How to get the most out of the prior art explorer

Drop a granted patent, a published application or a draft on the form (PDF, Word, text) and the title, claims and specification are read out of it, on this server. A scanned PDF with no text layer is not read: paste its text instead. The start page takes several documents at once, the drawings included.

Numbered, one claim per number, as you would file them. Dependent claims inherit the elements of their parents, so paste the whole set: the map scores the ladder, not just claim 1. No claims yet? Write a rough first set, even three or four claims: the search charts references against claim elements, so it needs them.

The description of the invention: a specification, or a plain write-up of the idea if that is all there is yet. It supplies the synonyms the search runs on, the vocabulary a searcher would use instead of yours, and the support quoted when an amendment is drafted. Paste the whole thing; long is fine.

The date the map is read against. References dated after it are kept and labelled but left out of the novelty scores; applications filed before it and published after it are treated as prior art under 35 U.S.C. 102(a)(2), or pre-AIA 102(e) if your application's effective filing date is before March 16, 2013.

Optional, semicolon separated. The inventors' own publications inside the one-year grace period are set aside from the prior art and marked, rather than counted against the claims.

One per line: a DOI, an arXiv id, a patent number or a title. Use it for the references you already have, an examiner's citations from an office action, or a competitor's patent. They are resolved and charted with everything else.

Tick the patent offices, the research literature, or both. Quick is a first look (three rounds, 80 references kept), Standard is the default (four rounds, 150), Exhaustive keeps searching longest (five rounds, 250). Every depth runs round after round until new queries stop returning new references.

Hexagons are your claims, sized by whether they are independent. Circles are papers and preprints, diamonds are patents. A red ring is anticipation risk: one reference reaches most of that claim. An orange edge is high overlap; a gold dashed edge is a 103 pair. A dotted outline marks a reference that is not prior art.

Network places references by how they relate to the claims and to each other. Radar puts the focus claim at the centre and rings the references by overlap. Timeline lays the references out by date against the priority line, which is the quickest way to see what is and is not prior art.

The Claim selector sets the focus claim. The chips add layers: Elements draws the claim's elements as nodes, Citations draws who cites whom, Similar draws references that read alike, 103 pairs draws the combinations, Not prior art brings the excluded references back into view. The Overlap slider hides references below a coverage.

Hovering a node says how it relates to the claim it is nearest, with the element it reaches best. Clicking opens the inspector on the right: the abstract, the element-by-element reading with its evidence, links to the record, and a button to put the reference on the IDS list. Find reference jumps to a node by title, number or author.

The Claim chart tab is the table an examiner keeps: each element of the focus claim against its closest references, marked disclosed, partial or absent, with the sentence that carries each reading. With element charting on, each reading is the model's and the sentence is quoted from the reference; without it, the overlap of terms is scored and the passage shown. It exports as CSV from the Export menu.

The 103 pairs tab lists the two-reference combinations that together cover the focus claim, best first, with which reference supplies which element and how similar the two are. A pair of references from the same field is a combination an examiner will make; a pair from distant fields needs a reason, and that is where your argument starts.

The Overview tab carries the novelty gauges (novelty is 100 minus the strongest coverage by one reference, or by two read together), what each independent claim faces, and the fallback positions: every dependent claim ranked by how much stands clear of the art. The Actions tab turns that into a plan per claim: the likely rejection, the elements that carry the claim, the strongest supported amendment with an Apply button, the references to cite with Add buttons, and the fallback claims.

Everything you change is on the History tab, with the score before and after, so a claim can be walked back one edit at a time. Undo and Restore original at the foot of the claims pane do the same in one press.

Press Edit on any claim, change the language, press Done. The elements are re-split, the chart is re-read against every reference already retrieved, and the novelty, the pairs and the map update in a moment. Edit all claims opens the whole set as text for a larger rewrite.

Draft amendment asks for three alternatives for that claim, each adding a limitation the retrieved art does not reach, with the supporting sentence quoted from your specification so the amendment does not read as new matter. Apply one and the map re-scores; keep the one you like and edit it further.

A revision can reach art the first search did not look for. Search revised claims runs a further search built from the new language and adds what it finds to the same map, at no charge. Do it once the claims have settled, not after every edit.

Read the fallback positions before amending claim 1. The dependent claim that stands clearest of the art is usually the amendment, and it already has support, numbering and antecedent basis.

Put a reference on the list from the inspector, or press Add prior art with 50%+ overlap on the IDS tab to start from everything close. The list comes out in the order the SB/08 forms want, patents and non-patent literature apart, with the references dated after your priority date flagged so you can decide about them.

Export, Interactive report writes one HTML file that opens anywhere, works offline, and re-scores the claims when they are edited inside it. Send it to counsel or to the inventor; nothing in it phones home.

The prosecution brief is a text file with the reading of every independent claim: the likely rejection and the references behind it, the elements the art does not reach, the recommended amendment with its support, the references to cite and the fallback claims, ready to paste into a memo. The claim chart and the IDS list export as CSV; Copy current claims puts the revised claim text on the clipboard; the session file (JSON) holds the whole map.

A map is saved to your account the moment it is made and has a link of its own; Copy link to this map opens it where you left off, for anyone signed in to the account. My maps lists them; a map can be deleted there at any time.

Paste the inventor's claims or the disclosure's key sentences as claims, run Quick, and read the map before a word of the specification is written. The elements nothing reaches are where the claims should sit; the pairs say what the examiner will combine.

Put the examiner's citations under Known references, paste the claims as filed, and open the chart on the rejected claim. The Actions tab names the element the cited references reach only partially, which is the amendment, and the pair they do not make, which is the argument. Apply the amendment, re-score, and download the brief before you write the response.

Run Standard or Exhaustive, add everything above 50 percent overlap to the list, walk the timeline for the references near the priority date, and export. Have a practitioner review the list: the readings are automated and the duty of disclosure under 37 CFR 1.56 stays with you.

The interactive report is the briefing. The overview page shows what stands clear, the chart shows why, and the recipient can try an amendment themselves. For a technology transfer office deciding what to file, that is the whole conversation in one file.