Since 2025, the U.S. patent system has moved into a period of active recalibration, with the PTAB at the center of policy, legal, and industry debate. Recent congressional discussions and USPTO actions frame this shift as an effort to restore balance by prioritizing patent quality, predictability, and rights that reliably support investment and commercialization.
At the same time, inventors, startups, and universities have raised a more fundamental concern: over the past decade, patents were often drawn into administrative review and weakened or invalidated before being tested in an Article III court. For stakeholders that depend on enforceable rights to license technology or build ventures, this has translated into uncertainty at the point where certainty matters most: commercialization.
The current reforms respond to that tension but introduce a new one. Declining institution rates, expanded discretionary denials, and increased Director control are reshaping access to PTAB.
For some, this is a necessary correction to an earlier imbalance; for others, it raises questions around transparency, consistency, and the ability of accused infringers to obtain timely review.
As a result, the debate has shifted beyond procedure to structure: how PTAB should interact with district courts, and where validity disputes should ultimately be resolved. With multiple reform proposals under discussion and the Director’s role becoming increasingly influential, the system is not just evolving, it is being redefined in real time.
This article brings together the latest developments, data, and practitioner insights to map what is happening at PTAB today and, more importantly, what it means for litigation, licensing, and patent strategy going forward.
01 – The Numbers That Redraw Every Calculation
Three years ago, filing an inter partes review (IPR) petition was a reliable form of leverage.
An institution rate hovering around two-thirds meant that challengers entered negotiations with a credible threat: proceed to trial and watch the claims dissolve.
That dynamic no longer exists.
The PTAB’s evolving posture, accelerated sharply since late 2025 under Director John Squires, has produced one of the most dramatic reversals in the system’s history.
Where institutions were once nearly automatic for well-prepared petitions, the gate has narrowed to the point where a strong set of prior art no longer guarantees a hearing.
| Period | Institution Rate | Environment |
| 2024 | 67% | Open Gate, near-standard review |
| 2025 | 44% | Slowing Gate, merit scrutiny rising |
| 2026 YTD | 21% | Closed Gate, record institutional low |
The Oct 2025 – Jan 2026 window makes the picture even starker. During that period, the Director assumed personal control over institution decisions.
Half of all petitions were discretionarily denied outright, many for vagueness or structural weakness rather than a merits-based examination of the art itself. Only about one in ten cases reached full trial.
The Board now often issues Summary Notices, bare lists of case numbers marked “denied” or “instituted,” without reasoned opinions, leaving petitioners with little diagnostic information about why they failed.
“A petition that reads like a dry claim chart is more likely to be denied than one that builds a compelling case for why this patent should never have been granted.”
Is a challenge still viable for your patent?
The sector breakdown adds another layer of complexity. Not all technology fields are treated equally under the new regime. A challenger in electrical and computer science faces a strikingly different environment than one operating in biopharmaceuticals.
| Technology Sector | Institution Rate (Since Mar 2025) | Challenge Level |
| Electrical / Electronics | 29% | Hardest to challenge |
| Mechanical & Business Methods | 44% | Moderate |
| Biopharmaceutical & Chemical | 56% | Most viable for challenge |
For anyone working with electrical and computer patent challenges, these numbers demand a fundamental reassessment of when an IPR is even the right vehicle. Proceeding without clean, non-cumulative, strong prior art in that sector is not a calculated risk; it is the near-certain path to a summary denial.
02: The Six-Year Rule: An Unofficial Threshold That Now Governs Scoping
The concept of “settled expectations”, the idea that a patent which has anchored a market for years deserves deference, has been debated for years.
In 2026, it had hardened into something approaching a bright-line rule, even if the statute does not formally require it.
The practical guideline articulated by patent invalidity experts at GreyB is direct. If a patent is more than six years old, an IPR petition will be discretionarily denied in the vast majority of cases on settled expectations grounds.
This is not a statutory rule, and edge cases exist. Still, they are difficult to prevail on, and attempting to overcome the threshold without a credible counterargument is a poor use of resources.
| Patent Age | Likely Outcome | Notes |
| Under 6 years | IPR may proceed | Standard analysis applies |
| Over 6 years | High risk of discretionary denial | Settled expectations likely to apply |
| Over 6 years + edge case | Possible to proceed | Very rare; requires strong additional arguments |
The two edge-case arguments that practitioners are using to bypass this threshold involve demonstrating either limited competition combined with high R&D investment (to argue the “clock” for settled expectations should start later than the filing date) or knowledge timing, showing the petitioner only recently became aware of the patent through a first-time assertion. Neither path is reliable without documentary support assembled before scoping even begins.
Operational Implication- Patent Families Carry Inherited Risk
A growing tactical hazard: patent portfolios with families in which some members are under 6 years old and others are over. The PTAB has, in rare cases, grouped newer patents with older ones, effectively applying the older patent’s settled expectations to the entire family.
This means a challenger can invest heavily in preparing a petition against a patent that appears new on its face, only to have the proceeding denied because it is treated as an extension of an older family member. Filing date checks must now run across the entire family, not just the asserted patent.
For older patents that do clear the threshold, the bar for the prior art itself is substantially higher. Where a patent is more than six years old and has been commercially deployed, the standard shifts from “good art” to “compelling art”, evidence that proves the original examiner made a clear, demonstrable error. Art that narrows claims or creates licensing leverage may still serve the client, but it will not survive a discretionary review on its own.
03: The Storytelling Imperative: From Claim Charts to Narrative
The shift in how petitions are reviewed at the Director level has made the quality of argument, not just the quality of art, decisive.
USPTO Deputy Director Coke Stewart articulated the principle clearly: a patent is not a list of parts; it is a solution to a specific problem.
An examiner, or in this case, a Director reviewing a petition, who understands why something was invented, can better judge whether the art being cited reflects a genuinely new leap or merely recombines prior ideas.
That understanding now shapes discretionary decisions.
“If it takes a mountain of paperwork to prove an idea is obvious, it probably wasn’t that obvious. A truly bad patent should be easy to debunk with one or two clear examples of prior art.”
The two-way combination of references, the workhorse of many obviousness arguments, must now be airtight. If the best available argument requires stacking four or more documents, the current PTAB is likely to view that as evidence of a weak case, not a thorough one. Quality over quantity is no longer an aspirational framing; it is what the institution standard now demands.
A strong petition story contains three elements working together:
- A clear narrative of the patent’s history: why it was filed, what specific problem it claimed to solve, and what the prosecution record reveals about how the examiner reasoned through it.
- Evidence that the problem was already solved: prior art demonstrating that the claimed solution existed before the priority date, presented in terms of what that reference was designed to achieve, not just which claim elements it maps to.
- A concise explanation of why the original examiner got it wrong: a specific, locatable error rather than a general suggestion of obviousness assembled across multiple references.
When presenting prior art, the reference’s context matters as much as its technical content: what problem it solved, when it solved it, and how that framing makes the challenged patent’s novelty claims implausible. Without this context, even technically strong art reads as an incomplete argument under the current review standard.
Also Read:
04: EPR’s Rise and the Critical Distinctions That Determine When to Use It
As IPR institution rates have collapsed, ex parte reexamination has emerged as the primary alternative.
EPR requests increased by 66.1% over 2024, reaching 726 requests, and 92–95% of those requests are being granted, compared with the institution rate of 21% for IPR petitions.
But EPR is not a substitute for IPR in most situations. It is a different instrument with different properties, suited to specific fact patterns. Both patent owners and petitioners, when given a choice, still prefer IPR when it is available.
Also Read:
The reasons are operational.
IPR: When to Choose It
Complex §103 obviousness arguments requiring expert testimony. When settlement optionality must be preserved. When speed and a defined 12-month schedule matter. When the petitioner needs to control the prior art record.
EPR: When to Choose It
Clean, single anticipatory §102 references. When attacking obvious patent-family variations (double patenting, not available in IPR). When the institution threshold of IPR cannot be met. When the CRU’s ability to independently introduce art is an advantage.
Two features of EPR require explicit client communication before any filing decision. First, the “no exit” problem: once an EPR is requested, the Central Reexamination Unit continues the examination to its conclusion regardless of whether the requester withdraws. If a client files an EPR primarily as negotiation leverage, they must understand there is no off-ramp if a licensing deal is reached mid-process. Unlike IPR, there is no settlement mechanism to end the proceedings.
Second, the “gap-filling” dynamic cuts both ways. CRU examiners are not passive; they can independently search for and introduce prior art beyond what the petitioner submitted. For a petitioner with imperfect art, this can be a safety net.
For a patent owner, EPR is a less predictable battleground than IPR because the examiner is not bound by the record submitted.
Advanced Tactic: The “Fire and Forget” Strategy
Running IPR and EPR in Parallel
The CAFC’s ruling in In re Gesture Technology Partners confirmed that a client may file an EPR while an IPR is pending. Because the USPTO, not the petitioner, maintains the EPR, the petitioner can settle or exit the IPR, and the EPR continues independently.
The strategic play: use IPR to apply pressure on settlement and to have complex §103 arguments handled by PTAB judges. Simultaneously file EPR on the same or related patent using §102 or double-patenting grounds. If the IPR resolves through licensing, the EPR lives on. The patent owner cannot simply pay off both proceedings; they face two parallel, independent challenges designed to attack on different grounds.
This works best when there is both a clean anticipatory reference and a complex obviousness combination. The two proceedings should be designed to complement rather than overlap, different grounds, different forums, different timelines.
05: After a Lost IPR: Physical Products and the Estoppel Escape
The CAFC’s ruling in Ingenico v. IOENGINE established one of the most operationally significant principles of the 2025–2026 period: estoppel from a lost IPR does not touch physical products.
The estoppel rule is well-established in broad terms: a petitioner who loses an IPR cannot use the same printed publications (patents, papers, manuals) to challenge the same claims again in district court.
What the Ingenico ruling clarified is what estoppel does not cover.
| Type of Prior Art | Estopped After Lost IPR? | Usable in District Court? |
| Printed publications (patents, papers, manuals) | Yes, blocked | No |
| Physical products/devices | No, NOT blocked | Yes |
| On-sale bar evidence | No, NOT blocked | Yes |
| Public use evidence | No, NOT blocked | Yes |
Even when a physical device is substantially identical to a manual or publication that was used in the IPR, the device itself is treated as a completely separate invalidity ground. This means that physical product evidence and on-sale bar documentation are now premium deliverables for any challenge that may end up in district court, whether or not the IPR succeeds.
Running a parallel track for historical product releases, trade show materials, sales records, and physical implementations of the claimed technology that predate the patent’s priority date is no longer optional for comprehensive challenge work. This evidence creates a second path in the district court, even when the primary printed art track is estopped.
06: Machine Learning Patents: Where the §101 Line Now Sits
The CAFC’s decision in Recentive Analytics v. Fox has drawn a hard line that directly affects how ML and AI-related patents are evaluated. The case involved claims for applying machine learning to TV scheduling and network map optimization. The court invalidated the patents under the Alice doctrine’s two-step §101 analysis.
The operative standard from the ruling:
| Alice Step | Question | CAFC Standard |
| Step 1 | Is the claim directed to an abstract idea? | Applying generic ML to a known field (scheduling, optimization) = abstract idea. Fails Step 1. |
| Step 2 | Is there an inventive concept, “something more”? | Must show the ML changes the machine itself or delivers a specific technical improvement. Using existing ML in a new field is NOT enough. |
“High-level functional language”, describing what an ML system does without explaining how, is now a near-automatic failure at Step 2. Being the first to apply machine learning to a particular industry or problem is not patentable under this standard. The patent must show a specific, concrete improvement to the underlying computer or algorithm itself.
This has two distinct consequences for how ML-related patent challenges are approached. For claims that rely on a generic ML application to a domain problem, the §101 vulnerability in the district court may be the most effective attack path, even if prior art search doesn’t produce a “killer” reference. That angle should be flagged to counsel regardless of what the art search yields.
For claims that do articulate a specific technical improvement to the ML model itself, the prior art search must go considerably deeper than standard patent databases. ML conference proceedings (NeurIPS, ICML, CVPR), academic preprints, and open-source implementations are the most relevant sources for this type of claim, and they are exactly the category of non-standard sources that current AI-assisted search tools systematically miss.
07: The Texas Effect: Litigation Is Up, and the Search Goal Has Shifted
The weakening of IPR as a default option has not slowed patent enforcement; it has redirected it. Patent litigation filings have increased 12%, and NPE litigation has grown 18.7% (adding 2,236 new defendants). The Eastern District of Texas has emerged as the dominant venue for a simple operational reason: its fast-track docket compresses timelines, amplifying pressure on defendants who can no longer neutralize enforcement risk with a cheap IPR threat.
The old equilibrium, in which the threat of a 70% claim-kill rate under IPR gave petitioners default-settlement leverage, has collapsed.
Patent owners, including universities and small businesses that previously avoided enforcement due to IPR exposure, are now asserting portfolios with confidence. Voluntary pre-litigation licensing has increased because both sides can now engage without the IPR “nuclear option” distorting the negotiation from the outset.
“We are no longer just searching for a kill switch. The goal is now to claim narrowing and licensing leverage, finding art that forces the owner to limit their claims, even if the patent survives.”
This shift changes what a complete deliverable looks like. Search output now needs to serve two parallel tracks simultaneously: the invalidation track for litigation counsel, and the negotiation track for licensing strategy. Art that cannot kill a patent outright may still force a narrowing of claims that protects the client’s product. Non-patent literature documenting “standard industry practice” at the priority date gives the licensing team evidence for lower royalty rate arguments even when validity is upheld.
Texas venue timelines eliminate the option of staged, phased search work. An Eastern District docket moves fast enough that a top-tier search must begin immediately on any case heading to that venue, with no runway for extended deep-dive phases once a trial date is set.
08: Operational Checklist: What Changes at Every Stage
Taken together, the 2026 PTAB landscape requires adjustments at intake, during search scoping, and in how results are presented and structured.
The following checklist consolidates the operationally actionable items from across all eight PTAB Masters sessions.
Patent age check is mandatory before scoping.
If the patent is six years or older, flag it before investing in search. Discuss discretionary denial risk with the client before committing resources. For patents over six years, determine whether compelling edge-case arguments (limited competition, knowledge timing) are available.
Check the entire patent family, not just the asserted patent.
Identify all parent and sibling patents and their filing dates. Flag any family member over six years old, as it may pull the whole family under the settled expectations review.
Verify all RPIs before any IPR filing.
The law has reverted to the strict 2015 standard. Failing to name all Real Parties in Interest, including parent companies, subsidiaries, and affiliates, is a jurisdictional defect that causes immediate dismissal, before the art is even reviewed.
The first shot must be the best shot.
Roadmapping, filing a weak first petition with the intention of strengthening it later, is being denied. Repeat challenges are not a viable recovery strategy. Research must be exhaustive the first time.
Run a parallel physical product track for each challenge.
Even if the primary strategy is print-based, simultaneously document historical product releases, trade show materials, sales records, and public implementations predating the priority date. This evidence is not estopped after a lost IPR and creates a district court path.
Cross-reference every result against the IDS.
If a reference was already cited in the patent’s Information Disclosure Statement, the search must identify a material error in how the examiner applied it, not merely that the reference exists. The Substantial New Question must be genuinely new to the prosecution record.
Search sibling patents for double patenting grounds.
EPR is the only venue where a patent can be challenged as an obvious variation of its own family members. Identify sibling patents as part of standard scoping for any EPR-track project.
For ML patents, flag §101 vulnerability to counsel regardless of art.
If the claims use high-level functional language without articulating a specific technical improvement to the ML system itself, document this and flag it. The district court §101 path may be more reliable than prior art, even absent a killer reference.
Do not rely solely on AI-assisted search.
AI tools cannot reliably interpret diagrams and figures in patents, and do not cover grey literature, regional filings, product manuals, and niche technical publications. Manual review of figures and non-standard sources remains essential, particularly where the key claim element is illustrated rather than described in text.
Warn clients before EPR filing if settlement is the goal.
An EPR cannot be withdrawn or settled once filed. If a client is using it primarily as negotiation leverage, they must understand it is a one-way door. If a licensing deal is reached mid-process, the CRU continues regardless.
09: The Case to Watch: Volkswagen and the Non-Delegation Argument
No immediate action is required from practitioners, but one case on the CAFC’s docket carries the potential to restructure the entire framework described in this article. Volkswagen is advancing a constitutional challenge to the Director’s discretionary authority, arguing that giving the USPTO Director essentially unbounded power to deny IPR petitions without clear congressional guidelines violates the Non-Delegation Doctrine.
The CAFC has consistently rejected Mandamus petitions from frustrated petitioners seeking to force institution, establishing clearly that there is no constitutional right to have an IPR instituted, and that the Director’s broad discretion is legally sound under current doctrine. Volkswagen’s argument operates at a different level: not that the Director exercised discretion incorrectly, but that Congress cannot constitutionally delegate unchecked authority of that kind to an executive agency without defining its boundaries.
If the argument succeeds, the practical consequence would be substantial. Congress or the courts would be required to establish clearer, mandatory rules for when an IPR must be instituted. The six-year threshold, the settled expectations framework, and the current wave of summary discretionary denials could all be restructured as a result. The entire operating environment described in this article, including how searches are scoped, how petitions are prioritized, and what “compelling art” means, would need to be reassessed.
The outcome is uncertain, but the direction of the argument is worth monitoring closely. A ruling in Volkswagen’s favor would be the most significant PTAB structural change since the AIA itself.
Invalidity Has Not Gotten Harder. It Has Gotten More Honest
In the past, you could rely on IPRs as a primary route. That path is now less certain.
Lower institution rates and higher discretionary barriers mean each challenge must be stronger from the start. There is little room to test weak arguments. Instead of asking “what maps to this claim,” the better question is where the actual inventive step sits and whether it was genuinely new.
Most independent claims don’t carry novelty across every element.
In one audio technology case, comparing two figures in the patent itself revealed that the claimed novelty was a single technical operation. That observation led to a 1993 Canadian patent that had documented both implementation approaches decades before the priority date, and it was found with a three-word query, not an exhaustive keyword sweep. In another, the breakthrough came from tracing inventor-collaborator networks across 2,000+ sources, in which following one co-author’s work surfaced IBM-assigned prior art covering both independent and dependent claims.
The searches behind most active cases were scoped for a different PTAB. A 67% institution rate allows for broader, more speculative art. A 21% one does not. That gap is where petitions are failing before the art is even reviewed.
Most matters at this stage raise the same hard questions:
- Does the patent family have members old enough to trigger discretionary denial, and has that risk been mapped before scoping begins?
- Is the prior art record genuinely exhausted, or was the search built around claim vocabulary rather than the invention’s technical history?
- If an IPR has already failed, is there a live district-court path via physical-product evidence, or a parallel EPR track that hasn’t been assessed?
- Does the technology sit in a field where the decisive art is in conference proceedings, product lineage, or inventor networks rather than standard patent databases?
If you are working through an active IPR, reassessing a failed petition, or trying to understand how the 2026 PTAB environment affects a specific patent family, fill out the form below. GreyB’s invalidity team will review the patent, identify which of these risk factors apply, and tell you whether a viable challenge path still exists.