What is prior art? A guide to patent searches for startup founders

When a startup founder tells me they've "checked for prior art" before filing, they almost always mean they Googled their idea, didn't find an identical product, and assumed they were clear.

But that's not a prior art search; it's a competitor check, and it misses the thing that actually kills patent applications. Read on as we explore what prior art is, how it works in practice, and, more importantly, how to use it strategically rather than just defensively.

What counts as prior art for startups?

Prior art is any publicly available information that existed before your patent’s filing date and is relevant to your invention. In practical terms, it's the body of knowledge a patent examiner will compare your invention against to decide whether it's novel and non-obvious.

Your own public disclosures

This is the trap that catches the most founders. Under U.S. patent law, you get a one-year grace period after your first public disclosure to file. Most other countries, including Europe, China, Japan, and South Korea, offer no grace period at all. Any disclosure before filing creates prior art that can permanently block your international patent rights.

The problem is that "public disclosure" covers far more startup activity than founders expect:

• Demo day presentations. If your accelerator records and posts demo day talks (and most do), that recording is a public disclosure. If you included system architecture details or described how your technology works under the hood, those details are now prior art.
• Investor pitch decks shared without NDAs. Most investor meetings don't happen under NDA, and most pitch decks get forwarded. If your deck includes detailed technical descriptions of how your system works, that information is now in the wild.
• Product documentation, white papers, and blog posts. Founders often publish technical content to build credibility before fundraising. A detailed white paper explaining your novel approach to solving a problem is exactly the kind of disclosure that creates prior art.
• Public product demonstrations (e.g., trade shows): Any of the tech involved in the demonstration would now be prior art.
• Offers to sell a product (even under NDA): This is a tricky one. If you sell your product even if the sale is confidential, the technology being sold will count as prior art

Academic research and preprints

If your CTO published a PhD thesis or conference paper three years ago describing an early version of the approach your startup now uses, that paper could be prior art. It doesn't matter that it was your own co-founder's work or that the startup didn't exist yet; what matters is whether it was publicly available before your filing date.

Beyond your own team's publications, the broader academic landscape is a common source of examination surprises. For AI/ML startups in particular, the arXiv preprint server is a minefield. There are thousands of papers from 2016 to 2020 describing approaches to problems that startups are now commercializing. A paper you've never heard of, written by a research group in another country, can be cited against your claims if it discloses the same technical approach. Similarly, IEEE conference proceedings, ACM publications, and PubMed abstracts all carry the same weight.

Open-source code

If you or anyone on your team contributed to an open-source project that involves core parts of the invention, those commits are timestamped, publicly accessible disclosures. This comes up frequently with ML/AI startups where the founding team was previously active in open-source communities.

But it's not just your own contributions. If an open-source library already implements the core method your invention relies on, that codebase is prior art, even if nobody has used it the way you're using it. The examiner doesn't care about your product context; they care about whether the technical approach was publicly available.

Existing patents from unrelated industries

This is one that surprises many founders. Prior art is not limited to the industry you’re building in. A method for optimizing resource allocation in semiconductor manufacturing might be structurally similar to your approach for scheduling meetings. A logistics routing patent might disclose the same graph traversal technique your startup uses for social network analysis. There are limits to the applicability of prior art, but patent examiners are often quite creative when applying references.

The applications are completely different, but the underlying method might be structurally similar, and that's what patent claims cover. This is why founders who only search within their own industry almost always miss the references that actually show up in office actions.

Products, teardowns, and third-party content

Commercially available products are prior art whether or not they were ever patented. If a competitor (or a company in a completely different space) ships a product that implements the same technical approach, the product's existence is a citable reference.

This also extends to product manuals, technical specifications, standards documents, and third-party analysis. If someone wrote a thorough teardown of a product or technology that happens to describe your inventive approach, that is citable prior art. Tech blogs, YouTube teardown channels, and detailed product reviews can all qualify.

What happens if prior art exists for your invention

Scenario 1: A single reference describes your exact invention

This is a §102 rejection or a “novelty” rejection, where the examiner found one piece of prior art that discloses every element of your claim. It's the most straightforward type of rejection and, paradoxically, sometimes the easiest to deal with.

If the reference matches your claim element-for-element, you have two options: identify a specific technical feature your invention has that the reference doesn't (and amend your claims around it), or argue that the examiner’s interpretation of your claim and/or the reference is not correct.

In practice, "exact match" is rarer than founders fear. The reference usually describes something close but not identical, and the differences, even small ones, can become the basis for patentable claims if they're non-obvious.

Scenario 2: Multiple references combine to make your invention "obvious"

This is the §103 rejection or obviousness rejection, which is the most common outcome for startup patents. The examiner finds two, three, sometimes four or more references that, taken together, cover the elements of your claim.

Here's what most founders don't understand about this: a §103 rejection is the start of a negotiation, not the end of one. The examiner has to demonstrate not just that the references exist, but that a person skilled in the field would have been motivated to combine them in the way your invention does, among other technicalities.

Those technicalities are where experienced practitioners push back. If the combination of references would require hindsight bias to come up with, are incompatible in some functional way, or would require non-obvious modifications to work together the way your invention does, the combination argument has weaknesses you can attack.

Over 70% of patent applications are amended after the first office action. This is normal prosecution. A first rejection based on prior art doesn't mean you're done; instead, it means you're entering the back-and-forth phase where claim scope gets negotiated between your practitioner and the examiner.

Scenario 3: Prior art covers the broad concept but not your specific approach

This is actually the most common and most strategically interesting outcome. You find (or the examiner finds) prior art that addresses the same general problem but uses a different method, architecture, or combination of features.

This kind of prior art doesn't block your patent. It shapes it. Instead of claiming the broad concept ("a system for optimizing delivery routes using machine learning"), you claim the specific way you've implemented it, whether it’s the particular model architecture, novel data preprocessing step, or specific combination of real-time inputs that makes your approach work.

The practitioner's instinct here is to ask: what did they NOT cover? Every reference has gaps, including specific features, implementation details, or combinations that it doesn't disclose. Those gaps are your claim territory. A good prior art analysis maps the white space where your invention can be most strongly protected.

Scenario 4: Your own disclosure is the prior art

We covered the types of self-disclosure that create prior art above. What's worth understanding separately is what happens when this comes up during examination, because unlike the other scenarios, the options are severely limited.

With a §102 or §103 rejection based on third-party prior art, your practitioner can argue claim amendments, challenge the examiner's interpretation, or attack the combination of references. With a self-disclosure problem, the prior art is often your own words describing your own invention, which means it’s usually a much closer match to your pending claims.

If the disclosure happened within the one-year U.S. grace period, you can file a declaration establishing that the disclosure can be traced back to you. But if you're outside that window, or if you need international protection in countries with no grace period, there's no prosecution strategy that fixes it. The claims covering whatever you disclosed are simply unavailable.

This is why filing even a provisional application before any public disclosure is the single most important IP decision a startup founder can make. 

What does prior art mean for your fundraising efforts

This is the angle most prior art guides miss entirely, but for startup founders, it's one of the most practical reasons to take prior art seriously early.

Investors increasingly review IP as part of due diligence, especially at Series A and beyond. A patent due diligence checklist typically includes questions about the novelty and defensibility of your core technology. If you've filed a patent but haven't done a prior art search, you're essentially telling investors "we will have IP protection" without knowing whether that protection will hold up under examination.

The timing trap is real: founders often want to file a patent before raising a round (smart), but that might cost money they don’t have. As we covered above, those public presentations can become prior art that limits your claims. The sequencing matters — filing even a provisional application before your first investor meeting establishes a priority date that protects you against your own later disclosures.

A prior art search done before filing also gives you a much stronger story to tell investors. Instead of "we filed a patent," you can say "we conducted a prior art analysis, identified the white space in our technology area, and filed claims specifically targeting the defensible territory." That's a materially different signal about the quality of your IP position.

How startup founders should approach prior art

Now that you understand what prior art is, how it shows up, and what it means for your patent, the practical question is: what do you actually do about it? There are several paths, and they vary significantly in cost, rigor, and risk.

Skip the prior art search entirely

This is what most founders do by default either because they don't know much about the nuances, or because they want to move quickly and cheaply. You go straight to drafting and filing, either yourself or through counsel, without anyone doing a dedicated search beforehand.

The risk here is significant. Without a search, you're filing claims in the dark. You might be claiming territory that's already well-covered by existing patents, or drafting broad claims that will get immediately narrowed during examination. You might spend $10,000+ on a filing only to discover two years later that the core of your invention was disclosed in a 2017 arXiv paper nobody on your team had seen.

Do a prior art search yourself

Some founders try to get ahead of the problem by searching Google Patents, Google Scholar, and similar free tools on their own. While the instinct is right, in my experience this approach almost always does more harm than good.

The problem is that a non-professional search creates a false sense of security. You'll find some references, miss others, and come away thinking you have a clear picture when you don't. Patent examiners use professional databases, classification-based searching, and years of experience in specific technology areas. A keyword search on Google Patents, no matter how thoughtful, won't replicate that.

There's also a practical risk most founders don't know about: many references you find during a self-directed search may need to be disclosed to the USPTO in an Information Disclosure Statement (IDS). A sloppy search can create disclosure obligations without providing reliable insight, which is the worst of both worlds.

Hire a standalone prior art search firm

For founders who want professional-grade intelligence before committing to a full filing, dedicated prior art search firms are a strong option. These firms employ trained searchers with access to professional patent databases (Derwent, PatSnap, Orbit) and non-patent literature sources that go well beyond what's available through free tools.

A professional search typically runs $500 to $1,500 depending on the complexity of the technology area, and it produces a structured report identifying the closest prior art references along with an analysis of how they relate to your invention. This is especially valuable for founders in crowded technology spaces where the volume of potentially relevant prior art is high.

The limitation is that a search report is just that — a report. You still need to figure out what to do with it. Most founders bring the results to a patent attorney or agent who then uses them to inform the drafting and claim strategy. That's effective, but it means you're managing two separate vendors and stitching the process together yourself.

Have your patent counsel handle search and filing together

This is the traditional law firm approach. You engage a patent attorney or agent, and they handle the prior art search, claim strategy, drafting, and filing as one engagement. The search informs the claims from the start, and you're working with a single team throughout.

The advantage is integration: the same people who find the prior art are the ones positioning your claims around it. The disadvantage is cost. At a traditional patent law firm, this end-to-end process typically runs $15,000 to $20,000+ for a single utility patent over the course of the application process, which is a significant commitment for an early-stage startup.

Use Patentext

Patentext takes a fundamentally different approach to the prior art problem. Instead of treating it as a separate step you have to manage before the filing process begins, Patentext runs a prior art search based on the structure of your invention in the background and uses the information it finds to help you make filing decisions before you even start drafting. You understand the novelty and non-obviousness of your invention before you've committed thousands of dollars to a filing that may need significant rework.

But the prior art piece is just the starting point. Patentext is the first end-to-end patent filing service built specifically for startups. We handle everything from patent strategy and prior art analysis through AI-powered drafting, patent agent review, and filing with the USPTO, all for a fraction of what traditional law firms charge. You're getting a single, integrated process where every step is informed by what came before it.

Our team has filed and prosecuted thousands of patents across deep tech, AI/ML, hardware, chemical processes, and SaaS. We built Patentext because we saw the same problems over and over: startup founders either couldn't afford quality patent work, or, when they could, the process was so painful that the patent portfolio never reached its full potential. Patentext solves these problems: rigorous patent filings at startup-friendly pricing, with a focus on an approachable and transparent user experience

If you're a founder looking to protect your IP without the $15K+ law firm price tag, join our waitlist (and lock in 25% off when we launch)

‍

This article is for informational purposes only and does not constitute legal advice. For guidance specific to your invention and situation, speak with a qualified patent professional.

‍