I’ve been a ChatGPT Plus subscriber for a while, and initially, I was satisfied — voice mode included. It responded naturally and directly, just like I expected.
But over time, I began noticing a change: voice mode became increasingly verbose. Instead of simply answering, the assistant started adding long preambles, repeated confirmations, and excessive “friendly” commentary — even though I had clearly set my Custom Instructions to avoid all that.
Here’s what I use in my Custom Instructions:
“Be concise and direct. Avoid filler phrases. In voice mode, no intros or summaries — just answer.”
Despite this, voice mode in Plus gradually began to ignore those preferences. It started feeling less like a smart assistant and more like a chatty host. So I decided to compare.
I used the same advanced voice mode on a free account, with the same device, app, and settings — and the difference was immediately obvious.
The free version responded much more cleanly and directly, with far less unnecessary commentary.
To clarify: this is not a comparison between feature tiers. Both tests used the exact same advanced voice mode.
The only difference was the account: one Free, one Plus.
Out of frustration, I paused my Plus subscription and kept using the free version — and voice mode behaved exactly how I wanted.
Once I re-subscribed to Plus, the verbose behavior returned instantly.
This strongly suggests that Plus users may be routed to a different voice behavior model, potentially designed to be more “personable” by default, even when tone preferences say otherwise.
My Questions:
Are Plus users routed to a different voice behavior layer or model?
Can we opt out of this more “personality-driven” version?
Can tone and behavior settings (like Custom Instructions) be enforced consistently, regardless of plan?
All I want is a voice experience that respects my settings — clear, fast, no unnecessary warmth or filler. Just the answer.
Thanks for reading.