About personalization through algorithms and AI
In his guest article, Arian Okhovat uses data, studies and cases to show how AI is changing editorial offices around the world and personalizing content in a targeted manner.
In his guest article, Arian Okhovat uses data, studies and cases to show how AI is changing editorial offices around the world and personalizing content in a targeted manner.
The way we consume news is (once again) on the verge of fundamental change - and behind this is, in particular, personalization through artificial intelligence (AI). What was considered an ambitious promise for the future just a few years ago is now becoming a reality in newsrooms all over the world - from small to large. Media companies and agencies are increasingly relying on AI-supported technologies to tailor content to the individual needs of their readers.
Studies confirm this trend: according to an analysis by the International News Media Association (INMA), the time users spend on news portals doubles when content is personalized. The Frankfurter Allgemeine Zeitung shows that subscribers who use personalized summaries return almost twice as often. A full 81% of users found the AI-supported summary function useful - a clear indication that this technology is taking the relationship between readers and publishers to a new level. Initial results from our PANTA RHAI research study, which we launched together with the University of Hamburg and the University of Rostock, also show that consumers are generally not averse to AI-generated content. However, it remains important to strengthen trust in the credibility of this content and to clearly communicate the benefits, such as personalization.
The advantages are manifold. At the Kölner Stadt-Anzeiger, an AI-based recommendation system led to an 80% higher click-through rate and a 13% increase in articles read in full. These examples show that AI not only improves the quality and relevance of content, but also brings measurable (economic) success.
As technological progress drives the media industry forward, the focus is increasingly shifting. From overambitious visions to tangible, real added value. At the INMA Innovation Week in September, where I was able to share my experiences with AI in the media environment and take a look into the future, one thing became clear: AI is more than just hype today - AI delivers resilient improvements and is becoming an integral part of modern business models in the media industry. Mastering this change is a challenge for many media companies. But also an enormous opportunity.
In the practice of many media companies, it is becoming increasingly clear how transformative the use of AI can be for the personalization of content. In my projects with editorial teams and media companies, I see how personalization is not only changing the consumption of content, but also editorial processes and the way news is prepared. In my opinion, three main approaches can be clearly distinguished:
Personalized preparation of content: The way in which content is consumed plays a crucial role in user engagement. More and more readers want news to be delivered in a format that suits their lifestyle. Whether as easy-to-understand text, on-the-go audio formats or personalized video, users expect this customized variety. Platforms such as the BBC and the Washington Post have been using text-to-audio technologies for years to make content more accessible.
Personalization through recommendation systems: In many media companies, personalization of home pages and feeds has long been standard practice. Recommendation engines based on algorithms and AI ensure that readers are shown targeted content that matches their interests and previous reading habits. As mentioned at the beginning, the Kölner Stadt-Anzeiger was able to increase its click-through rate enormously by implementing such systems. The number of articles read in full also increased significantly. Such success stories illustrate the direct influence these technologies have on user behavior and the performance of a medium.
Personalization of editorial input: Another personalization approach that I find very exciting is the use of different data sources to make content (hyper)local and relevant. It feels like data is now a dime a dozen - much of it is available free of charge via interfaces from local authorities/cities, the German parliament, public utilities, etc. Editorial teams can use live information such as local demographic data, weather or traffic reports to create articles that are precisely tailored to the needs of a region. A successful example of this is Ringier from Switzerland, which uses weather data in Blick to create daily articles that are particularly relevant to the region.
Long before the recent hype surrounding AI, media companies were already looking for ways to make content more flexible and accessible. The big question: how can news and information be made available to people in exactly the way they want to consume it? Whether as text, audio or even video - the form of consumption is playing an increasingly important role.
The BBC has been actively investing in text-to-audio technologies since 2015 in order to make articles audible. What is particularly exciting is that these audio offerings are not only available in English. This has led to a growing reach, especially internationally. The Irish Times was also an early adopter of audio formats. For several years now, they have been using Speechkit (now BeyondWords) to automatically convert 15 to 20 articles a day into audio.
A particularly original example comes from Norway and Sweden. The publishing house Schibsted has implemented synthetic voices of well-known editors in Aftenposten and other newspapers. This means that readers can consume articles that sound as if they are being read by "real" people they already know. This not only creates trust, but also a very special bond with the listeners. The audio function leads to 58% of listeners listening to the articles in full, which significantly increases interaction. Considering that 13% of readers cancel subscriptions due to lack of time, this shows how audio formats can play an important role in strengthening user loyalty and increasing the value of the subscription. In addition, the use of AI-generated audio articles was found to be almost on par with podcasts.
But it's not just the big international media houses that are focusing on personalization. Regional publishers are also active. The Kleine Zeitung in Austria offers content in plain language to make it easier for people with reading difficulties or language barriers to access news. Such initiatives significantly promote accessibility at a local level.
Personalization of content has by no means only been an issue since the rise of AI technologies. With today's technical possibilities - such as synthetic voices, text-to-audio or adapted language levels - it is clear that tomorrow's news consumption can and must become more flexible, individualized and accessible.
It's no longer just about serving content based on editorial recommendations - today, recommendation engines ensure that users get exactly what they are interested in at exactly the right moment. Social media such as Tiktok, Instagram, Snapchat and tech giants such as Netflix and Amazon are probably best known for this. But recommendation algorithms are also used beyond this to increase interactions and user loyalty.
The Kölner Stadt-Anzeiger newspaper uses automated recommendation systems to design the entire content of its homepage. The result? An 80% increase in the click-through rate and a 13% increase in user interaction. This personalization ensures that readers access content that is relevant to them in a more targeted manner. A clear win for the publisher - and an improved experience for the readership.
During the INMA Innovation Week in Helsinki, several Finnish publishers presented how they successfully automate a large part of their homepages using recommendation algorithms. At the Finnish media group Sanoma, where more than 75% of traffic is generated via direct traffic, the editorial team only determines the first three articles, while the rest is automated.
The Polish news portal Onet was able to personalize its content to a greater extent and significantly increase user loyalty through the targeted use of AI tools. A 600% increase in content variety and a 50% increase in dwell time show the success of data-driven personalization. Onet uses an AI-based engine that analyzes both user behavior and content to deliver tailored messages to different audiences. This has led to a growth in Onet audio app subscriptions and an overall increase in the number of loyal users by around 20%.
Spotify is probably the best known name when it comes to personalization in media. The Discover Weekly and Daily Mix playlists are a daily port of call for many users. By using algorithms based on listening habits, Spotify achieves significant increases in user loyalty and dwell time. With new features such as AI Playlist, which was launched in beta this year, Spotify wants to expand this personalization even further. If you want to delve even deeper, Spotify Research provides exciting insights into the technical strategy and logic of the algorithms behind these functions and successes.
Personalization through recommendation engines is no longer a nice extra - it is an essential part of the modern user experience. From publishers to streaming platforms, it is clear that data-based personalization not only increases user satisfaction, but is also (economically) extremely attractive and, above all, necessary for companies.
Personalization in the media industry goes far beyond adapting content to individual reading habits. One exciting example is the Swiss publisher Ringier, which works with weather data to automatically generate content. At Blick, for example, articles are generated that not only present the weather, but also offer readers suggestions for activities on rainy days - based on existing editorial content. This creates meaningful content with daily weather information that enriches readers' everyday lives. This personalized, data-driven content is well received by consumers - 7% of daily visitors also access it - and increases their interaction with Blick.
But the potential goes much further. Real-time data is a particularly useful way for smaller, local editorial teams to target their local readership. Local data sources - such as traffic statistics, data from municipal utilities or even databases from regional parliaments - can be used to create precise, (hyper)local, relevant content. Such data helps to identify important developments in the region and put them into a direct, user-friendly context. Demographic data - from population structure to income distributions - can also provide exciting insights for readers.
As part of a white paper by PANTA RHAI, which was createdin collaboration with Media Lab Bayern, we analyzed local data sources, among other things, in order to make trends in areas such as sustainability or demographic changes usable for journalistic purposes. For editorial teams, especially those with limited resources, AI-supported tools offer new opportunities to increase their relevance and implement data-driven local journalism more efficiently.
The personalization of content based on locally relevant data offers enormous potential to strengthen local journalism in the long term. By using various publicly accessible data sources, editorial teams can create content that is not only informative but also directly relevant to the everyday lives of their readers. This increases efficiency in editorial offices and promotes trust and relevance in local media.
Even though the potential of AI and personalization is enormous, this development does pose some challenges. In order to maintain a balance between technical progress and ethical concerns, we need to take a closer look at the challenges that affect not only implementation, but also the long-term consequences. Especially when it comes to the application in the media industry - and especially in journalism - we face some complex issues.
Personalization is partly based on user data - and this is precisely where one of the biggest hurdles lies. The more precisely content is tailored to the needs of readers, the more personal data has to be processed. Especially in Europe, where, for example, the General Data ProtectionRegulation (GDPR) strictly regulates the handling of personal data, media companies must pay particular attention to data protection. Transparency in the handling of data is essential. Trust is the key. Losing the trust of readers would cause long-term damage.
Another risk of personalization is filter bubbles. If recommendation engines only play out content based on users' preferences, there is a risk that they will become increasingly isolated. They will then only see what they like and less what may be contrary to their opinion. Media companies should make sure that their recommendation engines preserve the diversity of content. It's about getting users out of their comfort zone - without losing them.
There is also a risk that human influence will be pushed back too far, especially when it comes to personalizing editorial input. Automation may mean speed and efficiency, but the trained eye of editors must not be missing. Especially in journalism, which also deals with critical topics, human care is crucial. Without this, the quality of the content could suffer and with it the added value that subscribers are willing to pay for.
Technology itself also still has its limits. Although AI is capable of processing huge amounts of data, it often reaches its limits when it comes to the depth of content. Our experience with prototypes such as our local journalism GPT shows that while AI tools can provide useful support, in many cases they still struggle to grasp complex relationships correctly.
Finally, social responsibility also plays an important role. Journalism has the task of supporting an informed society and contributing to the formation of opinion. An excessive focus on personalized content could lead to certain topics or population groups being neglected. We need to ensure that diverse content continues to be delivered that is accessible to all - not just those whose interests and preferences are already known. Personalization must not lead to one-sidedness - this is especially crucial in times of disinformation and increasing polarization.
The introduction of AI and personalization in journalism has permanently changed the way news is consumed and created. What was recently considered a vision of the future is now a reality. Whether through personalized home pages or flexible formats (e.g. text to audio) - AI makes it possible to offer tailored and targeted content. This benefits both readers and publishers: user loyalty grows and the relevance of content increases. However, this change does not only bring advantages. Data protection, filter bubbles and increasing automation pose major challenges that companies must actively tackle in order to ensure long-term trust and quality.
If the trends are to be believed, hyper-personalization will be at the heart of this: Content will not only be individually tailored to users*, but also adapted to their context and situation in real time. We are talking about content that changes dynamically based on the user's current location, time of day or interests. I also find the rapid development in the area of multimodal personalization particularly exciting. Users increasingly expect to be able to decide in which format they want to consume content - be it as an article, podcast or video avatar. These new technologies offer an opportunity to evolve journalism and explore how we can make content even more flexible and accessible.
Nevertheless, it is important that we preserve the diversity of content. Personalization should not (only) lead to people being confirmed in their opinions - it should also open up new perspectives. It is up to media companies to consciously promote this and provide their users with high-quality content that sometimes goes beyond their habits.
At the same time, increasing automation in the editorial sector must not lead to human control and journalistic integrity falling by the wayside. Efficiency is great, but the human factor remains crucial here too. AI can support editorial processes, but it should never replace journalistic diligence.
What journalism needs today is the courage to change. We are only at the beginning of what AI and personalization can mean for the media world. It's not about replacing existing structures, but complementing and expanding them. New technologies such as AI enable us to offer content dynamically and individually - in a way that creates real added value for both readers and the company.
The challenge will be to develop ethical guidelines and ensure that personalization promotes diversity and opinion-forming rather than restricting it. At the same time, the digitalization of journalism and the media industry in general opens up opportunities that we could hardly have imagined a few years ago. With the right mix of curiosity and responsibility, we can make journalism fit for the future - and put the needs of readers at the center.
Arian Okhovat Alavian ist ein Hamburger Experte für Künstliche Intelligenz und Medieninnovation. Als Mitgründer und Geschäftsführer von PANTA RHAI hat er sich auf die Entwicklung von KI-Strategien und -Tools für Medienunternehmen und Agenturen spezialisiert. PANTA RHAI konzentriert sich auf die Nutzung und Entwicklung von GenAI und Machine-Learning-Modellen, sowie deren Implementierung in Unternehmen.
Arian ist aktiv in der Forschung zur Medieninnovation tätig und beteiligt sich unter anderem an AI for Good und EU-geförderten Projekten. Neben seiner beruflichen Tätigkeit engagiert er sich als Mitglied der Global Shapers Community des World Economic Forum und unterstützt weitere Initiativen, wie GermanDream, Social Impact Award, 2hearts und weiteren.