5 Signs You're Talking to a Stateless Chatbot (Not a Real Memory System)

If you're evaluating AI companion apps and want to know whether "it remembers you" is real or marketing, these are the practical tells.

1. It gets vague the further back you go

Ask about something from your very first conversation, not last week's. A summary-based system tends to lose specificity the older the reference — names blur, details get generic. Real persistent memory shouldn't degrade with time the same way.

2. It contradicts itself across sessions

If a character states a preference or fact in one session and casually contradicts it in a later one, that's often a sign nothing is actually being checked against stored state — the model is generating plausible-sounding continuity rather than retrieving real memory.

3. Starting a brand-new chat resets everything

Some apps tie memory to a single ongoing thread rather than to your relationship with the character. If opening a fresh conversation wipes the slate clean, memory is scoped to the chat, not to you.

4. It can recite facts but the relationship never deepens

There's a difference between an app that can echo back a fact you stated and one where familiarity, tone, or storyline actually shift over time. If nothing about the character's behavior changes no matter how long you've talked, memory (if present) is shallow.

5. Nothing happens when you're away

If a character has no sense of elapsed time and treats a conversation after a month-long gap identically to one continued five minutes later, there's no real model of continuity running underneath — just a chatbot responding to whatever's in front of it.