Our Offerings · Prior Art Search

The reference that kills your claim rarely uses your words.

Semantic search across 220 million patents from 108 jurisdictions and more than 100 million research papers, matched by meaning, run in graded rounds, and triaged by you. The results position the claim set the moment drafting begins.

Search report DOCX Feature matrix XLSX X · Y · A triage 220M patents · 108 jurisdictions

Why it matters

Why the search you run shapes the patent you get

The most dangerous reference in any prosecution is the one the drafter never saw: the document that describes the same invention in entirely different vocabulary, filed in another jurisdiction, indexed under a classification nobody thought to check. Keyword search misses it by design: it matches words, and the prior art was written by someone who chose different ones.

There is a second, quieter problem. Even when a search is thorough, it usually ends in a report: a list of references that the drafter reads, forms a view on, and then sets aside to start writing. The search and the draft are two separate jobs, connected only by whatever survived in the drafter’s head. Claim positions that should have been shaped around the closest art are drafted first and repaired later, in an office action, at office-action prices.

Esgenix treats search as the first half of drafting, not a preliminary. Agents plan the search strategy from your invention’s features, which you can workshop first (split them, merge them, brainstorm angles with the AI), and run it in rounds against the full corpus, semantically, so a reference is found for what it discloses rather than what it happens to be called. Each round has a named direction: go broader, take a different angle, move to an adjacent domain, or a custom instruction from you.

The search does not retire into a PDF. It becomes the spine of the application.

What comes back is graded, not dumped. References arrive as X (key match), Y (relevant in combination), or A (background), the grading vocabulary examiners actually use, and you triage them on a board built for speed, keyboard shortcuts included. A feature matrix maps your invention’s features against the top references cell by cell, with evidence quotes, so "how close is this really?" has a documented answer. Suggested combination groups show you the D1-plus-D2 arguments an examiner might assemble before an examiner assembles them.

Then the part that separates this from every standalone search tool: the results do not retire into a PDF. If you continue into drafting, the graded references are already in the session. The claim set is positioned against the closest art from the first draft, the specification carries fallback positions exactly where the art is dense, and the search that would once have been filed away becomes the spine of the application.

You direct every round and approve everything that leaves. What changes is the floor: the search that used to depend on the right keyword now depends on what the documents actually say.

The workflow

How a search runs

Run it for you, or walk through it step by step. Either way, you triage every round and approve the report.

  1. 01

    Describe the invention

    Paste the disclosure or describe the invention in chat. The agents extract the searchable features.

  2. 02

    Search strategy & scoping

    Review jurisdictions, classifications, and which features to search. Workshop the features by splitting, merging, and brainstorming angles before a single query runs.

  3. 03

    Rounds with named angles

    Agents run the search in rounds: go broader, different angle, adjacent domain, or your custom direction. Each round lands as a graded batch.

  4. 04

    Graded triage

    References arrive as X (key match), Y (combination), or A (background). Triage them on the board, keyboard shortcuts included, and direct the next round.

  5. 05

    Feature matrix & combinations

    Top references mapped against your features cell by cell with evidence quotes, plus examiner-style combination groups.

  6. 06

    Search report

    Edit the title, executive summary, and conclusion; approve; and export the report with the full reference record.

Capabilities

Built for how searches actually go wrong

Semantic matching

Search that matches meaning rather than keywords, across 220 million patents from 108 jurisdictions and 100+ million research papers, including the reference that used none of your words.

Feature workshop

Edit, split, and merge the features driving the search, or brainstorm angles with the AI. All of it happens before the search runs, when it is cheap to change course.

Named round angles

Every round has a direction you chose: go broader, take a different angle, move to an adjacent domain, or follow your custom instruction.

X / Y / A grading

The vocabulary examiners use: X for a key match, Y for relevant in combination, A for background. Every grade explained, every override yours.

Feature matrix with evidence

References mapped against features cell by cell as disclosed, partial, or absent, with the disclosing quotes attached, exportable as a workbook.

Examiner-style combinations

Suggested D1-plus-D2 reference groups covering your features: the combination arguments you will face, assembled before you file.

Non-patent literature

Research papers searched alongside patents in the same rounds, in the same grading system, not a separate database bolted on.

Coverage analytics

X/Y/A split, classification by round, feature coverage, and cumulative coverage across rounds, so you can see when the search has actually converged.

Handoff to drafting

Continue into drafting and the graded art is already in the session: the claim set is positioned against the closest references from the first draft.

Deliverables

What lands in your inbox

Documents you can hand over. Every one is reviewed and approved by a person before it leaves the platform.

DOCX · PDF

Search report

Executive summary, methodology, graded references, and conclusion, edited and approved by you before it exports.

XLSX

Feature matrix workbook

The full feature-by-reference matrix with grades and evidence, ready for the client or outside counsel.

PPTX

Presentation deck

The findings as slides, for the meeting where the filing decision gets made.

DOCX

Invention disclosure

Optionally, a structured disclosure document built from the same session, ready to seed drafting.

Questions

What practitioners ask

Keyword search matches the words a drafter happened to choose; semantic search matches what a document means. The reference most likely to block your claim describes your invention in different vocabulary. That is precisely the reference keyword search misses and semantic search finds.

X marks a reference that reads on the invention by itself; Y marks one that is relevant in combination with another; A marks background art. It is the grading vocabulary patent examiners use, so the triage board reads like a search report, because it becomes one.

Yes. That is the point. If you continue into drafting in the same session, the graded references position the claim set from the first draft, and the specification carries fallback positions where the art is dense. Search and drafting are one workflow, not two engagements.

You do. The agents grade and explain every reference, but you triage the board, override any grade, and direct what the next round should chase. The report does not finalize without your approval.

Run a search on a real invention

Bring a disclosure to the demo and watch the rounds come back graded, then see what the same session looks like when it continues into drafting.

© 2026 Syncinvent Labs Private Limited. All rights reserved.