GIPResearch.org • Open reference

CPC section G reference

Cooperative Patent Classification section G: official 2026.08 scope, technical class boundaries and reproducible evidence-search decisions. Follow the relevant hierarchy, source notes and edition-specific restrictions before interpreting any patent publication.

Save or follow this source

AI actions open a new tab. They do not guarantee that a service saves, cites or endorses this page.

Clarify the CPC section G research question

Working with measurement, optics, control, computing and information processing? Discuss which official sources and claim features should define the research scope.

Discuss this research question
Featured educational video

Featured patent and technology research video

Watch the featured explanation of patent and technology topics. The video is supplementary to the official sources and qualified limitations on this page.

Featured Work & Insights

Review CPC section G search boundaries

Need to distinguish relevant classes, exclusions and adjacent technologies? Discuss a traceable classification evidence map.

Discuss this research question

CPC section G: official definition and research use

CPC section G concerns physics. This reference specifically explains how the EPO/USPTO 2026.08 scheme supports source verification, warnings, class boundaries and cross-system retrieval for measurement, optics, control, computing and information processing. It does not assign classification codes to any particular invention. CPC classifications, especially finer subgroups and supplementary tags, must be interpreted under the actual scheme and its note system rather than inferred from the matching IPC letter. USPTO official CPC section G scheme (2026.08)

Evidence from the current CPC scheme

The current CPC scheme distinguishes G06F, electric digital data processing, from G06N, computing arrangements based on specific computational models. That distinction is especially important in AI patent analytics. An AI-labelled medical device patent might have a substantive measuring feature rather than a computational-model invention. A search defined as 'AI model inventions' should include a reasoned G06N criterion, whereas a search defined as 'clinical image-processing systems' may need additional measurement, optics and medical classes. USPTO official CPC section G scheme (2026.08)

Classes and technical functions to separate

A CPC research question about measurement, optics, control, computing and information processing should start with distinct functional categories. The candidate classes below provide a bounded orientation; the final set will depend on the technical claim features and any limiting references. The labels should not be read as a verified assignment of any product or document, and subgroup codes must be obtained from the official 2026.08 classification page.

Candidate classResearch orientationVerification condition
G01Measuring and testingCheck CPC definition, note, warnings and subgroup context
G02OpticsCheck CPC definition, note, warnings and subgroup context
G05Controlling and regulatingCheck CPC definition, note, warnings and subgroup context
G06Computing and calculatingCheck CPC definition, note, warnings and subgroup context
G09Education, display and signallingCheck CPC definition, note, warnings and subgroup context

Why CPC-specific warnings change retrieval

CPC G06F and G06N give an especially important official distinction: G06F covers electric digital data processing, while G06N addresses computing arrangements based on specific computational models. A paper titled 'AI in medicine' does not establish that G06N is the right primary place. CPC may also have detailed application-specific paths; a researcher should check the relevant subordinate definitions and use the current edition rather than assume all AI symbols sit under one group. USPTO official CPC section G scheme (2026.08)

A warning about an IPC subgroup absent from CPC is not an invitation to create an undocumented synonym. First identify the code on the source publication, then follow the official 2026.08 section page and dynamic CPC-to-IPC concordance. Store the mapping as a separate field, retaining any one-to-many relationships and the reasons for exclusion. A historical dataset should make both its unconverted and converted selection rules reproducible. CPC-to-IPC concordance and revision record.

Worked CPC research scenario

CPC's finer-grained physics and computing classes can improve retrieval precision, but they also create ambiguous intersections. A computer-vision claim may cover optics, signal pre-processing, model execution or feedback to a machine. A reproducible study should disclose whether the classifier is coding the algorithm alone or the larger apparatus. Use the official CPC notes and representative patent claims to determine whether joined classes represent one integrated invention or neighboring subject matter.

For this G-section scenario, maintain three queries: a precise CPC symbol leg, a functional-language leg describing the claimed technical mechanism, and a boundary-control leg involving the nearest competing class. The third leg helps expose otherwise overlooked disclosures. Examine representative documents retrieved only by one method and record the reason for their inclusion. A CPC label on a front page is not a substitute for reading claims and the specification.

Comparison with IPC: what a crosswalk can and cannot establish

The corresponding IPC section G offers an international baseline at the section/class level, but it must be researched as a separate official scheme. CPC may contain finer groups, specific warnings or indexing rules, so converting IPC to CPC by matching prefixes can misstate coverage. A defensible crosswalk should include source classification symbol, edition/date, proposed CPC destination, evidence of the mapping, and a note explaining whether the relationship is one-to-one or multiple. Never present a search-equivalence assumption as a verified class assignment. Official CPC concordance.

The most important functional comparisons here are Algorithm architecture versus measurement apparatus; optics versus image analysis; control method versus communications protocol. Each comparison should be checked against the actual wording of the relevant CPC entries, then tested on a small sample of patents with independent claims. Class placement describes the office classification of technological information; it does not by itself prove that all documents in the same class disclose the same claimed feature. Different applicants may also use different terminology for technically similar arrangements.

Version, revisions and historical corpus controls

The 2026.08 CPC edition entered into force on 1 August 2026. CPC revision notices, definitions and downloadable concordances are part of the source record and may affect the meaning of an apparent search pattern. Save the exact edition, retrieval date, classifications and the platform used to obtain patent records. When comparing cohorts from different years, distinguish movement caused by reclassification from changes in the underlying technology or patenting activity. This is especially necessary for long-running landscapes and historical invalidity research. Current CPC scheme and version history.

Independent CPC section G investigation: Optical sensor fusion with a trained defect-detection classifier

Consider an illustrative investigation into optical sensor fusion with a trained defect-detection classifier. The research is not a classification assignment to an identified patent. Its purpose is to separate measurement principle, optical formation, model inference and image analysis. The technical questions are different from the other CPC sections because the useful evidence, candidate classes, claim functions and likely false positives arise from this subject matter. The current official CPC scheme should control each symbol decision and any further subgroup narrowing.

Which alternative functions could change the search boundary?

A smart inspection system may be claimed as physical sensing, signal preparation, model learning or application-specific image processing. The presence of a trained model does not make every document a G06N invention; an optical sensor without a classifier can still be the most relevant prior art to claimed instrumentation. The task therefore begins with a written description of the feature the researcher intends to test. Only then should the investigator choose the relevant official CPC group definitions and cross-references.

Three questions to resolve before querying

  1. Which step produces the measured physical signal, and which step makes a prediction?
  2. Does the independent claim require optical geometry or just digital defect scoring?
  3. Are model training data, inference hardware and measurement configuration separate claim features?

Evidence design and reproducibility record

Use a split workflow: measurement modality, optical component, computer-vision processing, and model-estimation logic. Build a sample set of documents retrieved only under physical measurement and compare it with model-centered documents. Report overlaps without implying that a shared algorithm keyword proves a shared inventive concept. The researcher should identify families separately from publications, preserve a dated version of each query and store exclusions with reasons. The resulting conclusions remain limited to the actual corpus and claim elements reviewed.

For this example, the candidate classes to inspect—not presumed patent allocations—include G01N; G02B; G06N; G06V. The critical contrast is physical metrology versus optics versus computational classifier. A class symbol should be accepted only after opening its 2026.08 official scheme entry and applicable definitions. If a cited symbol was assigned under an earlier edition, the original code and any subsequent concordance or reclassification must be recorded separately. Avoid treating a shared high-level section as evidence of the same inventive contribution.

Official verification starting point: EPO/USPTO CPC scheme and definition table. Verify the precise subgroup and any limiting references in the source materials before publication of an invention-specific conclusion.

Audit checklist for the selected CPC section

A high-quality G-section CPC file should record the invention function, the relevant claim or abstract passage, the initial CPC classes and the explanation for any neighboring class considered. A reviewer should be able to see what was searched, why each code was used, what was excluded, and whether a symbol came from the official scheme or from a third-party database. Special notes and warnings at G06N or other relevant subclasses should be checked before narrower group use. USPTO official CPC section G scheme (2026.08)

  1. Retain the actual patent publication and original office-assigned classification.
  2. Inspect G06N and nearby CPC definitions, exclusions and references in the 2026.08 scheme.
  3. Record all additional-information codes and cross-sectional tags separately from the base technical codes.
  4. Build and execute independent technical-term and CPC-symbol search legs.
  5. Use the official CPC-to-IPC concordance for any proposed conversion; log ambiguous mappings.
  6. Read a source-document sample, preserve false positives, and state the completeness limitation.

Section-specific corpus design and exclusions

An AI patent corpus under CPC should not be defined by the presence of the letters 'AI' in a title. One practical research split is between model architecture, general computing system, signal measurement and applied control. For each candidate G06N or G06F hit, inspect whether the independent claim actually defines computation or instead specifies an apparatus using computation. Create a separate 'applied-domain' category rather than automatically excluding such results. This enables analyses of general model innovation versus application-specific deployment, while avoiding a claim that a class symbol proves model performance or legal patent eligibility.

What the evidence does not establish

Neither a subclass match nor a CPC tagging entry determines infringement, patent validity, inventive step or commercial value. The classification does not certify whether an algorithm works, a material achieves a laboratory result or an applicant has a granted enforceable claim. Scope and legal status need separate document-level and jurisdiction-level examination. If a class relationship is uncertain, retain it as a hypothesis, record the controlling source and do not invent a precise subgroup.

Official CPC references for independent verification

This is an informational classification method and a versioned source map. Before relying on any assignment, verify the exact published symbol and its source edition against the patent and the current official records. Use professional advice for patentability, clearance and legal decisions.

Discuss a versioned classification research packet

Planning prior-art or technology intelligence based on CPC section G? Contact Dr. Rahul Dev to define source versions, deliverables and limitations.

Discuss this research question
Content is general research information, not legal, patentability, freedom-to-operate, financial or other professional advice. Website use does not create an attorney-client relationship. Verify current rules with official authorities and qualified advisers. Full disclaimer.