Continued Influence Effect.

01Overview

The continued influence effect (Johnson & Seifert, 1994) describes what happens when misinformation persists in memory and reasoning even after it has been explicitly corrected. People may acknowledge the correction — and still use the original false claim when making judgments, explaining events, or deciding what to trust.

For designers, this is not an abstract media-literacy problem. It is a product-trust problem. A misleading tooltip, an outdated FAQ, a rumour in a community forum, or a botched launch announcement can keep steering user behaviour after you fix the copy. The correction lands. The original story keeps working.

02Detailed explanation

Research shows that retractions often fail to fully replace the initial narrative. The brain retains the causal story built around the false claim and uses it as scaffolding for later reasoning. In digital products, the pattern repeats across surfaces:

• Users continue to believe a feature was discontinued after a confusing banner — even after a clear follow-up email explaining it was a bug.
• A false pricing rumour on social media shapes purchase intent after the official account posts a correction that most users never see.
• Internal teams keep designing around a debunked user assumption — "everyone thinks we sell their data" — because the correction never became as memorable as the fear.
• Support macros that once stated incorrect policy language leave a residue: agents and users cite the old answer because it was encountered first and encoded with context.

Corrections update declarative knowledge unevenly. Automatic inference — what feels true in the moment — often still runs on the first version. For design, the first explanation users encounter is disproportionately expensive to replace.

03Why it exists

Coherent stories are cognitively cheap to store and expensive to rebuild. Once a causal explanation slots into place — why a fee exists, why an account was locked, why a feature vanished — dismantling it requires not just new facts but a new narrative structure. Under time pressure, the old structure wins.

In ancestral environments, revising a belief completely was less important than maintaining a workable model fast. Partial updating — accepting a correction in principle while still relying on the original inference — was often good enough. Product environments punish that compromise because trust, compliance, and safety depend on the details being right.

04Effects on users

Users who read a retraction may still avoid a feature, hesitate at checkout, or warn friends based on the first account they heard. They are not necessarily being irrational — the initial story may be more vivid, emotionally charged, or socially reinforced than the quiet update buried in a changelog.

They also misattribute which source they trust: a dramatic post feels more authentic than a sterile status page. The continued influence effect compounds with availability and illusory truth — repetition and emotional weight keep the wrong version alive.

05Effects on designers & teams

Teams encounter predictable failure modes when they treat corrections as one-and-done:

• Buried retractions. Fixing microcopy in settings while the misleading version still lives in onboarding, cached screenshots, and third-party reviews.
• Asymmetric salience. The false claim was alarming; the correction is calm. Alarm encodes better. Users remember the scare, not the footnote.
• Designing for the debunked fear. Research and roadmaps still chase a misconception users verbally disavow but behaviourally follow — extra privacy toggles nobody uses, or friction added for a threat that was never real.
• Assuming literacy equals behaviour change. Users can pass a comprehension check on the correct policy and still act on the outdated mental model weeks later.

06Practical takeaways

• Replace the story, not only the fact. Offer a complete alternative explanation for why users believed the misinformation and what is true now.
• Match correction salience to error salience. If the mistake was prominent — modal, banner, push notification — the fix should be equally visible, not a silent string change.
• Audit first-touch surfaces. Map everywhere the wrong information appeared: emails, help centre, in-app copy, sales decks, support snippets. Corrections must reach the same channels.
• Design for repeated true exposure. One correction fights repeated false exposure. Plan a brief campaign of accurate repetition through UX, not a single patch note.
• Measure behaviour, not agreement. Surveys may show users "know" the fix while funnels still reflect the old belief. Watch what they do after the correction ships.
• Pre-empt narrative gaps. Ambiguous UI invites users to invent explanations. Fill the gap with clear copy before misinformation does it for you.

07Design examples

08Ethical risks

When teams under-invest in corrections, vulnerable users pay disproportionately. Misinformation about billing, medical information, legal rights, or safety features does not fade fairly — it persists longest among users with less time, literacy, or platform fluency to chase updates.

Deliberately vague copy that later gets "clarified" exploits the continued influence effect: users act on the alarming interpretation while the organisation points to the fine print. That is not a communication oversight. It is a trust violation with a cognitive mechanism.

Self-test: Where in your product could a user still be acting on wrong information you already corrected — and how would you know?

09Related biases