---
name: Nagging
slug: nagging
category: pattern
status: published
created: 2026-08-21T00:00:00.000Z
modified: 2026-08-21T00:00:00.000Z
definition: Interrupting a task again and again with the same request, so that
  agreeing becomes the cheapest way to be left alone.
aliases:
  - name: persistent prompt
    source: community
  - name: prompt fatigue
    source: community
  - name: repeat ask
    source: community
  - name: maybe later loop
    source: community
tags:
  - consent
relations:
  contrastWith:
    - permission-priming
    - friend-spam
  variantOf:
    - dark-pattern
  partOf: []
  seeAlso:
    - rating-prompt
implementations: []
sources:
  - title: "Deceptive Design: Nagging"
    url: https://www.deceptive.design/types
demo: inline
exhibit: false
useWhen: the same prompt keeps returning until you give in
---

Nagging is the one deceptive pattern that needs no deception. Nothing in the prompt is a
lie, the decline button works, and the reader is free to press it as often as they like.
The trick is arithmetic: refusing costs a small amount of attention every time, agreeing
costs it once, and a design that returns often enough makes consent the cheaper option.
What is being measured as a conversion is really a surrender, which is why nagging shows up
in [Deceptive Design](https://www.deceptive.design/types)'s catalogue of named types rather
than in a book about persuasion.

The tell is what happens to the answer. A "not now" that is recorded is an answer; a "not
now" that resets a counter is a delay, and the counter is usually the whole mechanism: ask
on the third launch, then the fifth, then every session. Notice that the copy in the
specimen above never gets more aggressive. The third ask is word for word the first. All
that has changed is the reader, who by then has learned that the only reliable way to make
the prompt stop is to say yes to it. Platforms have started legislating against exactly
this, which is the clearest possible sign that it stopped being a design decision: browsers
now treat a dismissed permission request as a denial that sticks, and mobile app stores
limit how often a rating prompt may be shown.

The repair is not softer wording. It is remembering, which means treating a refusal as
durable state, and then moving the request somewhere it can be found when it is wanted. A
row in Settings asks nothing and is available forever. If the ask genuinely has to come back
(a trial about to expire, a security setting that matters), it comes back once, at a moment
the reader can connect to something they were doing, and it says why it has returned.

The nearest neighbour is [permission priming](/permission-priming), and the difference is
the whole ethics of the thing: priming asks at a moment the reader understands, having just
done something the permission is obviously for, while nagging asks again because the first
answer was not the one wanted. Priming spends the request carefully because it only gets
one; nagging assumes it gets unlimited ones. Below that, nagging is a
[dark pattern](/dark-pattern) by the standard test, since it works only because the
interface's interests and the reader's have come apart. And its usual vehicle is the
[interstitial](/interstitial), the screen that stands between a person and the thing they
came for, which is what makes the interruption expensive enough to be worth ending.