Careful - searching based on patent classification codes can mean that you miss relevant patents

November 22 2017 Patent classification codes, such as CPC and IPC codes, are used by many patent searchers and search systems to find similar patents. There are are good reasons for this - they are widely used and can be useful in some cases.

And yet there are good reasons not to rely them on too much - they can be very unreliable, and will almost certainly have you missing key patents.

To show what I mean, I am going to run a case study. In a previous blog we talked about US patent application, 20170063194, filed by Apple for a Electromagnetic Levitator.


Apple patent drawing.gif

Ambercite Ai can quickly provide a similar patents, including a combination of known and unknown (predicted) citations. Running the query below will provide a list of 50 patents (or up to 401 similar patents if you wish):

Apple patent query.gif


But for now we might focus on the top 10 of these patents, shown in the list below.  8 out of these 10 patents are known citations, which two are 'unknown' citations, or predicted citations. All refer to magnetically levitated objects. 

Similar patents found, in ranked order

Similarity Rank

Known Citation

CPC Classes

US20170063194 Electromagnetic Levitator, (Priority year 2017)

Query patent

Query patent

F16C32/0444; F16C32/0446; F16C32/0453; F16C32/0457, H02K11/215; H02K7/09

US7348691B2 Magnetic levitation apparatus (2002)



A63H33/26; B60L13/04; F16C32/044; F21V21/096; F21V23/02; H02N15/00; H02N15/02

US4585282A Magnetic levitation system (1983)




US9352665B2 Magnetically lifted vehicles using hover engines (2013)



A63C17/00; B60L13/04; B60L2220/12; B60L2220/50; H02K5/04;H02K7/09H02N15/0

US5168183A Levitation system with permanent magnets and coils (1991)



F16C32/044; H02N15/00

US5332987A Large gap magnetic suspension system with superconducting coils (1992)




US8500509B2 Entertainment device including a remote controlled magnetic mini-craft (2007)




US8258663B2 Magnetic levitation novelty device (2009)




US9325220B2 Propulsion and control for a magnetically lifted vehicle (2013)



A63C17/00; A63C17/16; B60L13/04; H02K7/09;H02N15/00

US7110236B2 Magnetic suspension system (2002)




US9245679B1 Wine bottle floatation device (2014)



A47G2200/10; A47G23/0241; H02N15/00

In the column on the right, we show the listed CPC codes. The CPC codes listed for the query patent are highlighted in blue - and so we can see where they are appear in the most similar patents. Only two of the the most similar patents found share any of these CPC codes.

The most common CPC code in these results (marked in pink) is H02N15/00, which refers to Holding or levitation devices using magnetic attraction or repulsion, not otherwise provided for , which is clearly relevant - and it yet was a CPC code not found in the Apple patent application.



In this case, if a searcher was to rely on the listed CPC codes alone to find similar patents, they would miss similar patents, unless they ran a very broad search.

In this case, the examiner for the Apple 2017 patent application clearly ran a search that went beyond the limitations of these CPC codes, which is why they found these other patents - or at least 8 out of these 10 most similar patents.

Yet, there are patent landscaping systems out there that do purport to find similar patents based on patent classification codes. And there are plenty of patent searches run that use patent classification codes as key limitations.

All which makes us wonder - what relevant patents are being missed by relying on patent classification codes?


The alternative - ambercite ai

Exactly as shown before, Ambercite does not rely on any assumptions regarding patent classification codes when finding similar patents - and instead finds similar patents by applying its AI engine to its database of 155 million patent citation and over 50 million patents.  Contact us to arrange a demonstration and free trial of just how easy it can be to find prior art and other relevant patents using Ambercite Ai.

Mike Lloyd