Somewhere in the world right now, an inventor is filing a patent application without having conducted a proper patentability search. They know their idea is brilliant. They have a gut feeling it is new. And that gut feeling, more often than not, is going to cost them.
The numbers tell a frank story. According to data compiled from USPTO records, approximately 86 to 90 per cent of patent applications receive at least one office action rejection before allowance and the single most preventable cause of those rejections is inadequate prior art searching before filing. Of all first-action rejections at the USPTO, roughly 21 per cent are for lack of novelty under Section 102, and nearly 47 per cent are for obviousness under Section-103. Both of these are, at their root, prior art problems – and both can be anticipated, and in many cases avoided, with a properly conducted patentability search.
After decades of working on patent matters across sectors ranging from pharmaceuticals and biotechnology to software and mechanical engineering – and managing a portfolio of over 65,000 patent applications across India and Southeast Asia – I have seen the same mistakes recur with a frequency that is almost predictable. The good news is that they are all avoidable. Here are the five that matter most.
Mistake 1: Searching Only Patent Databases
This is probably the most common error, and it is also the most understandable one. Patent databases – USPTO’s Patent Public Search, EPO’s Espacenet, WIPO’s Patentscope, India’s InPASS – are the natural starting point for any prior art search. They are purpose-built, searchable, and organised. The problem is that they represent only a fraction of the prior art universe.
A significant body of prior art lives in non-patent literature, or NPL as practitioners call it. This includes peer-reviewed journal articles, conference papers, PhD dissertations, technical white papers, product manuals, standards documents, and even YouTube demonstrations and GitHub repositories. An invention that was described in a 2018 journal paper, publicly presented at an IEEE conference, or detailed in a product brochure long before your priority date is prior art – regardless of whether a patent was ever filed around it.
In technology-intensive fields – biotechnology, pharmaceuticals, artificial intelligence, semiconductor design – NPL is often where the most relevant prior art lives. Academic researchers routinely publish findings years before commercial entities think to file patents. If your search covers only patent databases, you may be walking into the examiner’s office with a blindfold on.
Mistake 2: Relying on Keywords Alone
Keyword searching feels intuitive. You know what your invention does, so you type what it is into the database search bar and look at the results. The difficulty is that the same invention can be described in dozens of different ways by different inventors, in different countries, at different points in time. An ‘antibody-drug conjugate’ in one patent family may be described as a ‘targeted therapeutic payload delivery system’ in another. A ‘convolutional neural network for image segmentation’ in one application may appear as a ‘deep learning-based visual parsing algorithm’ in a filing from a different jurisdiction.
Relying solely on keywords produces two related problems. The first is false negatives – highly relevant prior art that your search fails to surface because the language used is different from your chosen terms. The second, which is less discussed, is search fatigue: keyword-only searches often return enormous volumes of irrelevant results, leading searchers to curtail their review before they have found what matters most.
Patent classification systems exist precisely to solve this problem. The Cooperative Patent Classification (CPC), jointly developed by the USPTO and EPO and now adopted by China, Korea, and several other major offices, organises all patents by technical subject matter rather than language. The International Patent Classification (IPC), used under the WIPO system, provides the global equivalent. A well-constructed patentability search uses classification codes to identify the relevant technical neighbourhood, and then applies targeted keywords within that neighbourhood – rather than using keywords as the only search mechanism.
Mistake 3: Searching Only One or Two Jurisdictions
Many applicants – particularly those filing first in India – focus their patentability search on Indian databases, and perhaps a brief look at the USPTO. The reasoning is understandable: they are most immediately concerned with patentability in their home jurisdiction, and the Indian InPASS database is the most convenient starting point. But this approach fundamentally misunderstands how prior art works.
Prior art is global. A patent filed in Japan in 2015, a utility model registered in China in 2017, a patent application published in Germany in 2019 – any of these can be cited by a patent examiner anywhere in the world to reject your application for lack of novelty or obviousness. Patent examiners, particularly at the USPTO and EPO, conduct thorough multi-jurisdictional searches as standard practice. If you have not searched where they will search, you will be surprised by what they find.
According to WIPO’s 2025 World Intellectual Property Indicators, 3.7 million patent applications were filed globally in 2024 – a 4.9 per cent increase and the fastest growth rate since 2018. China alone accounts for over 1.6 millions of those filings annually. The Chinese patent database CNIPA, Japan’s J-PlatPat, Korea’s KIPRIS, and the European Patent Register all contain vast archives of prior art that simply do not appear in a keyword search of the USPTO database alone.
Mistake 4: Defining the Invention Too Narrowly for the Search
One of the subtler – and more consequential – mistakes in a patentability search is defining the invention too narrowly when formulating the search. This happens in a predictable way. The inventor describes their specific embodiment: a particular mechanism, a specific compound, a named process step. The search is built around those specific parameters, and the results reflect what is already known about that exact configuration. The search looks clean. The invention appears novel. Everyone feels encouraged.
Then the patent examiner, who is not constrained by the inventor’s self-description, conducts their own search. They search for the underlying technical concept-the function the invention performs, the problem it solves, the principle on which it operates. And they find prior art that the inventor’s narrowly framed search completely missed.
A prior art search should be conducted at the level of the broadest reasonable interpretation of the inventive concept, not at the level of the specific embodiment you intend to claim. If your invention is a new type of drug delivery capsule with a specific polymer coating, your search should cover not just that polymer, not just capsule drug delivery, but the broader landscape of controlled-release mechanisms, oral drug delivery systems, and polymer-based pharmaceutical carriers. The question to ask is not ‘Has anyone done exactly this?’ but ‘Has anyone solved this problem in any way that would render my approach obvious?’
This broader framing matters because patent examiners frequently combine references in obviousness rejections. Under Section 103 of the US Patent Act – which accounts for nearly 47 per cent of all rejections – an invention can be rejected not because any single prior art document discloses it, but because a combination of two or more existing documents would have made it obvious to a person of ordinary skill in the field. A narrow search that looks only for exact matches will not surface these combination arguments before the examiner does.
Mistake 5: Treating the Patentability Search as a One-Time Exercise
The final mistake is one of timing and process rather than methodology. Many inventors and companies treat the patentability search as a single event: something done once before filing, reported once in a memo, and then set aside. In reality, the prior art landscape evolves continuously, and a search that was thorough in January may be materially incomplete by October if significant new filings have been published in the interim.
Patent applications are typically published 18 months after their filing date. This means that at any given moment, there is an 18-month window of filed-but-unpublished applications that are invisible to your search – but which may become citable prior art during examination if the examiner’s search date is later than your own. In fast-moving fields like artificial intelligence, where AI-related patent filings have increased by 33 per cent since 2018 and now appear in 60 per cent of all technology subclasses according to recent USPTO data, the landscape can shift materially within weeks.
There is also the question of freedom to operate. A patentability search and a freedom-to-operate search are related but distinct exercises. A patentability search asks whether your invention is novel and non-obvious in light of the prior art. A freedom-to-operate search asks whether practising your invention would infringe any existing, in-force patents. Conflating the two – or substituting one for the other – is a surprisingly common and potentially costly error, particularly for companies moving towards product commercialisation.
Why All of This Matters More Now Than Ever
There has never been more at stake in getting a patentability search right. Global patent filings hit 3.7 million in 2024 – the fastest growth rate since 2018 – and the USPTO’s pending backlog reached a record 1.2 million applications by the end of that year. Examination timelines at the USPTO now average 23 to 24 months for standard applications, and the cost of prosecution routinely runs into tens of thousands of dollars even before you consider litigation risk.
In India, where IP filings grew 44 per cent between 2020-21 and 2024-25 and the startup ecosystem continues to expand rapidly, the stakes of a poorly conducted patentability search are equally significant. A rejection on novelty or obviousness grounds that was entirely foreseeable had the search been done properly – means wasted filing fees, prosecution costs, and most importantly, lost time. In a competitive market, losing 12 to 18 months to a predictable prosecution problem is a real business cost.
A well-conducted patentability search is not a bureaucratic formality. It is one of the most consequential pieces of strategic analysis you will commission in the lifecycle of an innovation. Done properly, it is not just a gate-check before filing – it is a source of genuine competitive intelligence.
Author: Vipasha Srivastava. In case of any queries please contact/write back to us via email to [email protected] or at IIPRD.