Model
A model is the trained AI system that writes an answer. A platform like ChatGPT is the app around it, and may use several models.
A model is the trained AI system that actually writes an answer. GPT, Gemini, Claude and Llama are families of models. ChatGPT, Gemini the app, Copilot and Perplexity are platforms: the apps people use, which run one or more models behind the scenes.
The difference matters because the platform often decides which model answers, and neither you nor your customer gets a say. Two people asking the same platform the same question can get answers from different models.
Platform and model
A platform does more than the model. It decides whether to search the web, which searches to run, which pages to read, which model to send the question to, and how to show the answer. The model writes the words.
So when you ask "what does ChatGPT say about us?," the honest answer is "it depends which model, whether it searched, and who's asking." A signed-in paying user may get a bigger model than someone on a free plan. A quick question may go to a fast model, a hard one to a slower one.
Why models change your results
Each model learned from different text, at a different time, and was tuned differently. When a platform switches to a new model, the brands it names can shift overnight, even if nothing on the web changed.
For example, Fernhill Coffee is named in about half the answers to its prompts on one platform for months. Then the platform moves to a new model, and Fernhill's visibility there drops by a third in a week. Its competitors' numbers move too, though nothing on Fernhill's site or anyone else's changed.
In AnswerOrbit
We ask each platform in its own app, as someone who isn't signed in, and let the platform pick the model, the same as it would for anyone. Each answer shows the model that gave it. When a platform changes how it answers, our counting rules move to a new version, we mark any comparison across the change, and that platform's first seven days start again. See How we count.