The most dangerous opinion is the one that won’t admit it is one
It’s easy to recognize an opinion when it begins with “in my opinion” or “I believe that.” It’s harder when the same opinion is dressed up as fact: delivered in a confident, authoritative tone, with no caveats, sometimes even with numbers in the background. And it’s hardest of all when we don’t notice at all that something is persuading us, because it looks exactly like information that doesn’t need to be questioned.
This isn’t a niche skill, useful only to journalists or scientists. It’s one of the most practical competencies you can have today – because, as we’ll show in this article, even well-educated, attentive readers get this distinction wrong much more often than you might think.
The scale of the problem: what the hard data shows
The test most Americans failed
In 2018, the Pew Research Center conducted one of the most frequently cited studies on this topic. More than 5,000 American adults were asked to distinguish five factual statements from five opinion statements taken from real news headlines. The result? Only 26 percent of respondents correctly identified all five factual statements, and just 35 percent correctly recognized all five opinion statements. In other words – fewer than one in three respondents could flawlessly tell fact from opinion even in a carefully prepared test, under calm conditions, without the haste and emotions that accompany everyday news browsing.
Why it’s not a matter of intelligence, but of habit
The study revealed something even more instructive: people were much more likely to classify an opinion as a fact if that opinion matched their own political views. This isn’t an error of knowledge — it’s an error of mental habit, in which agreement with our worldview raises a statement’s credibility in our eyes, regardless of whether it can be verified with objective evidence at all. An interesting complement to this picture is the finding that younger adults consistently handled this distinction better than older respondents — regardless of whether a given statement was closer to left-wing or right-wing sensibilities. This suggests that the skill is to some extent learnable, not just a matter of temperament or intelligence.
What an opinion pretending to be a fact sounds like – the linguistic signals
Linguists and rhetoric theorists have long described the specific linguistic constructions used to make an opinion sound like a fact. In the English-language literature, they’re called “weasel words” – from the habit of weasels that suck out the contents of an egg, leaving an empty shell that looks untouched from the outside. A sentence built from such words looks like a claim, but it’s empty inside – it can be neither defended nor refuted, because it doesn’t indicate who is actually making it.
Specific constructions worth watching for
“It is said that…” / “It is known that…” / “It is widely believed that…” — the classic passive voice with no agent. The sentence sounds like an echo of common knowledge, though in practice no one knows who specifically claims this or on what basis.
“Experts say…” / “Studies show…” — without naming a person, an institution, or the title of a specific publication, this sentence borrows the authority of science while offering no way to check whether that authority really stands behind it. This is exactly the type of construction we warned about in our articles on verifying AI-generated citations — language models have a documented tendency to generate precisely such unsupported references to “studies” and “experts.”
“Of course” / “Clearly…” / “No one can deny that…” — words that add no factual content but socially make debate harder: anyone who asks “and how do we know that?” after the word “of course” risks sounding as if they don’t understand something elementary.
“Supposedly” / “Allegedly” / “Reportedly” — on the surface these sentences sound cautious, but in practice they allow the author to spread unconfirmed information without taking responsibility for it — the linguistic equivalent of turning your back the moment someone asks about the source.
Loaded adjectives — “controversial,” “radical,” “questionable” — don’t describe a fact; they suggest in advance how you should judge it before you even know the details.
Even AI does it – and often
It’s worth adding something that connects directly to the subject of our guide: an analysis published in a scientific journal in 2025 noted that texts generated by language models such as ChatGPT contain an exceptionally large number of “weasel words” — even though the model itself has no intention to manipulate; it simply reproduces the language patterns that are statistically most likely, and these are often full of imprecise phrasing. This is further proof of why the ability to recognize such constructions is more important today than ever — you’ll encounter them not only in the newspaper, but also in every conversation with AI.
When an opinion has an entire marketing budget behind it: sponsored content
A study that should worry every internet reader
This is one of the most sobering research findings we came across while writing this article. Michelle Amazeen, a researcher at Boston University, conducted an experiment in which participants were told outright that they were looking at an advertisement — and yet more than 9 out of 10 people were convinced they were reading a regular editorial article. In another, related study by the same author with 800 American adults, fewer than 1 in 10 participants could correctly recognize sponsored content as an advertisement, even when it was properly labeled.
Why it works so well
The point isn’t that readers are inattentive. Research shows that the very way advertising content is labeled matters enormously – labels such as “advertisement” or “sponsored content” are recognized seven times more often than vague terms like “brand voice” or “presented by.” The problem is that many newsrooms and brands deliberately choose the latter, less clear labels – because, as the same research shows, an article recognized as an ad loses credibility in the reader’s eyes. A further complication is that sponsorship labels disappear in more than half of cases when an article is shared on social media – exactly where most of us read today.
Checklist: specific questions worth asking yourself
Based on the research above, here is a set of practical questions that help distinguish fact from opinion in practice:
- Can the statement, in principle, be refuted with specific evidence? If not – it’s most likely an opinion, no matter how confident it sounded.
- Does the author indicate WHO claims what they present as fact? “Experts believe” isn’t an answer — it’s a question without an answer, disguised as a statement.
- Do the words “of course,” “clearly,” or “it is known that” appear in the text? That’s a signal that the author is counting on you not asking how it’s known.
- Is it definitely an article, or is it sponsored content? Check the top and bottom of the page – look for the word “advertisement” or “sponsored,” remembering that research shows that less distinct terms are sometimes deliberate.
- Do the adjectives describing the topic (“controversial,” “shocking”) already tell you what to think about it before you’ve learned the facts?
In practice, these questions are an extension of the “Investigate the source” step of the SIFT method, which we describe at length in our guide — applied here not to an entire website, but to a single sentence.
An example of applying it in practice
Let’s compare two sentences describing the same event:
“The city authorities’ controversial decision, which experts call a mistake, sparked a wave of outrage among residents.”
versus:
“The city council voted 15 to 12 to raise parking fees by 20%. The Urban Drivers’ Association published a statement criticizing the decision; three council members who voted for the increase did not respond to a request for comment.”
The first sentence sounds like information, but it consists entirely of imprecise judgments: who exactly called it a mistake? How many residents expressed outrage, and in what way? The second sentence, less emotional, actually gives you data for your own assessment — a specific vote count, a named organization, a clearly stated lack of comment from specific people. That’s exactly the difference this article is about.
Why it pays off at all
Both weasel words and unclearly labeled sponsored content are neither accidents nor oversights — they’re the result of the same economic mechanism we wrote about in our article on filter bubbles: content that sounds confident and emotional generates more clicks and shares than cautious, precise content. The market for native content (sponsored, looking like regular articles) in the United States alone was estimated a few years ago at more than 21 billion dollars a year, with a forecast that it would account for nearly three-quarters of all online publishing advertising revenue. Cautious language and clear labels aren’t rewarded in this system — they’re punished with fewer clicks.
Summary: one rule to finish
If a sentence sounds so confident that you don’t even feel like questioning it — that’s exactly the moment when it’s most worth stopping. The greatest strength of an opinion dressed up as fact is that it doesn’t look like something that needs checking. And at the same time — as the research cited here shows — even the most attentive readers regularly get this distinction wrong, so it’s not about shame or failure, but about developing a specific, repeatable habit of asking: who is saying this, and how do they know?
Want to learn this systematically?
This article shows one specific slice of a much broader set of skills that we develop step by step in our guide. There you’ll find a complete source verification method (SIFT and CRAAP), ways to recognize AI hallucinations, and dozens of other practical tools — not just for reading individual sentences, but for navigating today’s internet consciously and holistically.
📖 Buy the “How to Search?” guide at jakszukac.com – complete, practical knowledge in one place, with no filler (and no hidden opinions dressed up as facts).
Did you like this article? You can help us keep developing it:
☕ Support us on Suppi – we create these materials in our free time, and every bit of support really helps us write more and more often.
▶️ Watch our YouTube channel – the same topics, in a shorter video format, with concrete on-screen examples.