How to use generative AIs for repetitive journalistic work

In the rapid world of journalism, much of our time is consumed by repetitive tasks that can often go unnoticed. From sorting through endless emails to transcribing interviews, these tasks can be at least partially delegated to an AI assistant, allowing us to focus on what truly matters – building relationships, verifying information, and crafting compelling stories.

By embracing generative AIs with a method and a perspective in mind, journalists can free up valuable time and dedicate themselves to the core aspects of their profession. This also means that it’s not a good idea to use AI tools to overproduce content. Generative AIs should support the idea of enhancing quality over quantity in journalism, making a strong case for the principles of slow journalism.

Let’s explore why this is beneficial and what tasks we can partially delegate to these technologies, and understand the associated benefits and risks.

Repetitive tasks on the digital ecosystem we can at least partially delegate

Transcribing interviews is still one of the most requested activities that journalists want to delegate to AIs. Generative AI can automatically transcribe audio interviews, converting them into text with high accuracy. Tools like Otter.ai, Happy Scribe, and Descript use AI to provide quick and reliable transcriptions, saving hours of manual effort.

It’s important to remember that if you want to rely on these online tools, you need to be aware of the security implications, especially when transcribing sensitive documents or confidential sources. Devices like the Pixel Pro 8 now include this function natively and can work offline. However, always remember that the only truly secure device is one disconnected from the internet. Even then, if someone finds it and accesses the password, it can be compromised, of course!

• Sorting and categorising emails

Mail management tools powered by AI, such as SaneBox or Google’s AI features in Gmail, can help sort, categorise, and prioritise emails. These tools can identify important emails, filter out spam, and organise messages into relevant folders.

If you need to manage ordinary mail with no confidential information, an AI chatbot like ChatGPT, Claude, or Gemini could be useful for drafting emails with just a few instructions you can even dictate (for example, using ChatGPT’s app). Plus, they can help you make sense of those long, messy emails we all dread (let’s be honest, not everyone writes clear, concise emails, and we are all drowning in our inboxes).

This ensures that journalists spend less time managing their inboxes and more time responding to crucial communications and leads. Just be sure to check the terms of services of these tools to understand what happens to your emails, data, and so on.

• Content curation and research

AI-powered platforms like Feedly and Curata can automate content curation by aggregating news from various sources, filtering relevant articles, and suggesting topics based on predefined criteria. Tools like Perplexity can help you find stories, quote sources, and dive deeper into topics. Personally, I find it handy for checking if someone has already covered a story I’m working on and seeing their approach.

However, be aware that these tools can sometimes go off the rails with summarisation, so you still need to verify everything. If you prefer to use conversational chatbots for this task, make sure to force them to quote and link sources.

One small hitch: some newspapers block these chatbots with the robots.txt instruction (most notably, The New York Times), which means they are excluded from the sources these chatbots can theoretically consult.

• Generating summaries and reports

Tools like Wordtune and SummarizeBot, along with the well-known LLMs (Gemini, ChatGPT, Claude, etc.), can condense lengthy articles, reports, or transcripts into concise summaries. These summaries highlight key points and provide a quick overview of the content. This is particularly useful for journalists who need to quickly understand and report on extensive documents or research papers. However, remember to double-check the summaries.

A good trick when using chatbots is to ask them to summarise by splitting the content into parts: these tools often suffer from the “lost in the middle” problem. They’re great for a quick overview, but if you need specific sections summarised from a PDF, make sure to give a proper prompt (for example, “explore, analyse, and summarise in ten bullet points this PDF from page 35 to page 47”).

Please note that, according to a leak and the ordinary experience of using some tools like ChatGPT, there are high-level limitations. ChatGPT cannot provide summaries of more than 90 words in its prosumer version (even for paying users), to avoid regurgitating training material.

• Automated fact-checking

Fact-checking platforms such as Full Fact and Factmata utilise AI to verify the accuracy of information by cross-referencing it with trusted sources. These tools can automatically flag potential misinformation and suggest corrections. It’s very important to remember that the final word is always human: machine tools can lead to false negatives or false positives, and avoiding both is crucial in journalism.

When it comes to spotting and detecting fake images, videos, or audio content, the rise of deepfakes makes this increasingly challenging. There’s one golden rule to keep in mind: we do not publish a photo, video, or audio if it hasn’t been double-checked. This means thorough research and verification — sometimes it’s quick and easy, and sometimes it’s not. If we can’t be sure that a picture is real, even if it’s 99% likely, we should not publish it.

Several tools can help with this process. For example, the INVID Project remains useful even in the age of AI-generated content. Its forensic analysis can reveal parts of pictures or frames with a high probability of manipulation, though it can also produce false positives or negatives. So, always double-check and verify before publishing.

• SEO optimisation

Who wants to compile the ALT description of an image? No one (or the web would be full of well-compiled ALT descriptions, which it clearly isn’t). So, why not delegate these tasks to AIs?

AI-driven SEO tools like Clearscope and MarketMuse can analyse content for SEO performance, suggesting keywords, meta descriptions, and structural improvements to enhance visibility and engagement. But you can also use the “ordinary” chatbots. For example, you can ask Gemini to extract keywords, long-tail keywords, and search intent for a topic. You can ask the AI to create an SEO-friendly editorial plan, and you can ask ChatGPT to create the ALT description for a picture. Just remember, it’s always a good idea to provide a bit of context.

Prompt and output

You can also work with developers to simplify these tasks, but we should expect that they will be integrated into CMS very soon.

• Social media management

Platforms like Hootsuite and Buffer integrate AI to schedule posts, analyse engagement metrics, and suggest optimal posting times. They can also generate content ideas and automate responses to common queries. Conversational LLMs can be used to create different versions of the same post.

However, be aware that if you don’t provide clear examples of your writing style, tone of voice, and so on, the social posts could end up disappointing – full of clichés and overly emphatic. AIs need statistically relevant samples and references to help you create proper social-content drafts. And yes, you’ll still need to refine them.

Human journalism is alive

By delegating repetitive tasks to generative AI, journalists can reclaim valuable time and focus on the aspects of their job that require human intuition and creativity. Embracing these technologies not only enhances efficiency but also elevates the quality of journalistic work by allowing more time for in-depth reporting, thorough verification, and meaningful engagement with sources and audiences. As we continue to navigate the digital ecosystem, leveraging AI responsibly and ethically will be crucial in shaping the future of journalism.

AIs won’t replace us as journalists; instead, they can be used to enhance journalistic work and bring it back to its essence.

That said, I’m not naive—I know many publishers might see AIs as a way to overproduce content, chasing the myth of infinite growth and clicks. But I believe we have a duty to propose an ethical and useful way to take advantage of these tools to produce good, accurate, and human journalism.

Source of the cover photo: BoliviaInteligente via https://unsplash.com/

---