Audit site-scoped ChatGPT queries
Find ChatGPT queries scoped to your domain and map them to existing or missing answers on your site.
When to use this workflow
“What is ChatGPT trying to find on our site?”
Before fetching
Confirm the canonical domain, prompts, observed ChatGPT model, and dates. Do not mix another platform’s queries or ordinary brand searches into site: queries.
Complete the CLI quickstart and confirm the workspace with orc workspace current --json. Replace example IDs, URLs, dates, and settings. Apply only approved mutations.
1. Fetch the observed model and queries
Check the observed answers, then use their model_id to fetch queries. Do not invent an ID from a model’s display name.
orc answers list --prompt-id YOUR_PROMPT_ID --platform chatgpt-default --json
orc fanout-queries list --type search --prompt-id YOUR_PROMPT_ID --model-id YOUR_MODEL_ID --start-date 2026-09-01 --end-date 2026-09-07 --json2. Select queries scoped to the owned domain
Have the agent inspect the site: operator and compare normalized hosts. Exclude different hosts such as example.com.evil.test. Retain query IDs and times, and count recurrence within the fetched set.
3. Map each query to an answering page
Use supplied files or the agent’s browsing capability to inspect the site inventory and text. For each query, record an answering page, incomplete coverage, or no matching page. Absence from the source catalog does not prove a page is absent from the site.
Deliverable
Deliver queries, observed counts, query IDs, mapped URLs, and content gaps. Mark the audit unavailable when the provider does not expose query evidence.
On fetch failure, retain the error and request scope. Do not count empty data, incomplete pagination, or missing permissions as a completed investigation.
Next steps
If a failure remains, report it and verify the fix, including this workflow and the failed step.