# ChatGPT Citation-Readiness Worksheet

Status: local planning artifact only. It does not fetch your page, change `robots.txt`, touch a CDN/WAF, open an analytics account, submit URLs, or test a ChatGPT account.

## 1. Name one page and one real question

| Field | Fill in before any technical review |
|---|---|
| Public page URL proposed |  |
| Reader's exact question |  |
| Direct answer visible near the top |  |
| Useful output (checklist, calculator, template, proof, or decision) |  |
| Content owner |  |
| Technical owner |  |
| Reviewed date |  |

Stop if the page includes private customer data, paid/access-controlled material, sensitive records, unpublished claims, or a topic requiring expert approval that has not happened.

## 2. State the crawler-policy decision separately

Do not assume that search discovery and model training are the same choice.

| Decision | Owner decision | Evidence to retain | Approval required |
|---|---|---|---|
| Allow OpenAI search discovery (`OAI-SearchBot`) |  | Current first-party bot documentation and intended `robots.txt` rule | Site/security owner |
| Allow or disallow training preference (`GPTBot`) |  | Separate owner policy rationale | Site/security owner |
| Page should be indexable/public |  | Content and privacy review | Content + technical owner |
| CDN/WAF may permit the intended public crawler |  | Owner-approved infrastructure check | Technical/security owner |

No generic "allow every AI bot" rule belongs in this worksheet. Do not weaken authentication, security, rate limiting, or private-page controls to chase a citation.

## 3. Crawl and page-parity card

- [ ] An accountable technical owner confirms the intended page is public and returns the expected content to ordinary visitors.
- [ ] `robots.txt`, page-level robots directives, CDN/WAF, and host behavior have been reviewed by the owner for the chosen policy.
- [ ] One canonical URL is proposed and matches the public page.
- [ ] Internal links make the page findable from a relevant, public parent page.
- [ ] The direct answer, material evidence, author/editorial context, review date, and primary-source links are visible in the rendered page.
- [ ] Any JSON-LD describes only visible, truthful content and validates as JSON.
- [ ] No fake author, rating, review, FAQ, price, availability, or citation claim was introduced.

## 4. Evidence map

| Visible claim | Primary source URL | Source checked date | Reviewer | Keep / revise / remove |
|---|---|---|---|---|
|  |  |  |  |  |
|  |  |  |  |  |
|  |  |  |  |  |

If a claim cannot be tied to a credible source, direct experience, transparent test, or clearly labeled opinion, revise or remove it. Do not hide uncertainty behind schema or AI wording.

## 5. Safe AI role card

Safe preparation:

- Turn a human-supplied page outline into a source/claim checklist.
- Flag missing visible answer, source, date, canonical, or internal-link fields.
- Draft a human-reviewable metadata and JSON-LD proposal from visible page content.
- Create a blank observation log.

Human-only / approval required:

- Choosing crawler/training policy or changing `robots.txt`.
- Checking or changing CDN, WAF, firewall, DNS, headers, hosting, canonical, sitemap, index controls, or schema on production.
- Opening analytics, ChatGPT, Search Console, Bing, webmaster, or deployment accounts.
- Publishing, deploying, submitting, measuring private logs, or making a citation/ranking claim.

## 6. Observation log after approved publication

Record states separately. Blank is an honest result.

| Date | Query / context | Page public? | Chosen crawler policy verified by owner? | Source/citation observed? | `utm_source=chatgpt.com` referral observed? | Evidence saved by owner | Next review date |
|---|---|---|---|---|---|---|---|
|  |  |  |  |  |  |  |  |

Do not infer crawl, indexation, ranking, traffic, leads, or revenue from one state. Do not automate query testing or account use without explicit approval.