The distinction between fact and opinion is not based on the content of a statement, but on whether it can be verified. A fact can be confirmed or disproved by an independent source. An opinion expresses a judgment, preference or interpretation. Confusing the two undermines any effort to find reliable information online, and recommendation algorithms make this confusion more common than many people realize.
Recommendation algorithms and the amplification of disinformation
Recommendation systems on major platforms do more than distribute content: they rank it according to engagement metrics. An article that provokes anger or surprise will be shown to more users, regardless of whether it is true. The problem, then, is not just the misleading content itself, but the mechanism that amplifies it until it dominates a news feed.
The European Digital Services Act (DSA) requires very large online platforms and search engines to assess the systemic risks associated with their algorithms. This obligation also covers misleading content that is not illegal, shifting the focus of responsibility: the issue is no longer just removing posts, but analyzing why content is shown and amplified.
The Code of Practice on Disinformation, updated at the European level, includes commitments on the transparency of recommendation systems, researchers’ access to platform data, and the demonetization of misleading content. For readers, the practical takeaway is simple: the fact that content appears at the top of a feed or search results does not guarantee that it is reliable or neutral.
Media outlets specializing in fact-checking, such as those featured on libreinfo.org, make it possible to cross-check information against local and independent sources, complementing the fact-checking tools used by major newsrooms.

Verifying online sources: a practical method
Verifying a source involves more than just looking at the website’s name. The method known as lateral reading, used by professional fact-checkers, involves opening a new tab to find out what other sources say about the author or outlet, rather than first reading the content in depth.
Here are the technical criteria to apply systematically:
- Identify the author and check their previous work on other platforms. An unsigned article or one published under an untraceable pseudonym calls for extra caution.
- Check the publication date. Old content resurfacing out of context is a common vehicle for disinformation, even if the facts it originally reported were accurate.
- Trace the information back to its primary source. If an article cites a study, report or official statement, check that the document actually exists and that the statements have not been truncated or reframed in a biased way.
- Analyze the domain name. Clones of legitimate websites use subtle typographical variations to impersonate recognized media outlets.
Content shared widely has not necessarily been verified. Virality on social media is driven by emotional dynamics, not an editorial process.
AI-generated content and transparency labels
European regulations on artificial intelligence require transparency for certain content produced or modified by AI systems. An “AI-generated” label indicates the content’s origin, but it does not tell you whether the claims it contains are true.
AI-generated text can be factually accurate. Human-written text can be riddled with false information. Labels help explain where content came from, not whether it is reliable. Provenance markers (C2PA and similar initiatives) make it possible to trace the history of media (photos, videos, audio) from the point of creation, but they do not replace content verification.
The proliferation of deepfakes makes things even more complicated. A realistic-looking video is no longer proof: reverse image searches and verification of the original source remain the only reliable steps to take when faced with suspicious visual content.

Building a reliable and diverse news routine
The quality of the information we consume depends less on the tools we use than on how carefully we read. Consulting several sources on the same topic, while varying editorial perspectives and geographic coverage, automatically reduces the risk of confirmation bias.
Distinguishing facts, analysis and opinion within an article
A news article often combines three types of content. Facts are verifiable details (dates, places, quoted statements). Analysis interprets those facts through a particular lens (economic, legal or geopolitical). Opinion offers a value judgment. Identifying these three types of content in every article helps prevent an interpretation from being mistaken for an established fact.
Editorials and opinion pieces are explicitly presented as opinion. The problem arises when content presented as informative slips into opinion without making that clear, particularly through the use of evaluative adjectives (“scandalous,” “historic,” “alarming”) or biased wording.
Reducing exposure to filter bubbles
Algorithms personalize news feeds based on browsing and interaction history. To break out of these bubbles:
- Go directly to media outlets’ websites rather than relying exclusively on social media feeds.
- Use aggregators that offer a range of sources without algorithmic personalization.
- Subscribe to newsletters curated by established newsrooms, where journalists—not engagement algorithms—select the content.
Finding reliable information requires active effort, not passive consumption of content pushed by systems designed to increase the time users spend on a platform. The DSA has established a regulatory framework, but readers remain the last line of defense against disinformation.



