ClaimHit Releases v2 of its Patent Infringement Search Platform

ClaimHit, an AI-driven platform for patent infringement searches, has introduced version 2 of its claim-mapping engine. This patent-pending upgrade integrates invention parsing, a multi-model ensemble of large language models, parallel keyword and semantic web search, and an evidence-verification layer. It enables AI-based scans of any patent in under two minutes. New users receive three free credits upon signing up, with introductory pricing at $99 per credit. The team is releasing v2 publicly. They want practitioners to test it and provide feedback on what features to develop next. This approach applies both to infringement search and to a broader vision they see as foundational rather than final.

For patent attorneys, IP licensing teams, patent holders, and in-house counsel, the initial phase of infringement or licensing negotiations has traditionally been consistent: select a patent, pinpoint potential products that may infringe the claims, gather supporting evidence for each candidate, and compile it into a credible case for partners, clients, or courts. This process of discovery and verification, conducted before detailed claim analysis, has generally required days or weeks of manual research across vendor sites, datasheets, regulatory documents, and trade publications.

ClaimHit‘s v2 release condenses the front end into a single automated scan that completes in under two minutes. But, in the team’s own framing, the more interesting story is what they’ve chosen to build first, what they’ve left for later, and the kind of feedback they’re now seeking from the people whose workflows the tool is meant to support.

Where the team focused first – evidence over consensus:

The most common architecture for AI-assisted patent search to date has been either LLM with temperature control or some form of consensus voting: run multiple language models in parallel, count how many surface the same candidate, and rank by agreement. The intuition is that if four out of seven models name the same product, the result feels more credible than a name that only one model surfaced.

ClaimHit v2 explicitly does not do this.

“We spent a long time deciding what the ensemble’s job actually is,” said Bikram, Founder of ClaimHit. “The conclusion we landed on is that the ensemble is for discovery, widening the net so we don’t miss a real product because one model happened not to know about it. Ranking is a different question, and we think the right answer for ranking is evidence: have we independently verified, on the manufacturer’s own product information, that this candidate exists and looks like what the patent describes? When that’s the question, agreement among LLMs is no longer the right answer. So in v2, the ensemble surfaces candidates, web search both lexically and semantically extends the pool, and ranking is decided by the evidence we can verify per-candidate. We think that’s a better foundation for any work that’s going to hold up in a licensing memo, in litigation, in front of a partner, but we’re aware we may be wrong about which trade-offs matter most.”

What’s in v2:

The team describes v2 less as a feature release and more as the first version of the platform that produces results they are willing to put in front of users.

Invention parsing: ClaimHit extracts the inventive contribution and primary claim elements directly from the patent text, so users don’t pre-annotate or pre-process patents before scanning.

Single-patent and portfolio scans: v2 supports both single-patent deep scans and portfolio sweeps. A practitioner can analyze one patent in detail or run a patent family or full portfolio through the platform.

Multi-model LLM retrieval: Discovery runs in parallel across multiple independent frontier models. Different models surface different plausible products; v2 treats that breadth as the point.

Parallel keyword and semantic web search: Both lexical and meaning-based web retrieval running together, alongside the LLM ensemble, during candidate discovery.

Evidence verification before display: Before any candidate reaches the user, ClaimHit independently verifies it against publicly available manufacturer information. Candidates that cannot be evidence-backed are filtered out — addressing a class of failure practitioners have seen across LLM-based patent tools.

Evidence-based ranking with Good Match and Possible Match tiers: Candidates are ordered by the strength of the supporting evidence rather than by how many models surfaced the same name. The UI separates results into Good Match (well-evidenced) and Possible Match (worth reviewing, weaker evidence) tiers, giving users a clear cue on where to spend review time.

AI Hit Charts: From any surfaced candidate, users can generate a structured Hit Chart mapping the patent’s claim elements to product features, with citations to the source material.

Sub-two-minute scans: Most full scans complete in under 120 seconds end-to-end.

What v2 doesn’t have yet:

The team has been explicit, in conversations leading up to launch, that v2 is a foundation rather than a finished product. Real-time progressive results, a streaming UI that surfaces verified candidates as they complete rather than waiting for the full scan, are in active design but not yet shipped. Deeper portfolio analytics, broader verification of every plausible match the ensemble surfaces, and integrations with common IP workflow tools are on the roadmap, but deliberately not promised on a timeline. The reasoning, per the team: roadmaps written without practitioner input tend to optimise for the wrong things.

Where ClaimHit is heading – a foundation, not a thesis:

Patent infringement search is the first problem ClaimHit set out to solve, but the team’s longer-term ambition is broader, and worth describing because it shapes what they’re asking practitioners to weigh in on.

A meaningful fraction of patents, especially newly issued ones, deep-tech patents emerging from research institutions, and patents covering technologies still ahead of their commercial ecosystem, have no current infringers, because the technology is too early or because the market that would use it hasn’t formed yet. For those patents, asking “who is infringing” is the wrong question. The more useful questions, the team argues, are forward-looking ones: which companies could benefit from this technology if they adopted it, which startups working in adjacent areas might be candidates to license or acquire it, which emerging verticals could be unlocked by the underlying invention, who would be the right tech-transfer or commercialization partners, and which industry advisors or board-level decision-makers already have this technology in their line of sight.

Today, the strategic positioning work done by university tech-transfer offices, corporate IP licensing teams, R&D leadership, patent owners trying to monetise portfolios, and the IP advisory community more broadly takes weeks of research, a lot of judgment, and a non-trivial amount of luck to find the right counterparties. The team’s view is that this work needs the same kind of evidence discipline that ClaimHit applies to infringement search, applied to a different question.

The team is careful to distinguish what v2 is from what that broader vision will require.

“What we’ve built in v2 invention parsing, multi-model retrieval, evidence verification which is a necessary foundation,” said Bikram, “It is not a sufficient one. Answering ‘who could commercialize this technology’ or ‘who would be the right tech-transfer partner’ is a fundamentally different problem from finding existing infringers, and it needs layers we haven’t built yet. modelling what companies are actually capable of, reasoning about adjacency and commercial readiness, and evaluating strategic fit. The architecture we have is the first floor of a much taller building. We’ve focused on infringement first because it has the cleanest evaluation criteria, whether the candidate is a real product on a real manufacturer’s site, or it isn’t. Those upper floors will look different, and we don’t pretend to have them yet. We’re talking about them now because the shape of those layers depends on input from the people who do tech-transfer and commercialisation work today.”

The implication for a reader of this announcement, the team suggests, is that ClaimHit is not just trying to be a faster infringement-search tool, but it is also not yet the full commercialisation platform the team eventually wants to build. v2 is the part of the journey the team is willing to put a price on and a public release behind. The rest is the conversation they are now opening.

Who it’s for, and how to try it:

ClaimHit is designed today for patent prosecutors, IP litigators, patent holders, licensing professionals, and corporate IP departments. The kinds of work it’s intended to support today include identifying licensing targets, building infringement watchlists, conducting freedom-to-operate analyses, and pre-screening portfolios for monetization opportunities. The platform supports U.S. and international patents, and the team is interested in conversations with tech-transfer offices, deep-tech investors, and patent owners working on commercialization, primarily as voices in shaping where the platform goes next, since the layers needed to fully serve those use cases are not yet built.

ClaimHit v2 is available today at [claimhit.com]. New users receive three free credits on signup, enough to run patents through the platform without payment. During the v2 introductory period, additional credits are priced at $99 each.

ClaimHit‘s evidence-first claim-mapping methodology is patent pending under the title: “System and Method for Domain-Agnostic Mapping of Structured Content to Real-World Commercial Entities Using Heterogeneous Language-Model Ensembles, Independent Web-Evidence Cross-Validation, and Multi-Stage Class-Verification Filtering.”

An invitation:

The team is open to conversations with anyone who wants to weigh in on direction as a user, an advisor, a workflow partner, or a critic. If you’ve worked on infringement search, licensing target identification, freedom-to-operate, portfolio monetisation, tech transfer, or commercialisation strategy, and have views on what tools like this should and shouldn’t do, the team would like to hear from you. Feedback channels and contact information are at claimhit.com.

About ClaimHit:

ClaimHit is an AI-powered patent infringement intelligence tool building AI tools for patent professionals. Founded in 2026 by Bikram jit Singh.

ClaimHit is currently focused on revolutionizing technology commercialization workflow and is working with practitioners across global markets.

Media contact: [email protected]