Base vs grounded

What AI knows versus what it reads

Every model answers two ways: from memory, and with live web search. Proversial runs both, so you can tell a knowledge gap from a search gap and fix the right one.

Two answers, one clear diagnosis

If a model knows you from memory but drops you once it searches, the live web is where your problem sits. If it never knew you at all, the fix is a different one. We show both answers side by side.

Fix the right gap

Seeing base and grounded together points you straight at the work that will move your rank, so your effort lands where it counts.

See the live web, with sources

The grounded run comes with the pages the models pulled from, so you can see what shaped the answer today.

Why the two answers tell you different things

A model can leave you out of an answer for two different reasons, and the fix depends on which one it is. It might not know you well from its training, or it might know you and still favour other names once it searches the live web. On the surface both look the same: you are missing.

Running the model from memory and with live search separates the two. If you show up from memory and drop once search is on, the live web and its sources are where the work is. If you are absent both ways, the model itself needs to learn you, through the kind of presence that shapes future training.

Most tools report a single answer and leave you guessing which problem you have. Proversial runs both every day and shows them together, with the sources behind the grounded answer, so you spend your effort on the gap that is holding you back.

More from Proversial

The rest of what we do, and how it fits together.

See both sides of the answer

Base and grounded, every day.