Prior-art search beyond keywords: six strategies

Keyword searching finds documents that describe an invention in the words the patent uses. The references that end up mattering often use different words, sit in another patent office's file, or are not patents at all. Six strategies reach them: citation chains, family members in other jurisdictions, the inventors' earlier work, classification neighborhoods, period-appropriate terminology and non-patent sources. After a first pass, the work becomes iterative: search again for the claim elements no reference covers yet.

Start from the claim, not the abstract

Every strategy below feeds one table: the claim's limitations down the side, references across the top, and a note in each cell saying whether that reference discloses that limitation and where. The table tells you when you have found enough and where to look next. If the claim has not been split into limitations yet, do that first; how to break a patent claim into limitations walks through it.

Record the date you will measure against: the effective filing date of the claim you are testing, usually its earliest supported priority date. Every candidate gets checked against that date before anyone spends time charting it.

1. Follow citation chains in both directions

Backward citations are the references listed on the patent itself. They are the starting set, not the answer: the examiner considered them and allowed the claims anyway. Their value is as a door to the second generation, the references they cite, which are older and often closer to the core idea.

Forward citations of the target are later documents that cite it. They postdate the target, so they are not prior art to it. Their reference lists are still worth reading, because later applicants and examiners often cite older documents that the target's examiner never saw.

Steps:

1. List the target's backward citations and pull each one's own citations.
2. Pull the forward citations of the target and of your two or three best references, then read their reference lists.
3. Discard anything dated after the target's effective filing date.
4. Prioritize documents that appear in several chains; repeated citation is a useful signal of relevance.

2. Check family members in other jurisdictions

The same invention filed in Europe, Japan, China, Korea or through the PCT is examined by different examiners searching different collections. Their search reports often cite art the US examiner never considered. European search reports mark each citation by category: X for a document relevant on its own, Y for one relevant in combination, A for background. An X or Y citation against a claim similar to the one you are testing is a strong lead.

Steps:

1. Pull the full family, including PCT and national filings.
2. Read the search reports, written opinions and office actions for each member.
3. Compare the foreign claims to the US claim, since scope often differs after amendment.
4. Add every cited reference not already in your table, then check its date.

3. Trace the inventors' earlier work

Inventors publish papers, give talks, write theses and file earlier patents in the same area, often years before the application at issue. Under 35 U.S.C. § 102(b)(1), an inventor's own disclosure made within the year before the effective filing date is excepted from the prior art. Anything earlier than that year is prior art like any other document. This applies to patents under the America Invents Act rules (effective filing date on or after March 16, 2013); older patents follow the pre-AIA timing rules.

Steps:

1. Search each named inventor in patent databases, filtered to dates before the target's filing.
2. Search academic indexes, university thesis repositories and conference programs for the same names.
3. Check co-authors and former employers; earlier group work often describes the same system under a different project name.

4. Search the classification neighborhood

Classification codes cut across vocabulary. Two documents that share no keywords can sit in the same Cooperative Patent Classification (CPC) group because they solve the same technical problem.

Steps:

1. Note the CPC codes on the target and, separately, on your best references so far. They often differ, and the difference shows where else the idea was filed.
2. Read the definition of each group and look at neighboring subgroups and the parent group.
3. Search within those groups using a few broad terms rather than the patent's exact phrasing.
4. For older art, also check the classification schemes in use at the time, such as the earlier US classification.

5. Use the vocabulary of the period

Terminology changes. Before a field settles on its words, the same idea is described in older language, often borrowed from a neighboring discipline. A search in today's vocabulary misses those documents.

Claim 1 of US 6,285,999 B1 ("Method for node ranking in a linked database", priority 10 Jan 1997) scores "linked documents" based on the scores of the "linking documents" that point to them. A period-aware search would translate each element:

Citation analysis of academic journals and network analysis in the social sciences both worked with directed references between items long before web search existed. Those fields are where a searcher would look for the scoring idea described in non-web terms.

Steps: read review articles and textbooks from a few years before the filing, harvest the terms used in the oldest references you have found, and build one query set per limitation.

6. Go beyond patent databases

Much of the earliest description of a technology never appears in a patent: conference papers, journal articles, theses, technical reports, standards contributions, product manuals and datasheets, archived web pages and source code releases. These count as printed publications if they were publicly accessible before the critical date, meaning interested people skilled in the field could have found them with reasonable diligence (MPEP 2128).

Steps: search academic indexes and conference proceedings for the years before filing, look for manuals and datasheets of products sold in that period, check web archives for product pages, and record the evidence of each document's public date as soon as you find it.

Then iterate on the uncovered elements

After the first pass, the table shows which limitations are disclosed and by what. The empty rows are the next search.

1. Take each limitation with no disclosing reference and search for it alone, without the rest of the claim.
2. Run it through the strategies above: its own classification group, its own period vocabulary, the citation chains of whichever reference comes closest.
3. Update the table. A single reference that fills every row supports anticipation under § 102. Several references that together fill the rows feed an obviousness argument under 35 U.S.C. § 103, which then needs a reason a skilled person would have combined them.
4. Stop when new rounds return only references you already have, or when the remaining gap is the point the patent owner will argue is inventive. That gap is worth stating plainly in the report.

Keep a column noting which strategy found each reference. It makes the search explainable to counsel and shows which routes were tried for the elements that stayed uncovered.

Which search you are running also changes the cutoff date and what counts; FTO, patentability and validity searches compared sets out the differences.

How One Transform runs this

One Transform's Prior Art search works this way. Discovery runs several strategies at once: citation chains, patent families in other jurisdictions, the inventors' earlier work, classification neighborhoods and period-appropriate terminology. Non-patent literature comes in through academic search, conference archives and historical product documentation. Every reference is scored against the claim elements it discloses, and the search keeps iterating on elements that remain uncovered. The output is a ranked reference list in which each finding links to its source.

Next step

Take a patent you are evaluating and check which of its limitations your current reference set leaves uncovered. To run the six strategies in parallel, start a Prior Art search from the One Transform product overview.

This article is general information about patent practice, not legal advice. For advice on a specific matter, consult a registered patent attorney or agent.