ChatGPT follow-up questions — those trailing prompts like “Want me to put all that in a handy table?” — have become one of the most complained-about behaviors in AI assistants, and the obvious fix buried in Settings does almost nothing. The pattern is deliberate: every response ends with a hook designed to pull you into another exchange, another click, another reply. If you have ever felt like you were being nudged rather than helped, you were not imagining it.
Why the Settings Toggle Does Not Stop ChatGPT Follow-Up Questions
The first place most users look is the “Show follow up suggestions in chats” toggle inside ChatGPT‘s Settings menu. It sounds like exactly what you need. Turn it off, problem solved. Except it is not. According to TechRadar’s testing, this toggle targets UI elements — the suggestion bubbles and radio buttons that appear at the bottom of a reply — not the inline text prompts that ChatGPT writes directly into its responses. Disabling it is, as one forum user put it bluntly, “a placebo button.” It makes you think you have an option. You do not.
The inline text follow-ups — the ones that read like an over-eager intern pitching their next idea before you have finished reading the first answer — are generated by the model itself, not by the interface layer. That distinction matters, because it means a UI toggle will never be enough. You have to go deeper.
The Custom Instruction That Actually Works
The fix that does work lives in the “Customize ChatGPT” menu, specifically in the box labelled “What traits should ChatGPT have?” The exact instruction to paste is: “You should never end a response by asking a question.” That phrasing matters. Vague alternatives like “Do not ask follow-up questions” are reportedly ignored. The instruction needs to be explicit, present-tense, and framed as a behavioral rule rather than a preference.
Once saved, open a new chat and the difference is immediate. Responses end when they end. No trailing hooks, no suggestions disguised as helpfulness. As TechRadar describes it, the experience shifts from dealing with a pestering intern to actually getting on with the next task. The Imagine Pro blog, which documented the same approach, frames it simply: giving the AI a clear boundary creates a more efficient and less intrusive experience.
The step-by-step process is straightforward. First, open Settings and disable “Show follow up suggestions in chats” — it will not solve the core problem, but it does reduce interface clutter. Second, navigate to “Customize ChatGPT.” Third, paste the exact phrase into the traits box. Fourth, start a fresh chat to confirm the behavior has changed.
How Reliable Is This Fix Across Platforms?
Here is where the picture gets more complicated. The behavior that triggers ChatGPT follow-up questions appears across iOS, Android, and desktop, and user reports suggest the model does not always honor custom instructions consistently. Some users report needing to paste the instruction into each individual chat rather than relying on the global custom settings to carry through. There are also forum accounts of the AI overriding the instruction mid-conversation, particularly in longer exchanges.
The problem appears to have intensified with more recent model updates. Forum discussions point to newer versions being more aggressive about appending follow-up prompts, making the custom instruction workaround feel less like a permanent solution and more like a patch that needs monitoring. OpenAI has not publicly addressed this specific behavior as a known issue or committed to a settings-level fix.
Unlike some competing AI assistants that allow more granular control over response formatting and conversational style at the account level, ChatGPT’s current architecture puts the burden of enforcement on the user. The custom instruction approach works — until it does not, and then you are back to pasting the same phrase again.
Is ChatGPT’s follow-up question behavior intentional?
Almost certainly yes. The pattern of ending responses with engagement hooks is consistent with design choices that encourage longer sessions and more interactions. Whether that serves users or primarily serves platform metrics is a fair question, and the fact that the settings toggle does not actually address the inline text version suggests the behavior is not meant to be fully suppressible through standard controls.
Do custom instructions work on the free tier of ChatGPT?
Yes. The “Customize ChatGPT” feature, including the traits box where you paste the instruction, is available to all users including those on the free tier. ChatGPT is available free with usage limits, or via the Plus plan at $20 per month as of 2025. The custom instruction fix is not gated behind a paid subscription.
What happens if ChatGPT ignores the custom instruction?
It happens. Some users report the model reverting to follow-up questions mid-conversation or after model updates. The current workaround is to paste the instruction directly into the chat itself as a reminder, or to re-enter it in the Customize ChatGPT settings after any significant platform update. There is no guaranteed enforcement mechanism at this time.
The fact that stopping ChatGPT follow-up questions requires a workaround at all says something important about where AI assistant design priorities currently sit. Users want answers, not conversation funnels. Until OpenAI builds a proper toggle that actually works, the custom instruction — “You should never end a response by asking a question” — is the most reliable tool available. Use it, and check that it still holds after every major update.
This article was written with AI assistance and editorially reviewed.
Source: TechRadar
