---
name: content-freshness-qa
description: Use when asked to review a content-heavy site (blog, articles, docs, prompt or resource library, product guides) for stale, wrong, or broken content - outdated facts, prices, plan names, screenshots, dates, product names, dead embeds, missing images, duplicate or thin pages, inconsistent formatting, and metadata errors. Produces a prioritized editorial worklist for the owner; only fixes mechanical issues (broken formatting, missing alt, dead embeds) in a review PR. Does not rewrite content.
---

# Content freshness and editorial QA

Last reviewed: 2026-11-10

## Quick prompts

For anyone improving a site or app, whether a small-business owner or someone at a larger company practicing better prompts. Outdated facts, prices, or tool names quietly erode trust, especially on how-to and comparison content. Paste a prompt into any AI assistant that can see the site or repo.

1. "Review the 10 most-read pages on the site I work on. Flag anything that looks outdated, like prices, product names, or 'this year' references, and cite a current source for each. Report only."
   *Why this prompt works:* Ask for sources so each claim can be checked.

2. "Find pages on the site I work on with broken images, dead embedded videos, or leftover formatting code. List them by page before fixing anything."
   *Why this prompt works:* Naming the exact problems gives a mechanical, easy-to-verify list.

3. "Look for pages on the site I work on that cover nearly the same topic. List the pairs and suggest whether to merge, update, or keep each. Don't change content."
   *Why this prompt works:* Asking for a suggested action per item turns findings into decisions.

## Hard limits

- Do not rewrite, delete, unpublish, or merge content. The owner decides; you supply evidence and a suggested action.
- Mechanical fixes only in the PR (broken markdown, missing image dimensions, malformed front matter, dead embed swapped for a link only if the owner agrees). Mark each.
- Verify a "this is outdated" claim against a current primary source opened during the task, and cite it. Do not rely on memory for fast-moving facts (AI tool plans and prices, software versions, laws).
- No publishing to production. Review PR or a report only.

## Inputs to gather

1. Content source (Markdown/MDX in Git, CMS export, headless CMS) and front matter schema.
2. Content types and their freshness risk (evergreen essay vs. tool tutorial vs. pricing comparison).
3. Owner's style rules: date format, title case, author names, category list, image rules.
4. Analytics top pages, if available, to prioritize.

## Interface-type notes

- **Articles / blogs**: dated tool advice, "this year" references, author and date fields, broken in-body embeds, category sprawl.
- **Docs / KB**: steps that no longer match the product UI, screenshots, version numbers, code samples that no longer run.
- **Prompt or resource libraries**: tool names and plan limits, model names, links to tools, example outputs that no longer reflect the tool, duplicate entries with different slugs.
- **E-commerce content**: discontinued products referenced in guides, prices in copy.

## Steps

1. **Inventory**: list every item with title, URL, type, published and updated dates, author, categories/tags, word count, image count, and outbound links.
2. **Mechanical checks** (automatable): missing or duplicate titles/descriptions, empty or default excerpts, broken images, images without alt or dimensions, malformed headings (skips, multiple h1), unrendered markup (shortcodes, raw HTML entities, leftover page-builder variables), broken embeds, dead links (use link-and-redirect-integrity-check).
3. **Freshness checks**: flag items that mention versions, prices, plan names, years, "new"/"latest", or named products; check the claims for the highest-traffic and highest-risk items against current official sources.
4. **Duplication and thin content**: near-duplicate titles or bodies, very short pages, overlapping tags and categories.
5. **Consistency**: date format, author names, category usage, title case, CTA blocks.
6. **Prioritize**: risk (wrong or harmful) × traffic × effort.

## Validation

Rebuild after mechanical fixes; page count unchanged; spot-check fixed pages in a real browser; re-run mechanical checks to zero.

## Report format

1. Inventory summary (counts by type, age buckets, last-updated distribution).
2. Worklist table: URL, issue, evidence (with source link and date checked), suggested action (update, merge, redirect, retire, keep), priority.
3. Mechanical fixes in the PR. PR link, "not merged".
4. Suggested review cadence per content type (for example tool tutorials quarterly, evergreen yearly).

## References

- Google Search Central, creating helpful, reliable, people-first content: https://developers.google.com/search/docs/fundamentals/creating-helpful-content
- Google: byline dates: https://developers.google.com/search/docs/appearance/publication-dates
- Google developer documentation style guide (useful house style baseline): https://developers.google.com/style
- Nielsen Norman Group, content audits: https://www.nngroup.com/articles/content-audits/
- W3C WAI, images tutorial (alt text decisions): https://www.w3.org/WAI/tutorials/images/
