects
- Brands that budgeted for conversational marketing redirect spend toward channels with measurable returns, pressuring Messenger and Kik to bundle better analytics, templates, and ranking into their bot platforms to keep them.
- Tooling vendors step into the gap: Facebook's own Bot Engine, built on its Wit.ai acquisition, becomes the template for machine-learning middleware aimed at making third-party bots less brittle.
Third-order effects
- The episode establishes a durable lesson that recurs across generations of the technology: conversational interfaces fail on UX and distribution before they fail on intelligence, so each model upgrade alone does not guarantee traction — Meta's 2024 retreat from celebrity chatbots shows the pattern repeating even with far stronger AI underneath.
- If utility-plus-discoverability becomes the accepted bar, messaging platforms converge on curating a small set of transactional bots (ordering, booking, support) rather than an open app-store free-for-all, reshaping how conversational commerce is structured.
The trend: Messaging platforms repeatedly over-promise conversational software, absorb a hype-correction cycle, and rebuild it around utility and discovery as the underlying AI improves — a cycle running from 2016's rule-based bots to today's LLM-driven assistants.