Count Research Relative to the Crop

The question behind the count

Publication totals answer which crops receive the most papers. They do not answer whether that attention is large relative to how much of the crop the world grows.

The comparison needs a denominator.

Build a rate

Match each crop's AI plant-disease paper count to its 2023 global production, then calculate:

$$\text{papers per million tonnes} = \frac{\text{AI disease papers}}{\text{production tonnes}} \times 10^6$$

Portfolio allocation signal

Research intensity is concentrated away from staple volume

Tomato receives 11.5× more AI plant-disease attention per tonne than maize; sugarcane sits at the opposite end of the portfolio.

Coffee

39.8

Tomato

15.4

Apple

14.6

Grape

12.4

Potato

4.2

Rice

3.0

Maize

1.3

Sugarcane

0.2

Higher research intensitySelected staple crops

Evidence detail

Crop	Papers / Mt	AI papers	Production (Mt)
Coffee	39.8	441	11.1
Tomato	15.4	2,848	185.5
Apple	14.6	1,427	97.4
Orange	13.7	946	69.3
Grape	12.4	933	75.4
Potato	4.2	1,610	386.7
Banana	3.6	493	136.7
Rice	3.0	2,420	804.7
Wheat	1.7	1,315	794.6
Maize	1.3	1,658	1238.6
Soybean	1.3	465	370.9
Cassava	0.9	318	339.5
Barley	0.7	100	142.7
Sugarcane	0.2	443	2017.4

Check the denominator visually

A scatter plot keeps both original quantities visible. If research attention scaled with crop production, the points would rise from left to right.

Opportunity matrix

Agricultural scale and research attention are not moving together

The lower-right zone contains large production systems receiving comparatively limited AI disease research—a useful place to test portfolio priorities.

Sugarcane0.2 papers / Mt

2,017.4 Mt production · 443 AI papers

Cassava0.9 papers / Mt

339.5 Mt production · 318 AI papers

Maize1.3 papers / Mt

1,238.6 Mt production · 1,658 AI papers

Rice3.0 papers / Mt

804.7 Mt production · 2,420 AI papers

Evidence detail

Crop	Production (Mt)	AI papers	Papers / Mt
Coffee	11.1	441	39.8
Tomato	185.5	2,848	15.4
Apple	97.4	1,427	14.6
Orange	69.3	946	13.7
Grape	75.4	933	12.4
Potato	386.7	1,610	4.2
Banana	136.7	493	3.6
Rice	804.7	2,420	3.0
Wheat	794.6	1,315	1.7
Maize	1238.6	1,658	1.3
Soybean	370.9	465	1.3
Cassava	339.5	318	0.9
Barley	142.7	100	0.7
Sugarcane	2017.4	443	0.2

The matrix exposes a mismatch, but it does not explain it.

Test whether the field itself is growing

Raw paper counts rise as science publishes more. To separate general publishing growth from adoption of AI, calculate AI papers as a share of all plant-disease papers for each year.

Adoption signal

AI moved from a niche method to a material share of the field

AI appeared in 14.8% of plant-disease papers in 2025—nearly five times its 2005 share.

3.1%

2005

14.8%

2025

Most of the acceleration occurred after 2019.

Evidence detail

Year	AI share
2005	3.1%
2010	3.8%
2015	3.7%
2020	6.0%
2023	10.5%
2024	13.1%
2025	14.8%

Use this when

• groups have different population, area, production, or exposure;
• raw totals would reward size alone;
• the denominator has a defensible connection to the question.

Read the result carefully

Production tonnage is only one denominator. Crop value, disease burden, nutrition, research feasibility, and dataset availability can change what a fair comparison means. The normalized rate identifies an imbalance; it does not assign motive or prove neglect.