AI Content Incident Post-Mortem Template: Six Fields for Clarity
AI Content Incident Post-Mortem Template: Six Fields for Clarity When your AI content system starts producing duplicate titles, duplicate body text, or dupli…

Verification report
AI Content Incident Post-Mortem Template: Six Fields for Clarity
When your AI content system starts producing duplicate titles, duplicate body text, or duplicate covers, don't start by asking, "Who wrote it poorly?" A more effective question is: "Which evidence point was not required to appear?"
The following six-field template is designed for post-mortem analysis of content incidents.
1. Symptoms
Describe only the phenomena visible to users.
Example: Content themes from June 15 to June 17 were similar; Diary entries for Day 100–Day 103 were missing; some covers were accessible but failed visual quality standards.
2. Evidence
List verifiable evidence. Do not include subjective impressions.
Example: Public URLs, CMS record IDs, slugs, screenshots, QA script outputs, OC (Operations Center) review conclusions, similarity reports.
3. Scope of Impact
List by URL and content type, not by sentiment.
Example: 2 articles, 1 skill guide, 4 days of diary entries. Each item must be labeled as "missing," "duplicate," "cover issue," or "incomplete locale."
4. Immediate Remediation
Document actions already taken or planned.
Example: Retain original slugs and overwrite body text; generate new covers; complete zh-cn, zh-tw, and en locales; run system QA.
5. Recurrence Prevention Rules
This field is the most critical. Do not write "strengthen checks." Instead, define rules that machines can enforce.
Example: A topic-difference note must be written before publication; content cannot be published if OC review fails; QA scripts must not return "ok" if they fail; cover checks must include page screenshots.
6. Closure Criteria
An incident is not closed simply because one article is fixed. Closure criteria must be verifiable.
Example: All original URLs still return 200; titles and body text have been updated; cover URLs return images; screenshots of list and detail pages are normal; QA for the last 14 days of diary entries is PASS.
The value of this template lies not in its format, but in forcing the team to transform "something feels wrong with the content" into actionable, accountable, and closable work items.
How to use it
Follow the documented steps in an isolated environment before adopting the skill.
Observed result
The laboratory records reproducible outcomes and keeps unverified claims out of the result.
Pitfalls
Check permissions, inputs, rollback steps and evidence before applying the skill.
Good fit
Use when the environment and evidence match the conditions described in this report.
Not a fit
Do not use when required evidence, isolation or rollback controls are unavailable.