Evidence comes first
DeepSearch begins with deterministic public sources and web discovery. It normalizes the query, checks relevant public services, discovers likely profiles, and records the claims and URLs that support each candidate.
For social discovery, the system compares profile links, reused handles, bios, photos, verified status, and other corroborating signals. A shared name by itself is not enough for a high-confidence match.
Signals become an identity graph
Profiles that link to one another or share several compatible signals can be grouped into a candidate cluster. Contradictions, such as a clearly different full name or incompatible location, reduce confidence or keep profiles separate.
This is why the candidate picker matters: DeepSearch can rank likely people, but you provide the final grounding by choosing the person whose visible details match what you know.
AI organizes the grounded result
After evidence has been gathered, AI helps reconcile and explain it. Unsupported fields are discarded, conflicts remain visible, and source links stay attached to the profile.
The AI is useful for synthesis, not as a new source. When the underlying evidence is thin, the answer should be treated as thin too.