Last verified: October 5, 2026
What You Can Now Do
Context Memo now lets you click "Simulate prompt" on any Problem Graph node and open the Simulator with that prompt already filled in, skipping manual copy-paste setup. The Problem Graph shows you where your brand is weak across the prompts buyers actually run. Until now, acting on that required moving the prompt yourself. Now the gap and the test sit in the same motion: you see a node where a competitor owns the answer, you click, and you're looking at live model output for that exact prompt. The decision it changes is how many nodes you bother to verify. When testing a prompt costs one click instead of four steps, you check the whole cluster instead of the one node that looked worst.
Where It Is in Context Memo
The action lives in the Problem Graph view. Select any node and the "Simulate prompt" action appears with the node's other actions. Clicking it opens the Simulator with the prompt text already loaded in the input field, ready to run against your selected models. Nothing about the Simulator itself changed, same model selection, same output view. The only difference is how you get there and what's waiting for you when you arrive.
How to Use It
- Open the Problem Graph for the brand or category you're reviewing. The graph renders the prompt clusters where your visibility is contested.
- Select the node you want to test. Pick one where a competitor is cited and you aren't, or where the model's description of your category looks off.
- Click "Simulate prompt" on that node. The Simulator opens in place of the graph view with the node's prompt already in the input field.
- Confirm your model selection and run the simulation. You'll get the live answer each selected model returns for that prompt, including which sources it cites.
- Read the citations, not just the answer. If a competitor's page is carrying the response, that URL tells you what format and depth the model is rewarding for that prompt.
- Return to the Problem Graph and repeat on adjacent nodes in the same cluster. Patterns across three or four related prompts are more useful for memo planning than any single result.
Why We Built It
Customers told us they were copying a prompt out of the Problem Graph and pasting it into the Simulator to test it. That's a trivial action once. It's not trivial thirty times, and that's roughly what a serious review of a contested cluster looks like. The practical cost wasn't the seconds lost. It was the nodes people skipped. When verification requires a context switch, you verify the one or two items that look most urgent and take the rest on faith, which means the prompts where you're quietly losing never get checked. Removing the switch makes full-cluster review the default rather than the exception.
A second cost was transcription error. Prompts in the Problem Graph are phrased the way buyers phrase them, which often means long, specific, and awkward. A truncated or slightly reworded prompt returns a different answer, and you end up building a memo against a question nobody asked. Prefilling the field removes that failure mode entirely.
What It Does Not Do Yet
The action carries the prompt. It does not carry your model selection, persona, or any other Simulator configuration from the node context. If you run simulations against a specific model subset, confirm that selection each time you land in the Simulator rather than assuming it persisted.
It's also single-node. There's no batch action to queue every prompt in a cluster and return results as a set, so reviewing a cluster means returning to the graph between runs. If you need a consolidated view across many prompts at once, the scanning and share of voice reports remain the better instrument for that, with the Simulator serving the narrower job of checking one specific prompt against live model output.
Finally, simulation tells you what models return right now. It doesn't tell you what to publish. The citations in the output are your input to that decision, and the memo generation workflow is where the response gets built. Treat the Simulator as the diagnostic step, not the fix.