Define scope, collection conditions and comparable evidence first. Review publication, indexing and AI citations separately; track Baidu and Google page positions and real enquiries independently from AI mention rates.
Use explicit conditions and sources so each answer can be understood on its own.
Agree on the brand and aliases, questions, competitors, region, platform/model, device and entry point, keeping exact prompts. Separate questions that name the brand from discovery questions and preserve the scope during re-checks.
A valid sample has a complete, preserved answer, including answers without the brand. Collection coverage = valid answers / planned checks. Report failures and unknown results separately instead of removing them to improve a ratio.
Brand mention rate = valid answers mentioning the agreed brand / all valid answers. Website citation rate = valid answers citing agreed website URLs / all valid answers. State alias, domain and duplicate-citation rules for the same scope.
Record position only for an explicit ordered list; mark unordered answers not applicable. Compare dates, models and collection conditions, separating changes in method rather than presenting them as proven improvement.
Link each observation to the question, time, platform/model, region/device, full answer, cited URLs and screenshot. Summaries need scope, valid/failure counts, denominators and limits. Publish case materials only with permission.
Publication does not prove indexing; crawlability does not prove indexing; indexing does not prove AI citation; one citation does not guarantee a long-term ranking. Each stage needs its own evidence.
Use verified source materials and obtain public-use permission before publishing.
Preserve original records, re-check the agreed questions and identify model or entry-point changes. Keep unverifiable samples as unknown.
Support material claims with original sources. User behavior, search conditions and model updates can affect outcomes, so avoid a single-cause guarantee.
For web search, record engine, actual region or egress, device, query and landing page. For AI, record platform, entry point, web-search setting and exact prompt. Region parameters alone do not prove testing from the target city.
Accept only the work and evidence actually completed. Crawlable text, clear structure and reliable sources support retrieval, while real indexing and citations still require platform-specific observation.
Content updated：