There's a difference between what ChatGPT knows about your brand and what ChatGPT tells users today. Until now, AEO Copilot measured the first. From this release, paid plans measure the second: ChatGPT and Claude answer your tracked prompts from the live web, with real cited sources, the same way they answer actual users.
What changed
- ChatGPT and Claude runs on paid plans now search the web before answering, instead of relying on training data.
- Every response is labeled. Open any prompt and the ChatGPT and Claude tabs show a "Live web answer" or "Model knowledge answer" badge, so you always know how a result was measured.
- You control it. Settings has a new Live web search toggle under LLM Configuration, and each brand can override it: always on, always off, or account default. Turn it off if you want to keep measuring pure model knowledge.
- Free plans keep the current model-knowledge tracking. Perplexity and Google AI Overview already used search and are unchanged.
Why it matters
When someone asks ChatGPT for the best tools in your category today, it usually searches the web and builds the answer from the comparison posts and review sites it finds. An answer from training data can look very different. If you're optimizing for AI visibility, the live answer is the one that wins or loses you customers, so it's the one worth tracking.
Expect your numbers to move on the first live run. That's the real baseline: it shows which sources the engines cite in your category right now, and where your brand needs to show up next.