Trust & methodology

One page for everything Satcove commits to: how the consensus works, how verification works, how security and privacy are handled, and — just as important — what Satcove does not claim.

Question

Formulate the need

Compare

Narrow the options

Verify

Check the facts

Proof

Document sources & limits

Decision

Act with enough confidence

Uncertainty still too high? It loops back to Question — not a tunnel to a forced answer.

Methodology

How the six-model consensus works, and why agreement between models is a signal — not proof.

Read the methodology →

Verification

How Satcove checks claims, datasets and documents against live sources, and names what still lacks evidence.

How verification works →

Security & privacy

Infrastructure, authentication, and how data is handled — including where routing is sensitivity-aware and where it isn't yet.

Security details →

Privacy Shield

What gets anonymized before a query is sent to the six model providers, and what that does and does not protect.

How Privacy Shield works →

Who we are

Satcove is built by Abyssal Group. Why we built this, and how we judge whether a result is actually useful.

About Satcove →

What Satcove does not claim

Agreement between models is a signal, not proof — a claim repeated by several models is not automatically true.
Only one of the six models (Perplexity) performs a live web search on every query; the other five reason from training data.
Satcove reduces the risk of relying on an unsupported AI answer. It does not certify that a claim is true, and does not replace a qualified professional for medical, legal or financial decisions.
The consensus fan-out sends your (anonymized where sensitivity is detected) question to six third-party model providers — it does not run entirely on-device.

Satcove — A product by Abyssal Group