The cheapest work in the whole invention process is also the work most consistently skipped. A prior art search costs nothing but time if you run it yourself, and somewhere between a few hundred and a few thousand if you commission one. Drafting and filing a full application costs several times that, and an examiner will run the search regardless — eighteen to thirty months later, at a point when the money has already gone.
Searching first inverts the order. It is not a formality performed to confirm what you already believe; it is a genuine test, and a meaningful proportion of ideas fail it. Failing it in week four is a good outcome. Failing it in month twenty-six, after drafting, drawings and filing fees, is the same result with an invoice attached.
Scope · What counts
Prior Art Is Broader Than People Expect
Prior art is anything made publicly available before your filing date. Not published in your language, not published in your field, not necessarily published at all in the ordinary sense — merely available to a member of the public who was under no obligation of confidence. A granted patent counts. So does an application that published and was then abandoned. So does an expired patent whose protection lapsed forty years ago, and that category is the one first-time searchers most often overlook, on the intuitive but wrong assumption that expired means irrelevant. An expired patent is not enforceable, but its teaching is permanently in the public domain and defeats a later claim just as thoroughly as a live one would.
Non-patent material counts equally: journal papers, conference posters, product manuals, trade catalogues, marketplace listings, instructional videos, and forum posts. So do your own earlier disclosures. Understanding what a patent actually protects makes the breadth easier to accept: the bargain is disclosure in exchange for a time-limited monopoly, so anything already disclosed cannot be traded for one a second time.
An expired patent cannot be infringed and cannot be licensed. It can still destroy a claim, and it does so silently.
The category searchers skip
Method · Three passes
Keywords, Then Classification, Then Citations
An effective prior art search runs in three passes, each narrowing what the previous one produced. Keyword searching alone is where most amateur attempts stop, and it is the weakest of the three, because patent drafters have every incentive to avoid the obvious word. A device you would call a clip may be described throughout as a resilient retaining member.
The three passes
- Keywords. Build three vocabularies — the function, the structure, and the problem solved — and generate synonyms for each. Search combinations, not single terms. The goal is not completeness; it is finding two or three genuinely relevant documents.
- Classification. Open those documents and read their classification codes. The scheme sorts every patent document into a hierarchy of technical subject matter, and browsing your invention's subclass surfaces material no keyword would have reached, including documents in languages you do not read.
- Citations. Follow references backwards to what each document cites, and forwards to what later cites it. Forward citations are how you find out whether anyone built on the idea, and who.
Work through the drawings before the text. Patent prose is deliberately abstract, but the figures show the mechanism immediately, and a searcher can dismiss a document from its drawings in seconds. Keep a log of every search string, database and date — partly so the work is not repeated, partly because a drafter will want it, and partly because a documented search is the evidence that a decision to file was considered rather than hopeful.
Reading · Element by element
Reading a Result Without Talking Yourself Out of It
The instinct on finding something similar is to conclude the idea is dead. Usually it is not. The comparison that matters is element by element, not impression by impression. Break your invention into its constituent features, then ask of each document: does this single reference disclose every one of those features, arranged this way, doing this job? If it does, the claim is anticipated and no argument saves it. If it discloses four of six features and a second document discloses the other two, the question becomes whether combining them would have been obvious to a person of ordinary skill — a much softer test, and one that is regularly argued and regularly won.
Very often the search does its most useful work by redirection. The broad concept is anticipated, but a detail treated as incidental — the geometry that made it manufacturable, the sequence that made it reliable — is unclaimed and is the actual invention. The vibration-based device described in this account of a pain relief product reaching the market illustrates the pattern: the underlying principle was old and widely known, and the protectable contribution lay in a specific implementation of it.
Limits · What it cannot tell you
The Eighteen-Month Blind Spot
No search is exhaustive, and one gap is structural rather than a matter of diligence. Applications are generally kept confidential for eighteen months from their priority date. Anything filed in the last year and a half is invisible to every searcher on earth, including the examiner who will eventually handle your case. A search run in good faith today can be undermined by a document that publishes next spring with an earlier date than yours.
That is not an argument against searching; it is an argument against treating a clean result as a guarantee. It is also a reason to file rather than to keep polishing. Timing interacts with the route chosen, and the different application types and the proceedings that follow them carry different exposure to that blind spot. A search sits in the middle of a short sequence of decisions — record, silence, search, file — and works best when the other three are already running to the same ninety-day timetable.
Commissioned searches are worth their cost when the field is crowded, when the classification scheme is unfamiliar, or when a professional opinion is needed before committing real money. But the first pass belongs to the inventor, because nobody else knows which features are essential and which were convenience. An hour spent listing what the invention genuinely requires, before opening a single database, improves the result more than any amount of clever query syntax.
Search before you draft, log what you searched, and treat a clean result as encouraging rather than conclusive.