Case Study

AI tools powering innovation in Technical.ly’s newsroom

By Hayley Slusser

November 26, 2024

Artificial intelligence is here to stay.  

Though the emerging technology may seem new, it is already becoming a permanent part of the way many news organizations work.  

While it is normal to be skeptical of new technology — we’re journalists, after all — Technical.ly Editorial Director Danya Henninger highlighted several ways newsrooms can leverage AI to improve their reporting, streamline simple tasks, or expand the reach of their journalism.  

“Journalists have been creatively applying technology tools to the trade since before the printing press, from the telegraph to broadcasting equipment to pocket-sized recording devices to spellcheck. Technology has leveled the playing field and increased access to the profession,” Henninger said. “If you’re in need help with grammar, or learning how to summarize, or simplify, or pull out a good quote — that doesn’t have to hold you back.” 

As the name suggests, Technical.ly covers technology, but the following strategies will work for any newsroom, regardless of their size or coverage area. 

Creating an AI policy 

No matter how extensively your organization plans to use AI tools in its work, it is crucial to have internal and external policies in place to ensure you uphold your organization’s ethical and journalistic standards.  

Publishers, including Technical.ly, have created internal policies that encourage experimentation while ensuring that staff are transparent with both the audience and their colleagues about how and when they are using AI tools. Henninger said to “think of AI as a team of interns” — it can get the work started for you, but a human should always review and fact-check AI-generated content before it is made public.  

For audience-facing use cases, you’ll need to determine when and how you plan on disclosing the use of AI-generated content. You probably don’t need to tell readers that you used an AI-informed tool to spell check your story, but other, more substantial uses of generative AI should include a public label.  Below are examples of Technical.ly’s AI disclosures for translations (which are human-reviewed) and for photos.

 

If you don’t know where to get started when building your AI policy, Poynter created a free template for newsrooms with sections on audience-facing content, business uses, and reporting assistance. Henninger published Technical.ly’s public AI policy in a blog post.  

Everyday tools 

Chances are, you’re probably already using AI tools in your day-to-day work. Popular tools in the journalism space include Grammarly, which checks your writing for spelling, grammar, and clarity issues; and Otter, a recording and transcription service. Here at the Institute, we’ve used DeepL for translation services, which were ultimately reviewed by a human editor, of course. 

Henninger also recommended tools including  Alice, another transcription tool built by journalists that offers more privacy protections, and Alt Text AI, which adds alt text to photos and offers integrations with platforms including WordPress, Contentful, Chrome, and Dropbox. (Adding alt text not only makes your content more accessible, but it helps with SEO, too!) 

SEO headline optimization 

One of the top ways Technical.ly uses generative AI is to write headlines for search engine optimization, helping to boost its articles in search engines — meaning Google — and drive more readers to its website. 

If you’ve tried using generative AI, you know that developing precise prompts is an art in and of itself. When using AI tools to help with SEO, some key parameters Henninger said you may want to include in your prompts to ensure useful outputs are:  

• The character count for your headline (65 or less) 

• First three words carry the most weight 

• Generate 10 options and rank them on which would yield the highest number of SEO results 

Here’s an example of a prompt used at Technical.ly: 

You are an editor at Technical.ly, a local news site that covers technology and startups. 

You’re going to suggest SEO headlines for the following article. The goal is for the article to show up in search results when people interested in the topic search the web. You want the headlines to be user-friendly, but know they also have to be shorter than 65 characters and that the first 3 words count the most toward search engine optimization. Use sentence case for the titles. 

Come up with 10 options for the following article, and please estimate which would return the most SEO results. 

[paste article text]  

Henninger’s key tip? Try and try again. Feed the first suggestions back into the AI tool and ask for changes as you see fit. If your story is based on a proper noun, ask for headlines with and without the noun and whether including the noun would make a significant difference in SEO results. If the tool continues to produce similar, mediocre results, closing out of the window and starting a whole new chat may help.  

If you plan on using AI for SEO regularly, you can also create your own custom GPT that has your newsrooms preferred parameters pre-saved.  

Summarizing interviews and stories with key points 

Whether for internal purposes or for public-facing content, there will be times where you need to boil down an interview or story to a few key points. AI tools can be a time-saving solution. 

For interviews, many transcription platforms like Otter or Alice will automatically provide a summary of key points. Henninger said these summaries can be used to identify which parts of the interview you want to return to for additional details or can shape your questions for a follow-up interview.  

For news articles, you’ve likely noticed that TL;DR bullet points at the top of articles are becoming increasingly common, as they can help readers better understand longer news articles. Just like how Technical.ly generates SEO headlines, you can craft a prompt for an AI tool to yield a short, bulleted summary of an article. It may not be as precise as you’d like, but you can use the output as a starting point and refine it manually.  

Since generative AI trains itself on the content users put into it, you might be skeptical to copy and paste your journalism into tools like ChatGPT. Henninger said reporters can turn off chat training if they choose, but since Techinical.ly is paywall-free, it is less concerned about its content being used for AI learning.  

“As more and more online search queries are answered using generative AI, we should want our reporting to help make the results more accurate,” she wrote in Technical.ly’s public AI policy.  

Analyze and display data 

AI can analyze large data sets and identify patterns in a fraction of the time it would take a human. One example is how Technical.ly runs workforce data through ChatGPT and asks it to identify interesting or unusual patterns. This strategy resulted in stories on Arlington, Virginia’s rising tech industry or where tech workers in the Philadelphia region live. 

AI tools can also generate data visualizations, allowing you to test out different kinds of charts to determine which format would be most useful. You can either use the AI-generated visualizations directly in your publishing (provided they’re human reviewed) or recreate them using more advanced data visualization tools. 

Whether you’re analyzing data, writing headlines, or doing any other kind of AI-related work, Henninger said the fundamentals of journalism remain the same.  

“It’s the integrity of the information and the sources and the reporter and editor and publication that matter, not the tools used to put that info into digestible, understandable form. For journalists, and really for all people, reputation is the most valuable thing we have,” she said. “Disseminating factual, useful, timely information should be the goal, within the larger mission of helping communities find their truth.” 

The headline of this story was written by ChatGPT.