LKQ v. GM Ends Rosen-Durling: Graham Factors Now Rule

At least four separate legal publications reported the same core holding in a single news cycle: the Federal Circuit, writing en banc with the full court sitting together, has overruled the long-standing Rosen-Durling test for design patent obviousness. The case is LKQ Corp. v. GM Global Tech. Operations LLC, and the disposition language converged across outlets — 'overruled,' 'invalidated,' 'ends' — all pointing to one outcome.

What fell was a two-part inquiry whose first step demanded a single primary reference: one designated starting design against which the accused design had to be compared. Commentators singled out that gate as the rigidity problem, since it forced every obviousness analysis through one 'basically the same' design before the comparison stage could even begin. Without such a reference in the prior art, a validity challenge often stalled at step one.

What replaces it is familiar ground for utility-patent practitioners: Graham factors. The court found Rosen-Durling's requirements too rigid to survive Supreme Court precedent in KSR International Co. v. Teleflex, and aligned design-patent obviousness with the utility-patent approach. For prior art searches, the practical shift is immediate — stop hunting for a single matching primary reference and start mapping how combinations of references would have rendered the claimed design obvious.

How It Works

On May 21, the Federal Circuit didn't amend design patent obviousness law — it swapped the operating system. Writing en banc, meaning the full court sat together rather than a routine three-judge panel, the court invalidated the Rosen–Durling test outright, according to the analysis in "Deere in the Headlights: Federal Circuit Invalidates Rosen-Durling Test." The en banc posture is the mechanical detail most summaries skip: a panel decision invites years of panel-to-panel drift, while a full-court ruling locks the replacement framework in circuit-wide. Nobody waits for a friendlier panel to soften the change.

The demolished machinery ran in two parts. Under Rosen–Durling, an examiner or challenger first had to locate a single primary reference "basically the same" as the claimed design, then show secondary references modifying it to yield the same overall visual appearance. That two-part inquiry collapsed because it collided with the Supreme Court's holding in KSR International Co. v. Teleflex Inc. — per Sterne Kessler's coverage, KSR is the named anchor for the flexibility requirement that Rosen–Durling failed. KSR rejects rigid, checklist-style obviousness tests, and a rule forcing every design case through one "basically the same" gate is rigidity in its purest form.

What runs instead is the Graham framework — the same multi-factor apparatus long used for utility patents, now expressly extended to designs through post-overruling judicial guidance, as reported by both "Full Federal Circuit Overrules Old Design Patent Test" and the "Deere in the Headlights" commentary. Mechanically, the search changes shape: rather than hunting for one near-identical design and stopping once found, you map the scope and content of the field, measure the claimed design's differences against a person of ordinary skill in the art, and articulate why a designer would have arrived at the claimed appearance from that art. Commentary titled "Overhaul of the Rosen-Durling Test" brands the shift "A New Era for Design Patent Obviousness," and the era's defining feature is that a primary reference now serves as a starting point for reasoning, not a jurisdictional admission ticket.

One myth deserves burial here: that the old two-step was a lean convention and Graham simply piles on costly, unnecessary steps. It's backwards. The two-step wasn't efficient — it was a filter that discarded most of the prior-art universe before obviousness was ever assessed, and art packages organized around it are now framed for a test that no longer exists. The Graham extension doesn't add bureaucracy; it redirects effort toward the documentation courts will actually credit.

Key terms, and their status after the ruling:

The concrete move for 2026: pull every design clearance or invalidity opinion drafted before the May 21 ruling, and check whether its prior art was selected and argued around a single "basically the same" primary reference. If it was, rebuild the analysis on Graham-factor scaffolding before relying on it in an office-action response or a PTAB petition — the underlying references may survive, but the argumentative architecture built for Rosen–Durling will not.

Key Factors to Consider

Zero. That is the number quietly reorganizing design patent prior art searching after LKQ Corp. v. GM Global Tech. Operations LLC: the count of "basically the same" primary references a searcher is legally required to find. Under Rosen-Durling, the entire obviousness analysis hung on locating one prior design creating the same visual impression as the claimed design — a search that surfaced none was, functionally, finished. The en banc court deleted that prerequisite when it moved design patents onto the Graham framework, so the anchor-first protocol many search teams still run is executing a step the law no longer contains. The court said it plainly, as reproduced in Redesigning Design Patent Validity: Rosen-Durling is "out of keeping with the Supreme Court's general articulation of the principles underlying obviousness."

Three criteria now determine whether a search protocol fits the post-Rosen world. First, analogous-art scope. The court's guidance on Graham factor one turns on the analogous-art question, according to the Full Federal Circuit Overrules Old Design Patent Test analysis — the old same-field-of-endeavor ceiling is gone, replaced by whether an art form bears on the designer's problem. Semantically, that converts field membership (a coarse categorical filter, e.g., a Locarno class) into problem-to-problem similarity (a graded judgment). A search keyed to classification codes alone now under-covers; one that maps where the same visual problem recurs — fastening, airflow, seam treatment — tracks the factor as written.

Second, combination logic. According to Sterne Kessler's account, the court found Rosen-Durling too rigid — incompatible with KSR International Co. v. Teleflex Inc., which demands a flexible approach to obviousness. The decisive search output is therefore no longer a single exhibit but a set: two or more references whose identifiable visual features, combined as a designer of ordinary skill would, yield the claimed ornamentation. Capture why a reference would have been combined — a shared design problem, a market pressure, a modular interface — because under KSR-style review, "could be combined" without a stated reason reads as hindsight.

Third, evidence timing. Under the bespoke two-step test, objective indicia were peripheral; under the aligned framework, Graham factor four — commercial success, copying, industry acclaim, long-felt need — carries independent weight. Build that record during prosecution, while products are in market, rather than reconstructing it after an accusation lands.

The numbers worth memorizing are structural, not statistical. Two steps became four factors: the bespoke Rosen-Durling sequence gave way to the full Graham analysis — prior art scope, differences, level of ordinary skill, secondary considerations — which, as the Federal Circuit Overrules Rosen-Durling Test For Design Patent Obviousness coverage frames it, aligns design patents with the utility-patent approach. One Supreme Court case, KSR, supplied the demolition charge. And the mandatory anchor count fell from one to zero. The concrete next move: rewrite your search memo template so the analogous-art map and the ordinary-designer definition precede any reference pull — under the new framework those two inputs define what counts as prior art, and retrofitting them afterward is how searches come back legally incomplete.

Common Mistakes

The costliest prior art error after LKQ Corp. v. GM Global Tech. Operations LLC is not missing references — it is deleting them. Under the old regime, independent coverage in "Overhaul of the Rosen-Durling Test: A New Era for Design Patent Obviousness" described a two-step process whose first step demanded "a single primary" reference that was "basically the same" as the claimed design. Searchers internalized that screen so deeply that many still discard anything failing it. That reflex is now backwards: the references that fail the "basically the same" test are precisely the secondary art a Graham-factor analysis consumes. The myth to retire is that a search produces nothing usable until one dominant reference appears — the en banc Federal Circuit removed that gatekeeper when it, per the "Deere in the headlights" commentary, "aligned the obviousness standard for design patents with the approach used for utility patents."

Consider a concrete triage, in the product category the caption itself names — exterior vehicle panels, the territory contested between challenger LKQ Corp. and GM Global Tech. Operations LLC. Say the claimed design is a fender with a distinctive wheel-arch flare crossed by a character line. A legacy-filtered search surfaces Reference A (matching flare, different crease), Reference B (matching character line, flat arch), and Reference C (a quarter panel showing how the flare transitions across a body side). The old habit bins all three: none is "basically the same." The post-ruling read promotes A to the Graham starting point, feeds B in as the modification evidence, and uses C to establish that a designer of ordinary skill had the ornamental vocabulary to bridge them. Applicants make the mirror-image error in office action responses, arguing "no single reference is basically the same" — language that now signals you have not engaged the framework the court adopted.

Pitfall two is the overcorrection: reading the alignment with utility practice as a license to stack references and declare the result obvious. It is not. The pre-ruling advocacy recorded in "LKQ En Banc Argument Suggests CAFC Could Soften Test for Design…" attacked the rigid test as "inconsistent with the dictates of the Supreme Court" — but the remedy the court supplied is Graham's structure, including its demand for a reason a designer of ordinary skill would arrive at the claimed design. In design cases that reason lives in shared ornamental language, not problem-and-solution functionality talk. An invalidity position listing five references with no articulated path loses; so does one ignoring that a combination producing a different overall visual impression supports the patentee, not the challenger.

The immediate move: pull the references your last search rejected at the "basically the same" stage, re-tag them by shared ornamental feature, and strip the retired vocabulary from every response template — then verify your framing directly against the en banc opinion rather than any summary, including this one.

Insider Tactics

Retire the anchor hunt. The highest-leverage change to a design patent search after LKQ Corp. v. GM Global Tech. Operations LLC is behavioral, not doctrinal: stop treating retrieval as a quest for one "basically the same" reference and start building factor-tagged clusters. Under Durling, a secondary reference had to be so tightly coupled to the primary that combining them was nearly automatic, so searchers rationally optimized for a single dominant lookalike. According to "Deere in the headlights: Federal Circuit Invalidates Rosen-Durling Test," the replacement regime is explicitly "A More Flexible Approach" — and flexibility flips the optimization: any reference can now enter the combination, so the marginal value of a non-lookalike hit exceeds the marginal value of a tenth near-duplicate.

The disposition language confirms this is permanent infrastructure, not a pendulum awaiting reversal. One outlet states the court "overruled the long-standing Rosen-Durling test used to evaluate obviousness of design patents"; another headlines "Invalidates"; this guide's own title reads "Ends Rosen-Durling." When independent coverage converges on three different verbs pointing the same direction, budgeting for reinstatement is the expensive fantasy.

Non-obvious strategy: run three parallel streams — design patents for visual similarity, utility patents for functional overlap, and trade literature for what an ordinary designer of the relevant type plausibly knew. Tag every capture with the Graham factor it feeds (ornamental similarity, function, ordinary-designer knowledge), and treat "similar" as a graded vector rather than a binary; the office-action response then drafts itself from the tags. From the computational side, add one diagnostic: search prosecution histories for "Rosen" and "Durling" separately, not just the compound name. Commentary pairs the surnames uniformly, but office actions cite In re Rosen and Durling v. Spectrum Furniture Co. individually — and a rejection leaning on either surname alone signals an examiner still reasoning under repealed law. Translate that rejection into Graham factors yourself rather than litigating the dead test on its own terms.

This is also where the money myth dies. The belief that the conventional approach wastes money on unnecessary steps gets the accounting backwards: the "extra" streams — utility art, catalogs, designer-knowledge sources — are precisely the Graham inputs that decide outcomes. According to coverage of the en banc argument ("LKQ En Banc Argument Suggests CAFC Could Soften Test for Design Patents"), critics charged that Rosen-Durling "has allowed large corporations to seize monopolies for designs," meaning challengers now hold doctrinal room they previously lacked. A trimmed search is the cheapest way to donate that room to them.

Timing tip: two levers. First, widen the lookback window backward — tight-coupling logic concentrated search budgets in the months before the priority date, but motivation evidence usually lives upstream, in the years when the style lineage formed. Second, sequence defensive audits oldest-first: grants prosecuted under the strictest version of the old test carry the thinnest combinability records, so they fall first to a challenger pairing distant references. For pending rejections written in legacy vocabulary, put Graham-mapped art on the record with your next response; after allowance, the same disclosure routes through costlier supplemental proceedings.

If you execute only one move, run the corpus split — it generates the raw material every other move consumes. Concrete close: pull your ten oldest active design patents, search each file history for both surnames, and rank them by how many Graham factors the record never addressed. That ranked list is your exposure queue. Verify current USPTO design-patent examination guidance before relying on either framework's vocabulary, since agency materials continue absorbing the ruling.

Comparison

Four separate legal publications — among them "Federal Circuit Overrules Previous Obviousness Rosen-Durling Test for Design Patents" and the pointedly titled "Deere in the headlights" — reported the same core holding within a single news cycle: Rosen-Durling overruled, Graham factors installed. Roughly two years on, the comparison that actually matters to a searcher is not old doctrine versus new doctrine. It is two retrieval workflows with opposite failure modes. And the honest caveat up front: no published dataset yet prices the difference, so the side-by-side below runs on structural counts — gates, stages, factors — that you can verify against any docket, rather than invented dollar figures.

The old test decomposes cleanly, the way one patent-research thread compressed it: "rosen-durling: demolition basically compare." Stage one locked exactly one "basically the same" primary reference; stage two ran the comparison against secondary references. The Federal Circuit's en banc ruling deleted stage one's gate and expanded stage two into the four Graham factors imported from Graham v. John Deere: the scope and content of the prior art, the differences between it and the claimed design, the level of ordinary skill in the art, and secondary considerations. According to the commentary assembled in "Redesigning Design Patent Validity," the court gave two grounds at once — Rosen-Durling conflicted both with the Supreme Court's general articulation of obviousness principles and with its specific treatment of design-patent validity.

Neither workflow wins outright; the density of the art decides. Anchor-first still wins in crowded fields — automotive wheel designs, graphical user-interface icons — where a near-twin genuinely exists, because one dominant reference carries the entire obviousness story and hands a judge or jury a single villain instead of a committee. The Graham sweep wins when the design space is fragmented and no single reference comes close: there, obviousness lives in combinations, and only the four-factor frame lets you build — or demolish — a mosaic of references. It also wins whenever commercial success, industry praise, or copying might rescue a patent, because objective indicia have no home in a rigid two-step. The edge case is the hybrid: run the anchor pass as a tripwire, not a gate. A fast hit signals a dense field worth mapping narrowly; a miss escalates immediately to the full combination sweep. That preserves the diagnostic value of the old habit without letting it gate validity — retiring the anchor hunt as a quest, not deleting the anchor as a signal.

Kill the lazy takeaway while you are at it: the conventional anchor ritual did not waste money on unnecessary steps. It bought a narrower answer at lower retrieval cost, and its true price surfaced downstream as missed invalidity combinations and thinner clearance opinions. Treat the two regimes as different allocations of risk, not different price tags.

Concrete next move: pull your three most recent design-patent clearance memos and tag every cited reference as anchor or auxiliary. A memo built predominantly on anchors predates the new logic — re-run it through the four-factor frame before relying on it in current prosecution or litigation.

What to do next