How and why the use of AI will better your data stories

By Leo Knight

AI even since the early introduction of automated systems was being used to trim down and make data sets more understandable and easier to interpret. It also allows you to focus more on finding a narrative within the data and what to use as the story. AI possess the ability to sift through huge datasets which in the case of sports datasets can be plentiful, there are a number of different ways and skills to utilise it in the right way though like using it in a way where it can also help you with idea and content generation, which this article will explore.

First of all you will need to find a data set, this can stem from a number of different ones; for example, shots on target, corners, win percentage away/home and cards given out by referees. The options are endless. A good website to use can be found at Football Results, Statistics & Soccer Betting Odds Data. The Dataset will normally be from Excel and will include a large set and numbers and information.

Screenshot from Excel, Dataset from 2023/2024 Premier League season.

Now that you have your dataset, make sure you know what the abbreviations and key is so you can paste that into openAI so they understand it fully and can make no mistakes. The notes/key will look like the picture below.

Screenshot of key for data set from football-data.co.uk/notes.txt

Once you have your data and the key you want to start a prompt which will look something like "I'm writing a data set which is from the premier league season of 2023/2024. i am pasting in the first week of results. find me five key features of the first week and stats that stand out in no less then 150 words." Along with this you will also want to paste in the key to avoid openAI confusing itself.

Screenshot from ChatGPT, Prompt for dataset article, taken 11/07/2025.

Now you have pasted in your data into openAI you will receive a response based on your prompt and go on from there identifying what story or idea you want to run with. there are a number of key things to consider however, typically with data stories you will want to identify what type of story it is; this can range from, ranking, relationship, scale and variation.

Screenshot from ChatGPT, Response to prompt about dataset, taken 11/072025.

From the response above, AI has provided several options for what narrative you could go with, for example you could go with the 2nd option provided. This could be a ranking story in how may goals were scored in the first week of this premier league season compared to others. You could also look at the average goals of the week and compare them to the next for weeks of that season and go in to depth about what that means for teams, whether they need to strengthen in the transfer market etc.

When writing a data story also you may receive data in a form or number that needs to more understandable for the audience, you may get mean and median metrics from some datasets while for others it will be in decimals which will need to be converted into percentages. For example in the fourth story offered, Wolves had 26 shots but only 6 on-target, this means that only 23.8% of shots were on-target. You could also change this by saying 76.92% of shots were off-target and develop that in your article.

Key features when writing up your data story must include a number of different contexts, facts and information. A good griping headline is key, it must draw in the audience to want to find out why or how your story has happened, for example 'Manchester United are the most disciplined team and its not going to get better.' This will draw in United fans to see what is going on and why it will not improve. It would also be hugely beneficial to get quotes for this story to deepen and back up the information and dataset, you could interview fans who run fan pages online to get their thoughts or even ex-players who could give valuable insight on how or why the stats look like they do.

overall, some limitations to the idea and content generation side could be seen through the biases of the AI, for example you could prompt AI not to focus on certain aspects that they might give bias on. Also AI might not have prior information to the league season you are inputting the data for, so they will only have the data set to work off. LLM (Large language models) are usually conditioned into giving only positive story and article ideas which is not something you want, you want shock, intrigue and something different which will engage the audience. Finally, always make sure to double check the data was input in the correct way as it is key to be transparent and careful when using AI.