Artificial Intelligence is proving to be a Rorschach test for publishers all around the world. Some of them look at the inkblot and see an existential threat or a parasite that has latched onto journalism, while others see incredible possibilities, with generative AI products used to automate onerous, even impossible tasks, or to create hyper-personalized content in Big scale.
There are certainly real elements in each perspective, but the reality of generative AI is neither subjective nor open to interpretation. O ChatGPTGive her, Bard and its counterparts are not just another shiny, shiny object easily discarded and ignored. They are powerful tools. Generative AI has the potential to be transformative for many industries, changing the way we work in the classroom, in the courtroom and, yes, even in news writing. In a recently released survey by Adobe, 92% of respondents said AI is having a positive impact on their work and more than a quarter (26%) consider AI a miracle.
With the right instructions and guidelines, generative AI can excel at tasks such as summarization, content optimization, and transformation between content types. Because generative AI understands the syntax of the language so well, the results are surprisingly good at turning a long news article into a scripted text. podcast or in a sequence of tweets. And the promise of conversational search and data analytics prompts exciting.
However, it is not designed for all tasks. As is well documented, generative AI can be unreliable with facts and quotes, prone to hallucinations. This is one of the main reasons why publishers They have such different stances and why it is essential to have a methodological approach to adoption.
Establishing rules of engagement
First and foremost, it’s important to start by articulating your company’s initial position on the use of AI-generated text, images, audio, and video. Just like you have a social media manual detailing what’s allowed and what’s not, generative AI requires the same care. What do you use generative AI for? Where will it never be used? Be specific, both for internal and external audiences. For example, perhaps you allow AI-generated images, but only in technology or medical articles to illustrate complex, theoretical concepts. Technology is evolving rapidly, so be prepared to reevaluate your position and guidelines periodically.
For example, Insider published a letter from the editor detailing its position on AI, taking an approach to experimentation and innovation.
I’ve spent many hours working with ChatGPT, and I can see that having access to it will make me a better global editor-in-chief for Insider. My conclusion after a fair amount of experimentation with ChatGPT is that generative AI can make all of you better editors, reporters, and producers alike,” Nicholas Carlson said in a statement.
The Guardian published a similar memo but took a more cautious stance.
When we use generative AI, we focus on situations where it can improve the quality of our work, for example, helping journalists interrogate large sets of data, assisting colleagues with corrections or suggestions, creating ideas for marketing campaigns, or reducing the bureaucracy of time-consuming commercial processes
The New York Times was even more conservative, with executive editor Joe Kahn saying at the recent World Media Congress in Taipei that the paper enthusiastic about experimenting, but extremely cautious about what we would be willing to present as a final product to our readers.
One of the most common provisions for using generative AI in newsrooms is external transparency. If an article or video is created or augmented by AI, it should be labeled as such. The Associated Press has relied on data-to-text AI for years to create short articles based on structured data, such as earnings reports. In all cases, they contain a notice detailing their origin and original source: This story was generated by Automated Insights using data from Zacks Investment Research.
Another common limit that news outlets are applying to this nascent technology is that human review is generally required for any output seen by the public. In an industry built on trust and accuracy, tools and workflows must include the opportunity for a validation step to prevent the spread of misinformation. Over time, as technology improves, it will be possible to achieve more end-to-end automation. Today, full automation would require a certain tolerance for risk.
Define a development roadmap
Once you’ve established your framework for engaging with generative AI, the next step is experimenting with purpose. Choose one or two projects with the direct goal of unlocking value or solving a problem in the newsroom. Make sure it has a reasonable development schedule. No very ambitious projects for your first effort. You want to dive into this and understand the limits.
Then, set clear and measurable goals. If you are trying to optimize headlines, what is the definition of success? Will you track adoption and usage metrics, such as an increase in the number of tests? Or perhaps a performance metric like click-through rate (CTR). And where will the product be located? If it’s just another browser tab that the audience team needs to have open, they may face an adoption barrier. Also make sure that the product team or innovation squad that develops generative AI tools is not isolated. Bottom-up suggestions about workflow efficiencies come with the advantage that the people doing the work are stakeholders in the success of this new tool.
One of the early adopters of AI initiatives is Forbes. The publisher’s content management system, called Bertie, can recommend article topics to contributors based on their previous offerings. A politics writer would receive suggestions about politics, while a technology writer might receive ideas from Silicon Valley. It can also suggest headlines and images to reduce production time.
Finally, you are not alone in this. There is a lot of good work being done globally to help educate and support publishers big and small. The Center for Cooperative Media has published an excellent guide on using prompts and use cases for generative AI, specifically aimed at journalists. The Journalism AI initiative at the London School of Economics and Political Science offers everything from classes to case studies.
What can generative AI do for me?
When we approached this question at Taboola, we started by roughly mapping the end-to-end production flow in a newsroom. While all newsrooms have different roles, the process is generally the same: you have an idea for something you want to cover, you gather information, you create it, you publish it, and then you work hard to attract an audience to that piece of journalism. Following this progress, we have identified many tasks and areas where AI shows promise.
Discovery tasks
How can generative AI help a reporter decide what article to write about today? There are actually several ways, surprisingly.
Generative AI can help with data analysis, identifying sources, analyzing trending topics and generating/indicating ideas, as occurs at Forbes. I can also imagine a system in which it could even help forward tips or news messages.
Often, the most powerful articles, those with impact, require the human intelligence and intuition of a journalist. Generative AI won’t directly help you develop sources, build that individual empathy, although it can help you find them.
Creation tasks
AI can help with creating articles from structured data, transcribing notes, researching topics, editing style, packaging content, transforming content into alternative story forms, and intelligently sharing content across networks.
With oversight and fact-checking, I’ve also seen generative AI help with research and create compelling data visualizations. However, generative AI cannot take photos or videos of a protest scene or conduct interviews with protesters. Again, there is no substitute for in-person reporters.
Optimization/delivery tasks
We have found this to be perhaps the richest area of workflow efficiency. Generative AI can help with content translation, SEO/SEM, headline optimization, social sharing, browser and native app alerts, email and newsletter alerts, content curation, and real-time data analysis.
If you take a solutions-based approach, you’ll find many areas where generative AI can help a newsroom, suggesting related links or headline options, hopefully freeing reporters to do impactful work rather than commoditized tasks.
What happens next?
In fact, even if your newsroom doesn’t do anything with generative AI, the world is changing around you.
Google’s Generative Search Experience (SGE) is still in development, but even with recent changes that highlight more sources, it’s likely to reduce referral traffic. How impactful is that? In a new Taboola survey of 3,700 sites active since 2019, the Google, in the first half of 2023, represented 34% of all reference sessions. For many publishers individuais, Google may represent half or more of all referrals. Any erosion in searches means a significant impact on the audience.
Specifically, we believe that utilitarian searches (those related to getting an answer to a specific question) will likely decline first and most quickly. Questions like When is the World Cup?, What is my horoscope? or travel destinations in Sydney will be answered without clicking on a link. Searches related to brands, authors and real-time news events should be more durable. Ecommerce and affiliate links tied to search performance can be eroded as questions like What is the best blender? can be aggregated and displayed without needing to visit a website.
However, Google Bard currently excludes medical, legal, and financial advice, so these sectors could represent a long-term opportunity. You probably also won’t be able to compensate for lost searches by using generative AI to create content at scale. By definition, Google says content created by AI is spam and would result in demotion or removal.
In addition to the challenges of the searches, the publishers Those focused on subscriptions will also likely face difficulties. Firstly, search traffic converts reliably, far ahead of social networks, for example. Therefore, any decrease in searches will have an impact. Secondly, the interfaces from SGE/GenAI isolate information from its source, reducing brand trust and affinity. The world may be less aware of where that information came from and, in the long run, less aware of how valuable you are as a publisher.
What does Taboola recommend?
First, try using generative AI, whether developing it in-house or using one of the thousands of tools available. This is a transformational technology that isn’t going away. Use it to make journalists’ work easier, freeing them from mundane tasks. Also, have fun with her. See what she can do.
Second, some loss of search traffic is inevitable, probably not in 2023, but soon. Prepare by performing a risk analysis of what would happen if vulnerable SGE searches were to disappear. Because this can happen.
Finally, become a destination website. Be a source of credibility in a world of chaotic content. You publishers Those who can maintain and expand direct relationships with readers and viewers will see greater success. Use and build tools that reduce the distance between you and your audience. And this advice is as valid today as it will be a decade from now.
* This text does not necessarily reflect the opinion of Adnews
Follow Adnews on Instagram,LinkedIneThreads. #WhereTransformationHappens
