Find prior art beyond the keywords you thought to search.

Google Patents and USPTO search are keyword-and-classification tools — you get back exactly what you searched for, and nothing you didn't think to search for. Overstand reads and understands the entire active patent corpus at the concept level, so it surfaces prior art that predates or anticipates a claim even when the reference uses completely different terminology than the one you searched. It's built for the "did I miss something" question litigation actually turns on, not just retrieval.

Every patent Overstand surfaces is a real, existing filing you can pull up on the USPTO or a foreign patent office's own records — Overstand never invents a patent number, and never invents a claim, filing date, or citation that isn't actually in that patent's file history. Where a general-purpose model like Claude or ChatGPT can occasionally invent a plausible-sounding citation, Overstand's results are constrained to verified, real filings only — every result is traceable back to its source.

Yes. The same full-corpus prior art search runs in either direction. To invalidate, Overstand finds patents and publications that predate and anticipate the target's claims — the basis for an invalidity argument. To defend, it runs the identical search against your own patent's claims before your opponent does, so you know exactly what they'll find and can prepare your enforcement position accordingly.

These are the three deterministic filters patent searchers have always used: keywords (the terms an examiner or attorney expects a patent to use), CPC codes (the Cooperative Patent Classification system's technology categories), and art units (the USPTO examining groups organized by technology area). Senior litigators build a mental map of these over years of practice. Overstand applies all three automatically and exhaustively — then goes further, searching by concept so it isn't limited to the exact terms, codes, or units a human searcher thought to check.

Yes. Overstand is available as an MCP (Model Context Protocol) server, so Claude, ChatGPT, or Gemini can search Overstand directly — the same way they can already search the web. You get Overstand's full-corpus, verified-citation search from inside the AI tool you already use, without learning a new interface.

All work product — search queries, target patents, findings, and reports — is protected with military-grade encryption throughout the process. Overstand never trains any model on client data, ever.

The entire active corpus — over 12 million active US patents and published applications, plus international filings across PCT, EPO, and other major offices. Every prior art search scans the full corpus; there's no sampling and no keyword-limited subset.

We're onboarding litigation teams in order as we open up access. You'll hear from us directly with next steps — including how to connect Overstand to your existing workflow, whether that's the Overstand platform itself or Claude, ChatGPT, or Gemini via MCP.