Doing research well is a craft. A careful researcher does a set of things an AI does not do on its own: framing the real question, checking what can be trusted, reading sources against each other, testing a conclusion before relying on it.
AIGroundwork writes that craft down as a framework an AI can follow, so the research you get is sound, not just quick.
Directly apply a prompt for a single research step, or add the system prompt to reuse across reports.
Paste a single prompt into a chat when you want one research move in the moment: chase a shaky figure, stress-test a conclusion, check a draft. It works in the conversation you are already in.
Go to manual promptsInstall the framework into a project or an agent once. It then applies the full process to every research task, including the parts you would not have thought to ask for.
Get the system promptIt helps to see research as five steps, each resting on the one before. The how it works page explains them in full.
Where a researcher works out the real question before searching, an AI answers whatever it is handed.
Where a researcher checks what can be trusted one source at a time, an AI reads the top results and stops.
Where a researcher reads two things against each other for what it reveals, an AI gives a summary.
Where a researcher tests a finding and looks for what would disprove it, an AI states the most common view.
Where a researcher writes the finding up so others will read it, an AI leaves it as raw output.
Standards check the quality of each step, because every step builds on the ones before it. Each one applies where it is relevant, not at every step.