Pok&eacute;mon Trading Cards reveal visual stereotypes of natural minerals

Chen, Cheng Hung

doi:10.5194/egusphere-2026-2574

Preprints

https://doi.org/10.5194/egusphere-2026-2574

Preprints

20 May 2026

| 20 May 2026

Pokémon Trading Cards reveal visual stereotypes of natural minerals

Cheng Hung Chen

Abstract. Popular media often simplifies Earth materials into idealized forms, shaping public understanding of geology. To investigate this effect, this study analyzed 202 mineral illustrations from the Pokémon Trading Card Game between 1999 and 2026. By quantifying visual characteristics through Multiple Correspondence Analysis, a dominant "universal crystal template" was identified, defined by large, transparent, high-symmetry prismatic crystals in cave environments. This representation contrasts with natural systems, where minerals commonly occur as irregular, rock-forming aggregates. Comparisons between human interpretation and AI-based identification further show that these stylized images lack sufficient diagnostic information for consistent classification. While human interpretation incorporates contextual cues, AI models rely on simplified geometric features and exhibit strong anchoring bias. These results demonstrate that widely circulated media not only simplifies geological reality but systematically reinforces visual stereotypes of minerals.

Received: 04 May 2026 – Discussion started: 20 May 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 2176 KB)

Supplement (493 KB)

Download & links

Cheng Hung Chen

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-2574', Edward McGowan, 07 Jun 2026
General comments
I strongly believe this is a great manuscript that draws attention to popular types of media that non-geologists can easily encounter in the public domain. It is therefore important to understand how geology is represented in these media. Video games have been covered in several papers in the past few years, but this manuscript makes the reader consider other media forms, especially those other than film/TV.

The manuscript was well written, and the author, who is clearly a PTCG collector, considered many of the key aspects about the PTCG during their research, such as focusing only on English cards to avoid missing other language-exclusive artworks, and to make the dataset more manageable. They clearly put a lot of effort into analysing the data and presented it clearly.

I do, however, have some concerns about the content of the manuscript, such as whether it is plausible enough to consider that PTCG collectors would use AI to identify minerals in the artwork, especially when it is so rare for a card to include a mineral. Perhaps the AI aspect would be more appropriate as its own manuscript, investigating how accurately they perform against popular media forms (e.g. stills from movies, or in comic books as well). Furthermore, the final summary points and conclusion feel rushed and leave the reader lacking in their understanding of why this research is truly important.

Specific comments
Does the author truly believe that there is a significant number of PTCG collectors who would use AI to identify minerals within the artwork? As a geologist and PTCG collector myself, I have never once considered the accuracy of the illustrations on the cards, nor thought to use AI to identify them. In my experience of over 20 years in the hobby, collectors are more likely to appreciate nice artwork and not question the realism of it. This can potentially even go as far as most collectors recognising it is depicting a fantasy world and therefore it is acceptable for the artwork to not be accurate.

If Pokémon cards that contain minerals only account for 1% of the entire English printed roster, how many ‘mineral cards’ would the average collector actually have/see? Taking a reasonable number of 2000 in a collection, that would equate to someone only having 20 ‘mineral cards’. This seems very low and might not have a significant impact on their understanding of true mineral structures. Furthermore, is it not possible collectors are able to understand that the artwork contains fantasy elements given they depict fantasy monsters.

Ice was identified by a human expert (the author) as the most common mineral (Figure 4). However, it is barely spoken about within the text. Instead, the text focuses mainly on the other five identifiable minerals (e.g. quartz, calcite, and zircon). What is ice’s significance to the results in terms of accuracy of illustrations, reasonable environments etc? Either explain the results regarding ice or remove it from the criteria/data.

The implications of this research (Sections 4.3 and 5) feel very rushed. The results strongly support the author’s conclusion that mineral depiction within PTCG is oversimplified. However, the reasoning provided for this is lacking, only stated as influenced by gemstone markets and museum displays. The author does not appear to consider the fact that many PTCG illustrators likely lack a geological background and that some art styles may purposely choose not to take inspiration from realism or detailed accuracy (e.g. Shinji Kanda or Hyogonosuke). The author does allude to the latter in Section 4.1 but not considered again. After 30 years of producing Pokémon cards, the hobby is arguably more popular than ever, and a lack of accurately illustrated minerals does not appear to dissuade collectors. Therefore, is the Pokémon Company not justified in its approach to continue to “reproduce this selective imagery”? It would be great to expand further on these considerations.

Continuing the last point, the author simply says “geoscience educators might consider addressing these media representations”, which I agree with. However, it would be great to also expand on this further and provide examples. This would improve the significance of the manuscript by providing educators with potential starting points that they could adapt into their lessons.

Technical comments
Section 2.1, Line 39: How did you reach a total of 19,875 cards? I cannot find an online reference to this total anywhere, even on Pikawiz. Also, it is worth noting that some artworks are reprinted, not just for mini-sets but in full sets (e.g. “Professor Elm’s Training Method” has the same artwork used in E-series base, Unseen Forces, and Dragon Frontier). Therefore, this total would not represent unique artworks.

Section 2.1, Line 40: Did you include Promo cards or other mini-sets such as the McDonalds or Trick & Trade sets as well? Most of the latter reprint existing artwork and count it as a new card by including a new holo pattern or stamp.

Section 2.1, Line 40: Was the Pikawiz database cross-referenced with any other database to confirm all cards were accounted for?

Section 2.1, Line 43: How confident are you that the coverage and list you came up with was “exhaustive”? Did you manually screen the cards once or more times? A quick search on Google and Bulbapedia shows there are several cards from the HGSS series that include evolution stones (e.g. HGSS base Arcanine #1), and there are at least three different artworks for “Old Amber”. From my interpretation of your methodology, these cards agree with your criteria.

Section 2.1, Line 45: “distinct geological formations” is a very broad term that could capture many different elements that do not appear to be included, such as volcanoes. Also, how can you accurately identify a “limestone cave”? You later state on Line 98 that it is not possible to identify the lithology of host rocks. The definitions you describe in the Appendix (Line 285: Major rock types) are much better. I would recommend using them in the main text body as well as the Appendix.

Section 2.1, Line 48: “202 cards”, again I believe there are more.

Section 2.1, Line 48: “84 different artists”. I would be interested to know later in the manuscript if there are particular artists who like to include minerals in their artwork, or if there is a general average of 2-3 cards per artist.

Section 2.3, Line 85: How does ChatGPT and Gemini fair against identifying real minerals?

Section 2.3, Line 86: Why was Rock Identifier used as the specialised app? Did it have the most positive reviews?

Section 3.1, Line 96: Provide an exact number/ percentage of how many minerals occurred in caves and that were fully exposed.

Section 3.1, Line 101: It is not explained or alluded to until the Appendix how the length of minerals are calculated. I would recommend including a short explanation earlier in the text, such as in Section 2.1, so that readers are aware.

Section 3.2, Line 135: Provide an exact percentage for the rock-scanning app to highlight how much less this agreed compared to the other AI apps.

Section 3.2, Line 139: Provide one or two examples of “fantasy elements” identified in the cards.

Figure 3 caption: Reference back to Section or Appendix as to what the “13 categorical features” are so that readers can quickly look at the figures and know where to go to in the text.

Figure 3 caption: Include a mention about the Benzécri correction in the main body of the text.

Section 4.1, Line 167: Has the lack of paragenesis and focus on rare crystals been reported in other forms of media? If so, it would be good to cite that here.

Section 4.1, Line 179: Artists may also be more familiar with particular mineral types or lack true awareness of other forms. I highly suspect many illustrators only have a limited background in geology, if any background at all.

Section 4.1, Line 182: Again, it may not be an “artistic shortcut” if the illustrators are not fully aware of other crystal types. If their knowledge of minerals comes from limited geological media such as museums (as mentioned in the text), then they are only replicating what they have seen themselves.

Section 4.1, Line 190: What do the statistics show when considering ice only?

Section 4.1, Line 192: Again, perhaps the illustrators are unaware of other minerals such as amphibole or mica.

Section 4.2, Line 213: “Failure” is a strong word for saying this manuscript is investigating the minerals of a ‘fantasy world’. Numerous aspects of the Pokémon world are impossible in the real world. Why can minerals not be included in this?

Section 4.2, Line 216: “A human could recognize”. A trained geologist such as the author is capable of easily doing this, but could a non-geologist?

Section 4.3, Line 235: Hyper-realistic digital illustrators such as 5ban Graphics and REND are relatively new to illustrating cards. Although their work is hyper-realistic, they have yet to illustrate cards with minerals. Perhaps with time, new cards will appear.

Section 4.3, Line 240: Explain further why this is beneficial and how educators could use the PTCG as effective communication tools.

Section 5, Line 245: What about ice? This was the most common mineral identified.

Section 5, Line 247: It has been previously mentioned AI also fall short in their ability to correctly identify minerals correctly for several other reasons such as not being able to consider surrounding environmental aspects or struggling with 2D image identification.

Figure A1, Graph: The Scarlet & Violet series ended in September 2025 and was succeeded by Mega Evolutions, which is likely to continue well into 2027.

Appendix B1, Line 293: “Official average heights” were used. Individual Pokémon, even from the same species, can drastically range in size. Also, where were the measurements acquired? Were they written on the cards or from Bulbapedia, for example?
Citation: https://doi.org/10.5194/egusphere-2026-2574-RC1
- AC2: 'Reply on RC1', Cheng Hung Chen, 29 Jul 2026
  
  For the reviewer's convenience, a marked PDF of the revised manuscript is attached.
  
  Citation: https://doi.org/10.5194/egusphere-2026-2574-AC2
RC2:
'Comment on egusphere-2026-2574', Anonymous Referee #2, 20 Jul 2026

general comments
The manuscript analyzed 202 mineral illustrations from the Pokémon Trading Card Game between 1999 and 2026. Multiple Correspondence Analysis was applied to identify the visual characteristics and found the contrasts with natural systems. Comparisons between human interpretation and AI-based identification further showed that these stylized images lack sufficient diagnostic information for consistent classification.
specific comments
This article calls for the importance of popular science education on minerals and conducted in-depth quantitative analysis of the minerals in Pokémon Trading Card Game, which are good. It would be even better if further suggestions could be given on how to enhance professionalism.
technical corrections
Because the mineral images in the Pokémon Trading Card Game are more artistic than practical, is it fair to use them test AI tools?

Citation: https://doi.org/10.5194/egusphere-2026-2574-RC2
- AC1: 'Reply on RC2', Cheng Hung Chen, 29 Jul 2026
  
  general comments
  The manuscript analyzed 202 mineral illustrations from the Pokémon Trading Card Game between 1999 and 2026. Multiple Correspondence Analysis was applied to identify the visual characteristics and found the contrasts with natural systems. Comparisons between human interpretation and AI-based identification further showed that these stylized images lack sufficient diagnostic information for consistent classification.
  Response
  The author sincerely thanks the reviewer for the positive evaluation of this study. The encouraging comments regarding its quantitative approach and contribution to mineral science communication are greatly appreciated.
  
  specific comments
  This article calls for the importance of popular science education on minerals and conducted in-depth quantitative analysis of the minerals in Pokémon Trading Card Game, which are good. It would be even better if further suggestions could be given on how to enhance professionalism.
  Response
  The author thanks the reviewer for this valuable suggestion. To address this point, Section 4.4 has been substantially expanded to provide more concrete recommendations for enhancing the scientific value of future PTCG illustrations while preserving artistic freedom. The revised discussion suggests that illustrators could draw inspiration from museum collections, mineral markets, and geological references, while collaboration with mineralogists and economic geologists could introduce scientifically recognizable mineral properties, crystal habits, and geological settings. It also provides practical examples of how these illustrations can be incorporated into classrooms, museum exhibitions, and science outreach activities.
  
  technical corrections
  Because the mineral images in the Pokémon Trading Card Game are more artistic than practical, is it fair to use them test AI tools?
  Response
  The author thanks the reviewer for this important comment and agrees that Pokémon card illustrations are artistic creations rather than scientifically realistic depictions. Accordingly, the purpose of introducing AI was not to evaluate the artistic quality of the illustrations, but to investigate how current AI systems interpret the same stylized images that educators or learners may use in future science communication activities.
  To clarify this rationale, a baseline experiment using photographs of real mineral specimens has been added. The results show that current AI systems identify real minerals accurately but perform much less consistently on stylized illustrations, indicating that the observed discrepancies primarily reflect the limited diagnostic geological information in the artwork rather than a general inability of AI to recognize minerals.
  The Discussion has also been revised to emphasize that AI serves as a comparative analytical tool for understanding how media-driven visual simplification influences both human and machine interpretation, while providing guidance for future educational applications.
  
  Citation: https://doi.org/10.5194/egusphere-2026-2574-AC1

Cheng Hung Chen

Supplement

https://doi.org/10.5194/egusphere-2026-2574-supplement

Cheng Hung Chen

Viewed

Total article views: 379 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
203	159	17	379	90	12	11

HTML: 203
PDF: 159
XML: 17
Total: 379
Supplement: 90
BibTeX: 12
EndNote: 11

Views and downloads (calculated since 20 May 2026)

Month	HTML	PDF	XML	Total
May 2026	101	66	10	177
Jun 2026	45	25	2	72
Jul 2026	50	55	4	109
Aug 2026	7	13	1	21

Cumulative views and downloads (calculated since 20 May 2026)

Month	HTML	PDF	XML	Total
May 2026	101	66	10	177
Jun 2026	45	25	2	72
Jul 2026	50	55	4	109
Aug 2026	7	13	1	21

Viewed (geographical distribution)

Total article views: 349 (including HTML, PDF, and XML) Thereof 349 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 22 Aug 2026

Short summary

To explore how media shapes geological perception, this study analyzed mineral art in the Pokémon Trading Card Game using human and artificial intelligence evaluations. Media heavily relies on a simplified template, predominantly depicting minerals as giant, flawless, transparent prisms isolated within caves. Consequently, artificial intelligence fails to identify these stylized images. This visual bias ultimately distorts public understanding of real Earth materials.


Total:	0
HTML:	0
PDF:	0
XML:	0