Top 4 things to be careful of when using AI in journalism

By Leo Knight

As AI is starting to take the journalism world by storm, there are a few things you should be especially wary about when using AI, this article will prepare you and inform you about some of the dangers using AI so you do not make the same mistakes some others have already.

1. Bias- biases in AI are a huge danger and issue when using LLMs or other forms of AI.

   
   

   This is due to two different forms of it, built in AI biases are common throughout most LLMs with a built in system of information and historical data which it calls upon to provide responses but this has pre-built biases that reflects social inequalities. Therefore, their perception on some topics is incorrect, for example AI might see underrepresented groups like women, ethnicity and geographical location, so in datasets or overall and choose not to give you an insight on that as the information they possess not worthy or not of note. The other form of bias is the bias of the user, whatever you input to AI is you placing your subtle biases that you might not even notice but it is definitely there. An example of AI bias can be found below.

   

   Screenshot of ChatGPT, Taken 11/07/2025. Prompt inputting/asking for five footballers to interview that are local to Coventry.

   
   

   

   Screenshot from ChatGPT, Taken 11/07/2025, Response from the previous prompt.

   
   

   The biases within the screenshots above are apparent, there are no suggestions of female or representation of people who are multicultural in any way, all people given for interview ideas are male and white.

   
   

   How to solve the bias issue: Input into AI and forcibly say you want female, old and diversity included in the response, this eliminates the bias of the AI and your subtle bias.

   
   

2. Hallucinations- AI hallucinations are a key feature and flaw within AI which has caused a problem for many news companies across the world. This is where AI flaws itself by wanting to flatter the user, thereby, they come up with completely made up sources or claims. Examples of this can be seen at AI is polluting truth in journalism. Here's how to disrupt the misinformation feedback loop. - Bulletin of the Atomic Scientists. AI hallucinations stem from the LLM simply not knowing the answer for what you are trying to demand, so they lie or make something up as they morally do not know that its the wrong thing to do.

   
   

   How to solve the hallucinations issue: This can be a simple fix in theory, you can simply tell AI that if they do not know something they can simply say "I'm not sure." Also you can double check on the sources they provide to you to see if they are all real so you in affect verify them.

   
   

3. Transparency issues- Another main issue when using AI as a journalistic tool is transparency, as you are reporting and doing most of the processes as a journalist it is key that you remain honest and transparent to your audience. You are trusted to report the truth and not make up sources, quotes and events. This can also be seen as using AI but not declaring as such to your audience.

   
   

   How to solve transparency issue: By using prompts such as chain of thought so you fully understand AIs process of how they came up with the response they provided. Declaring AI use in bodies of work. Actively telling AI to be transparent with you as the user which will avoid made up sources and or outright lies.

   

   Picture of BBC Logo, From @BBC, Taken 13/05/2025.

4. Legal and ethical issues- With AI recently there has been a number of policies and rules introduced in several newsrooms which provide training and outright ban AI use, for example the BBC who ban AI use as it is not deemed 'transparent' to the audience. Also ethically using AI can bring its own issues like whether its an honest tool to use, also how it may effect your work; for example CNET in 2023 were found to be using AI to write financial articles which contained clear errors and plagiarism (CNET Is Quietly Publishing Entire Articles Generated By AI), AI can also fail with the privacy of certain situations which AI lacks the empathy to understand which can lead to legal issues.

   
   

   How to solve legal and ethical issue: You can read and become familiar with professional practises, sites like Ofcom’s strategic approach to AI - Ofcom and JournalismAI who promote safe and responsible use of AI and help to educate on the dangers of using AI.

   
   

   

   Picture of CNET logo, from @CNET: Product reviews, advice, how-tos and the latest news Taken 12/05/2025.