Research note 007
Why model versions matter in AI visibility research
“ChatGPT” is not one fixed system. Models, retrieval modes and interfaces change continuously. An observation means little without knowing what produced it.
What should be recorded
- Provider — OpenAI, Google, Anthropic, Perplexity or other.
- Model — The model identity returned by the provider, where available.
- Date and time — When the observation was collected.
- Retrieval status — Whether web search or retrieval was used.
- Prompt and context — Exact prompt text and any location or language context.
Why “a ChatGPT result” is not enough
Two observations labelled “ChatGPT” may come from different models with different retrieval behaviour. Comparing them as if they were the same system can attribute a model change to a change in the firm's visibility. FirmRanker reports results per system and treats model changes as a variable to track over time.