1How would a student in 2025 go about finding information for a school assignment about the country of Taiwan? This could be an assignment in a Swedish school where students have to find and analyse information. Would she go to the school library and look it up in a printed encyclopaedia? Would she start by typing “Taiwan” into a search engine like Google and clicking on the first link (often a Wikipedia page)? Would she search for Taiwan on TikTok or another social media platform? Or would she instead turn to one of the many generative AI apps that have come onto the market since the end of 2022?
2Young people are constantly immersed in streams of information through social media and other aspects of today’s digital infrastructure. Popular culture, news, and fact-oriented content are interwoven with political messages on platforms built on a commercial logic. In a politically turbulent time, the increasing presence of AI-generated, synthetic content has heightened concerns – both in Sweden and globally – about the harm that disinformation can cause to both individuals and society at large. Various approaches have been proposed to address this problem. For example, digital platforms employ different forms of moderation, while established media rely on fact-checking, either conducted in-house or by external agencies. An alternative to these forms of information control is the promotion of media and information literacy, in particular the promotion of students’ ability to critically evaluate information sources.
3In Sweden, the importance of evaluating information has been repeatedly emphasised, especially in formal education at different levels (Limberg et al., 2008). However, today I will argue that the question of how to assess information is analytically underdeveloped. I will discuss two crucial challenges when it comes to preparing young people to navigate conflicting truths in a digital society: what does it mean to be critical of information in a digital infrastructure increasingly shaped by algorithms, datafication, and AI-generated content? How can a critical evaluation of sources be conducted when sources are increasingly difficult to recognise due to a digital landscape permeated by AI? In the following sections, I will first explore the concept of being critical per se, before discussing the challenges of evaluating sources in a digital environment dominated by synthetic information.
4If the hypothetical student mentioned in the introduction of this paper were to use a search engine to find information about Taiwan, she would be confronted with a vast number of links. As with many search queries, the first result is often a link to Wikipedia. As of 29 January 2025, the English-language Wikipedia article on Taiwan states (with supporting references): “Taiwan, officially the Republic of China (ROC), is a country in East Asia” (Wikipedia, 29 January 2025). However, if the student consults the Mandarin-language version of the article, it states that Taiwan is a part of China – also with supporting references. This example illustrates that what you learn about Taiwan depends on the source you read. And the sources you read are influenced by the search engine’s ranking algorithm and the sources you trust.
5In Sweden, interest in the critical evaluation of sources in school has developed in parallel with the spread of the internet and social media, where information is less scrutinised before publication compared to traditional printed media. In recent decades, the ability to search, identify, and critically evaluate information sources has been considered an integral part of education, from compulsory schooling to university. In contrast to simply learning about a topic – such as Taiwan – through the content of a textbook, the critical evaluation of sources involves scrutinising the credibility of the textbook and other sources. Rather than focusing solely on content, a critical evaluation of sources also considers paratextual elements such as the author, publisher, publication date, and other contextual factors.
6In the Swedish educational system, the critical evaluation of information sources is often referred to as “källkritik”. The concept of källkritik was developed within the academic discipline of history as early as the 18th century. In Sweden, it has been used in recent decades to describe an approach to determining whether digital information should be trusted. Concepts are difficult to translate, but in English källkritik could be translated as criticism of information sources or evaluation of information sources. Getting to know Taiwan can also become an exercise in learning to identify and scrutinise the sources of information used for the assignment.
7But what does it mean to be critical when evaluating sources of information? In my research (e.g. Haider and Sundin, 2022), I have identified different approaches to the critical evaluation of information: a sceptical approach and a pragmatic approach. Through interviews with young people in their late teens, I have found that the notion of being critical is often interpreted as being sceptical, or even cynical – you do not believe in something unless you have seen it with your own eyes. This sceptical approach is underpinned by distrust and even suspicion of established sources of information such as the so-called mainstream media, university research, and other public knowledge institutions. The sceptical approach is rooted in an extreme form of empiricism that disregards the evaluation of sources in favour of individual responsibility for evaluating content. However, unless one already has detailed knowledge of a topic, evaluating the accuracy of information content is extremely difficult. How can an ordinary student critically assess the content of a text about Taiwan – a country (or part of a country) about which they have little prior knowledge?
8For the evaluation of information sources in school to be a meaningful assignment, there must be a common understanding in society about which sources – including the institutions behind them – are of good quality and which are not (Haider and Sundin, 2022). The critical evaluation of sources can only be an effective method for determining trustworthiness if there is at least some degree of societal agreement about which sources are reliable and why. However, this is not always the case. In a contemporary classroom with students from different backgrounds, do all students have the same ideas about which sources they can trust when reading about Taiwan? The evaluation of sources and trust in their credibility are closely intertwined. If teaching students how to critically evaluate sources is an educational goal, this should go hand in hand with discussions about which sources they can trust and why. That said, trust should never be a goal in itself – information sources, as well as the institutions that produce them, must earn trust by being demonstrably reliable and credible.
9The second challenge I would like to address is what I see as the growing trend of information sources becoming increasingly invisible within the information ecosystem (Sundin, 2025). If the hypothetical student from the beginning of this paper were to turn to generative AI for her assignment on Taiwan, she would face a different set of challenges. The response she receives would depend on how she formulates her prompt and which application she uses. However, unlike the well-referenced Wikipedia article on Taiwan, texts generated by so-called large language models (LLMs) are not traditionally linked to specific sources. In early 2025, the Chinese AI application Deepseek attracted a lot of attention in Western newspapers. According to an article in The Guardian, when the student prompted “Is Taiwan a country?” she received a different answer than when she used one of the other leading chatbots. According to the article, Deepseek replied: “Taiwan has always been an inalienable part of China’s territory since ancient times” (Lu, 2025).
10Over the last decade, Google and other major search engines have evolved from simply providing links to sources to providing information directly without needing to click on a link. This can be seen, for example, in Google’s Knowledge Panel, which presents not only links but also fact-based summaries, mainly from Wikipedia. For many search queries, this is not a problem. However, with the public breakthrough of generative AI tools and the increasing integration of generative AI models in search engines, the challenges are becoming greater.
11With the increasing invisibility of sources, traditional approaches to evaluating information are becoming problematic. Several guidelines have been developed to provide students with checklists for evaluating the quality of information sources. A common feature of these guidelines is the assumption of a stable source with a specific originator (such as an author). They often include recommendations such as checking the date of publication, assessing the originator’s reputation, and considering the author’s underlying motivation. But what implications does this have on the teaching of media and information literacy when traditional sources as we know them no longer exist?
12Texts produced by generative AI are the result of a prompt and the subsequent workings of an LLM. Moreover, the workings of these models are based on the data on which they are trained and the way in which they are moderated. Critical voices have cautioned against equating texts generated by LLMs with search engine results (e.g. Shah and Bender, 2022). Unlike traditional search engines, which match queries against an index and redirect users (via links) to various sources, LLMs generate responses based on statistical models formulated in a humanlike natural language (Bender et al., 2021). All LLM-based applications, including the aforementioned Deepseek, include some form of moderation in terms of the types of text they can generate. However, the focus and extent of this moderation varies. At the same time, AI applications are increasingly being integrated into search engines, leading to features such as the summary of top-ranked links in natural language. As sources become increasingly invisible to students (and other users), it becomes apparent that traditional methods of critical evaluation are increasingly difficult to apply.
13To prepare young people for the digital society, they need to develop their ability to search, identify, and evaluate information sources. However, these activities are not just about acquiring digital skills. More importantly, they require an analytical understanding of how to evaluate sources in an increasingly AI-driven digital ecosystem. The first challenge I have addressed concerns the importance of critically evaluating information sources. This involves avoiding blind trust, while at the same time recognising the fine line where motivated, critical questions can turn into cynical distrust. The second challenge I have addressed is how we can enable students to identify and evaluate information sources when these sources are becoming increasingly invisible within the information ecosystem. There are no easy solutions to these challenges in education, but they are crucial to an educational system designed to prepare young people to participate in a democratic society.
Flux RSS
