# Reviewed teaching example and change record

Review date: 6 October 2026.

Reviewer: **Codex-assisted; owner review pending**.

This is a review of the retained response to `content-prompt.txt`, using frozen source pack `B05-KEYWORD-METRICS-v1`. It is a bounded teaching exercise, not client copy, a model comparison, an accuracy benchmark or a ranking study. The proposed author has not provided final sign-off.

## Result and limitations

The 22-row inventory includes the title, description, each substantive body statement, proposed alt text and both JSON values. Actions: **20 keep, one qualify, one investigate**. No factual corrections or removals were identified within the checked scope. These counts describe review rows, not a model accuracy percentage.

The qualification clarifies aggregation using an additional official source, D, consulted after the response. The original average-position wording matched the supplied pack; it was not contradicted. The unresolved row concerns whether proposed alt text matches an image that has not been made. The title, description and JSON values were retained. The two JSON strings were what the prompt requested; they do not constitute complete or validated structured data.

The ChatGPT web composer displayed `Pro`. A model menu inspected before the paired exercises displayed `6`, with `Latest` selected. Those interface observations are not an independently verified backend model identifier. Backend snapshot, temperature, sampling seed and hidden account context were unavailable or not audited. The prompt requested no tools, browsing, other conversations or personal memory; that instruction does not establish an isolated execution environment. See `method.md` for the full record.

The source pack and unedited response remain unchanged. The reviewed payload below preserves the returned sections and wording except for the changes listed after it. Proposed alt text remains proposed and must not be attached to the separate claim-review workflow infographic as though it described that graphic.

## Complete reviewed payload

<!-- REVIEWED-EXAMPLE:BEGIN -->
# How to Read Keyword Research Metrics

**Meta description:** Learn how historical keyword metrics, ad forecasts and observed site performance differ, including key caveats for reading Search Console data.

## Explanation

Keep three categories separate: historical keyword metrics, advertising forecasts and observed site performance.

Keyword Planner’s **average monthly searches** averages searches for a keyword and close variants over the selected period, location and Search Network settings. The default period is 12 months, and values are rounded. [A]

**Competition** describes advertiser competition for a keyword under the selected targeting, not organic ranking difficulty. [A] The **low and high top-of-page bid ranges** approximate the 20th and 80th percentiles, respectively, of historical top-of-page advertising bids for the selected location and network. [A]

**Forecasts** concern possible advertising performance, not historical search volume. Forecasted impressions account for bid, budget, seasonality and historical ad quality. [A]

Search Console’s **Search results Performance report** describes a site’s Google Search performance, including clicks and impressions. **CTR** is clicks divided by impressions. [B] **Average position** uses the topmost result for the relevant site or grouping, averaged across impressions. [B, D] A fresh manual search can show a different position because results depend on time, location, device and search history. [B]

Treat the visible query table as incomplete: some queries are omitted for privacy, and additional rows may be omitted because of internal limits. [C]

Finally, **exported zeros need context**. Search Console converts report values shown as `~` or `-`, meaning unavailable or not a number, into zeros during export. [B]

## Proposed alt text

Teaching diagram comparing historical Keyword Planner metrics (search volume, advertiser competition and bid ranges) with advertising forecasts and observed Search Console performance: clicks, impressions, CTR and average position.

## Suggested BlogPosting values

```json
{
  "headline": "How to Read Keyword Research Metrics",
  "description": "Learn how historical keyword metrics, ad forecasts and observed site performance differ, including key caveats for reading Search Console data."
}
```
<!-- REVIEWED-EXAMPLE:END -->

## Source key for the reviewed payload

- **A, supplied to the model:** Google Ads Help, About Keyword Planner forecasts. https://support.google.com/google-ads/answer/3022575?hl=en . Checked 2026-10-06.
- **B, supplied to the model:** Google Search Console Help, Performance report (Search results): Overview and basic setup. https://support.google.com/webmasters/answer/7576553?hl=en . Checked 2026-10-06.
- **C, supplied to the model:** Google Search Console Help, Performance report (Search results): Dimensions and data groupings. https://support.google.com/webmasters/answer/17011259?hl=en&ref_topic=17010766 . Checked 2026-10-06.
- **D, added during review only:** Google Search Console Help, What are impressions, position, and clicks? https://support.google.com/webmasters/answer/7042828?hl=en . Section: What is position? and Average position. Checked 2026-10-06. The documentation explicitly describes averaging position values across impressions. This source was not in the frozen model input.

No exact publication or update date was displayed for these source bodies. The check dates identify this review and must not be presented as publication dates.

## Exact change log

1. **C13, aggregation clarification and adjacent citation.** Replaced the exact source text `**Average position** uses the topmost result for the relevant site or grouping.` with `**Average position** uses the topmost result for the relevant site or grouping, averaged across impressions. [B, D]`. This makes the average explicit and identifies the additional review source. It does not correct a contradicted claim or alter the original input.
2. **C20, punctuation only; image check remains unresolved.** Replaced `metrics—search volume, advertiser competition and bid ranges—with advertising forecasts` with `metrics (search volume, advertiser competition and bid ranges) with advertising forecasts` in the proposed alt text. The categories are unchanged. The change follows house punctuation style; it does not verify an unmade image.
3. **Everything else in the payload was retained.** The title, meta description, body wording outside C13, section labels, and headline/description JSON values are unchanged. The source key and review notes surrounding the marked payload are editorial annotations, not part of the raw model response.

## Decisions still required

- Owner/human review of the teaching example, sources and article before publication.
- If the proposed keyword-metrics diagram is produced, inspect it and approve or revise its alt text. Otherwise keep that material as a proposed exercise artifact rather than deploy it as image text.
- If the JSON values are used in a page, integrate them into appropriate complete markup and validate that actual output. This exercise did not request or test a complete BlogPosting object.

## Retention check

- Frozen source-pack SHA-256: `2d47b2c2bf8de39c7aeb3f212d017f9b1d9d2e31189c282a53cc9f0e6a1a5d5d`.
- Raw-response SHA-256: `a6dd5a8f799f32b33cf325a88a78181db8864baaca4d175d7f68c66820e79e0c`.
- The reviewed payload can be reproduced by applying only change-log items 1 and 2 to the raw response.