Artificial intelligence and data scraping

At this point, we’ve all heard about how artificial intelligence (AI) is changing our lives (for better or worse!). It is a very powerful tool that is being used and marketed to be a real gamechanger for everything from learning complex topics like math and science, helping with language and grammar, or creating meal plans and recipes.  

Where does it come from? Big tech companies like Google, X, Microsoft and OpenAI now offer their own platforms featuring large language models, or LLMs. The purpose of an LLM is for the user to simply type and ask a question, then AI goes out and finds the answer by digging through its database to find the relevant information. This is effective because the databases were trained on an extensive range of sources, enabling them to answer a wide variety of requests and generate relevant text or images by connecting the dots. Twenty years ago, we said “Google it” to find quick and easy answers—the Oxford dictionary made “google” officially a verb in 2006. Fast forward to 2026 and we now have AI. This decades old concept is finally becoming easy to use every day and it’s more readily available on our phones, computers and electronics.

AI makes it simple, but remember, agriculture is a complex system

Farming is one of the oldest professions. Civilization advanced to today’s level partly because we domesticated agriculture so well that we were able to feed the masses by producing food from very few farms and allowing humanity to focus on other important tasks. Agriculture remains a complex system because biological diversity and environmental variability are constantly interacting with each other, which makes it difficult to predict their behavior.

It’s only natural to look for simple connections, but remember, just because two things happen at the same time doesn’t mean one made the other happen. In agriculture, for example, we know that feeding soils with the right amount and type of nutrients (like nitrogen, phosphorus and potassium) will lead to strong yields. But if the weather doesn’t cooperate by providing timely rains, sunshine and adequate temperatures, even a field with bountiful levels of nutrients will produce a very disappointing yield at harvest.   

What we do know and what we don’t know

I recently asked a series of questions to Open AI’s ChatGPT about helping me write a seeding rate prescription for a 2026 corn field. It’s important to remember that AI has a memory of who we are given any prior interaction. These conversations don’t come from a “clean slate” since the chatbot is familiar with my work in precision agriculture.

Here’s my prompt: Can I create a custom seeding rate prescription in a corn field for 2026?

At first glance, it appears ChatGPT did a good job highlighting the necessary steps to successfully design and execute a seeding rate prescription for this year’s corn crop. Research suggests that most farmers develop seeding rates from their own experiences, but will also include recommendations from private crop consultants, university and Extension services.

It’s important to recognize that many of these steps require additional critical thinking, which depends entirely on how each farm manages their crop, equipment, location and their future crop rotations/plans. Resist the urge to feed LLMs your own farm data without having specific objectives identified.

The following are some example objectives, prompts and results from Microsoft Copilot.

Objective:

Provide a table of average April precipitation on a field in the past 4 years.

Prompt:

Can you provide me with a table of historical rainfall at my field from the past 4 years? Here are the coordinates: 42.773º N, 84.458º W.

CoPilot results:

*[extremeweatherwatch.com]

Notes on interpretation:

• Data are derived from NOAA station–based records summarized for Lansing, MI, which is the closest long‑term station to your field.
• Differences between this station and a specific field are usually small for monthly totals, but localized storm effects can still matter for field‑scale decisions.
• If you want a gridded product instead (e.g., PRISM 4‑km estimates centered exactly on your coordinates, or daily totals rolled up to April), I can generate that as well and compare the two.

Objective:

Provide a table of weekly urea price trends in 2026.

Prompt:

Can you provide me with a figure or table of weekly urea prices per ton?

CoPilot results:

*Late‑March values widened substantially depending on region and timing of quotes as global supply disruptions intensified.

Sources: DTN Retail Fertilizer Trends articles for February and March 2026; USDA‑AMS Illinois Production Cost Report corroboration. [weather.gov], [weather.gov], [weatherspark.com], [forecast.weather.gov]

Objective:

Determine amount of P to apply in spring for corn.

Prompt:

Based on Tri-State Fertilizer Recommendations, can you provide me with a baseline amount of phosphorus I should apply at planting to my corn crop?

CoPilot results:

I can give you a Tri‑State–consistent baseline, but I need to be explicit about the assumptions, because Tri‑State phosphorus (P) recommendations are soil‑test driven, not blanket rates.

Below is a practical baseline you can use at planting, followed by how to adjust it once soil test data are available.

*[www.canr.msu.edu/soilfertility/Files/Main-page/FINAL%20PRINT.pdf]

These examples show that Copilot did great in summarizing data from my prompts and showing me sources. It found rainfall totals from the closest weather station to my field. It accurately summarized the recent price hike in urea fertilizer. In the prompt regarding P application in spring corn, it reminded me it cannot blindly assume a rate without knowing prior information about my field that would normally come from a soil test.

Another important consideration when using AI for prescription services is that neither AI nor humans can predict the weather. Effective fertilizer rate decisions, especially for nitrogen in grain crops, requires an understanding of past conditions, current weather and plausible future weather scenarios.

Click here for more Industry News.