AI tools helped gather research and write the first draft of this article. I edited it and checked the quotes against the original documents (study abstracts via Europe PMC and Crossref; IPSO's Editors' Code; FCA and SEC press releases). The cover illustration and the narration voice in the videos are AI-generated. Details at the end.
"Fees crashed 40%!!!" in a huge font, with no source, feels convincing. "Median fee per transfer fell about 40% in October, according to a public dashboard, on one network" (an invented example, worked through below) feels smaller and duller. Only the second one can be checked, and only the second one tells you how far to trust it.
Confidence is a tone. Evidence is something a reader can open, compare and question. When a claim sounds certain, ask what it rests on: a source, a method, a date. If those are missing, the certainty has nothing underneath it. For writers, the same rule runs the other way: sound as confident as your evidence allows, and no more.
This is the fifth part of a series on making editorial work inspectable. Earlier parts covered showing your evidence, disclosing how AI tools were used, correcting mistakes visibly and attributing other people's research honestly. This one is about the voice those facts are delivered in.
Why confident claims persuade
The pull of a confident claim is not a sign that readers are careless. It comes from shortcuts that usually work, and that are easy to exploit.
People prefer confident advisers, even overconfident ones. Paul Price and Eric Stone ran three experiments in which college students evaluated two fictional financial advisers who judged the likelihood that various stocks would rise. One adviser was reasonably well calibrated; the other was overconfident. "In all three experiments, participants tended to prefer the extreme advisor." In two of the experiments, the advisers' confidence shaped how knowledgeable participants thought they were. The authors describe this as a "confidence heuristic": people assume a more confident adviser is right more often and knows more [1]. (Peer-reviewed experiments with student participants.)
But confidence backfires when it turns out to be wrong. Elizabeth Tenney and colleagues found, in two experiments on witness testimony, that "errors in testimony damage the overall credibility of witnesses who were confident about the erroneous testimony more than that of witnesses who were not confident about it." After an error, "less confident witnesses may appear more credible than more confident ones." Their interpretation is that people judge whether a source's confidence matches its accuracy, which the authors call calibration [2]. (Peer-reviewed experiments.) In other words, once an error comes to light, people judge a confident source more harshly than a cautious one.
Easy-to-read claims feel more true. Rolf Reber and Norbert Schwarz showed participants statements such as "Osorno is in Chile" in colours that made them easy or hard to read. "Moderately visible statements were judged as true at chance level, whereas highly visible statements were judged as true significantly above chance level." Their conclusion: "perceptual fluency affects judgments of truth" [3]. (Peer-reviewed experiment.) Big type and high contrast make a claim easier to process, not more accurate.
Repetition makes claims feel true, even when we know better. Research on the illusory truth effect shows that repeated statements are easier to process and so seem more truthful. Lisa Fazio and colleagues tested whether knowledge protects against this and found that "illusory truth effects occurred even when participants knew better" [4]. (Peer-reviewed experiments.) A claim that is everywhere in your feed can feel established simply because it is everywhere.
Taken together, these findings match, in my view, how a lot of crypto content is built: confident, visually loud and repeated. None of these studies was done on crypto readers, and the effects come from laboratory tasks, so they show that the mechanisms exist, not how strong they are in your feed.
Does admitting uncertainty cost trust?
Writers often fear that hedging will make them look weak. The best evidence we found says the cost is small, and depends on how uncertainty is expressed.
Stated uncertainty barely dented trust. Anne Marthe van der Bles, Sander van der Linden, Alexandra Freeman and David Spiegelhalter ran five experiments, including a preregistered replication with a national sample and a field experiment on the BBC News website, with 5,780 participants in total. People did notice the uncertainty, but "we observed only a small decrease in trust in numbers and trustworthiness of the source, and mostly for verbal uncertainty communication." The authors suggest the results "could help reassure all communicators of facts and science that they can be more open and transparent about the limits of human knowledge" [5]. (Peer-reviewed experiments, PNAS 2020.)
Numbers work better than vague words. A systematic review by Abel Gustafson and Ronald Rice covered 48 experimental studies. It found that most negative effects came from experiments that presented uncertainty as disagreement among scientists. By contrast, "uncertainty in the form of quantified error ranges and probabilities ... has had only positive or null effects, not negative effects" [6]. (Peer-reviewed systematic review, not a meta-analysis.)
The practical reading: "about 40%, give or take a few points" costs you little or nothing. "Some experts think it might possibly have fallen" costs more, and tells the reader less.
What the rules say
IPSO Editors' Code (UK). Clause 1 says "the press must take care not to publish inaccurate, misleading or distorted information or images, including headlines not supported by the text." It also says the press, "while free to editorialise and campaign, must distinguish clearly between comment, conjecture and fact" [7]. (Industry code, enforced by IPSO for its member publishers.) The code does not ban strong opinions. It requires that opinion and guesswork be clearly marked as such, and bans headlines that promise more than the article delivers.
UK Financial Conduct Authority, crypto promotions. Under rules that took effect on 8 October 2023, the FCA requires that "those promoting crypto must also put in place clear risk warnings and ensure adverts are clear, fair and not misleading." Its announcement also quoted Sheldon Mills, executive director for consumers and competition: "Those who invest should be prepared to lose all their money" [8]. (Regulation of financial promotions; it applies to marketing, not to independent journalism, but it shows where regulators draw the line on confident selling.)
When a confident label was the problem. In June 2024 the US Securities and Exchange Commission announced that Terraform Labs and its co-founder Do Kwon agreed to pay more than $4.5 billion after a jury found them liable for securities fraud "after less than two hours of deliberation." The SEC said the May 2022 collapse of the UST token "wiped out $40 billion in market value nearly overnight," and its enforcement director, Gurbir S. Grewal, called the claim of creating an "algorithmic stablecoin" a lie [9]. (Official statement, regulator's press release; the case concerned fraud, not tone, but it shows how much weight the word "stable" was asked to carry.)
A worked example
The network and the dashboard below are invented for illustration.
A public dashboard tracks transaction fees on a mid-sized network. In October, the median fee per transfer fell from about $0.50 to about $0.30, roughly 40%. The dashboard counts only simple transfers, not smart-contract calls. September had an unusual spike because of a popular token launch.
Version A, the loud one. "FEES CRASHED 40%!!! This chain is now the cheapest in crypto." No source, no period, no definition of "fees". The second sentence is a comparison with every other chain that the data does not support. The capitals and exclamation marks do the work the evidence should do.
Version B, the hedged-into-nothing one. "Some observers suggest fees may have come down somewhat recently." Nobody can disagree with it because it says nothing. Who observed? How much is "somewhat"? It is verbal uncertainty, the kind that cost slightly more trust than numbers in one large set of experiments [5], and it gives the reader nothing to check.
Version C, confident and calibrated. "The median fee per simple transfer fell about 40% in October, from roughly $0.50 to $0.30, according to [dashboard] (link). Part of the drop reflects an unusual spike in September, so the longer trend is smaller. Contract calls are not included."
Version C is not timid. It states a clear number in the first clause. It then tells the reader where the number comes from, what it covers, and the one fact that would change how much it means. A reader can check every part of it.
A phrasing guide for writers
| Your evidence | Sound like this | Not like this |
|---|---|---|
| A measured number from a named source | "Fell about 40% in October, according to…" | "Crashed", "exploded", "collapsed" |
| One study | "A 2020 study of 5,780 people found…" | "Science proves…", "Studies show…" |
| A trend over a short period | "Over one month…", "So far this quarter…" | "Is now", "From now on" |
| Data on one network or market | "On [network]…" | "In crypto", "Everywhere" |
| Your judgement | "I think…", "In my view…" | Stating it as fact |
| Genuine uncertainty | A range: "between 30% and 45%" | "Possibly", "might", "some say" |
Three habits matter most. Lead with the number, then give its limits, so confidence and honesty arrive together. Use ranges and probabilities instead of vague hedges; the research suggests readers handle them well [6]. Keep the size of the claim and the size of the type in proportion: if a headline needs capitals to land, the evidence is probably not carrying it.
Three questions for readers
When a post, headline or influencer sounds certain, look for three things:
- Source. Where does the number come from? Can you open it?
- Method. How was it measured, and what does it include or leave out?
- Date and scope. When, over what period, and for which network, market or group?
If you cannot find them, the confidence is decoration. That does not make the claim false. It means you have no reason yet to believe it, and repeating it would pass the problem on.
Common mistakes and how to fix them
1. Volume instead of evidence. Capitals, exclamation marks, "huge", "massive". Fix: replace each intensifier with the number it is trying to suggest.
2. Scope creep. One network becomes "crypto"; one month becomes "now". Fix: name the network and the period in the same sentence as the claim.
3. Vague hedging. "Might", "possibly", "some say", stacked until the sentence means nothing. Fix: give a range, a probability or a named source who disagrees.
4. Certainty by repetition. The same claim across ten posts feels confirmed [4]. Fix: before repeating a claim, find its original source, not the loudest copy.
5. The confident label. "Stable", "safe", "guaranteed", "risk-free". Fix: use such words only when you can show what they rest on, and say what would make them stop being true.
What this article cannot tell you
The studies cited here were mostly done in laboratories or online experiments, often with students, and none of them tested crypto media. They show that confidence, fluency and repetition affect how true claims feel, and that clear uncertainty need not cost much trust. They do not tell you how much any of this moves a particular audience. A calibrated tone is also no substitute for checking: a carefully worded claim built on a bad source is still wrong.
How this article was made
| Task | What the AI did | What a person did | How it was checked |
|---|---|---|---|
| Research | Searched for studies, codes, regulations and cases; summarized them | Chose the questions and the sources to rely on | Study abstracts [1]–[6] retrieved from Crossref and Europe PMC and compared with the quotes; IPSO [7] via the Editors' Code Committee page, FCA [8] and SEC [9] pages opened directly |
| Draft | Wrote the first draft from the notes | Edited the text and the structure | A separate fact-check compared figures and quotes with the sources |
| Images and video | Generated the cover illustration; produced the narration voice of the companion videos | Chose the illustration; designed the explanatory diagrams; wrote the narration scripts | Labelled as AI-generated in the disclosure line and alt text |
| Not done by AI | — | The decision to publish and responsibility for the text | — |
Conflicts of interest: none. The author has no positions in the assets named; the network in the example is fictional.
Responsible for this article: Dmitriy Lastovka, author. Corrections: contact form.
Sources
Labels: code, regulation, official statement, research (peer-reviewed unless noted).
- Price, P. C., & Stone, E. R. (2004). Intuitive evaluation of likelihood judgment producers: Evidence for a confidence heuristic. Journal of Behavioral Decision Making, 17(1), 39–57. doi:10.1002/bdm.460. Research.
- Tenney, E. R., MacCoun, R. J., Spellman, B. A., & Hastie, R. (2007). Calibration trumps confidence as a basis for witness credibility. Psychological Science, 18(1), 46–50. doi:10.1111/j.1467-9280.2007.01847.x. Research.
- Reber, R., & Schwarz, N. (1999). Effects of perceptual fluency on judgments of truth. Consciousness and Cognition, 8(3), 338–342. doi:10.1006/ccog.1999.0386. Research.
- Fazio, L. K., Brashier, N. M., Payne, B. K., & Marsh, E. J. (2015). Knowledge does not protect against illusory truth. Journal of Experimental Psychology: General, 144(5), 993–1002. doi:10.1037/xge0000098. Research.
- van der Bles, A. M., van der Linden, S., Freeman, A. L. J., & Spiegelhalter, D. J. (2020). The effects of communicating uncertainty on public trust in facts and numbers. PNAS, 117(14), 7672–7683. doi:10.1073/pnas.1913678117. Research.
- Gustafson, A., & Rice, R. E. (2020). A review of the effects of uncertainty in public science communication. Public Understanding of Science, 29(6), 614–633. doi:10.1177/0963662520942122. Research (systematic review).
- Editors' Code of Practice Committee / IPSO, Editors' Code of Practice, Clause 1 Accuracy. Code. https://editorscode.org.uk/the-code/ ; https://www.ipso.co.uk/editors-code-of-practice/
- Financial Conduct Authority, "FCA introduces tough new rules for marketing cryptoassets" (8 June 2023). Regulation (press release). https://www.fca.org.uk/news/press-releases/fca-introduces-tough-new-rules-marketing-cryptoassets
- US Securities and Exchange Commission, Press Release 2024-73, "Terraform and Kwon to Pay $4.5 Billion Following Fraud Verdict" (13 June 2024). Official statement (regulator's press release). https://www.sec.gov/newsroom/press-releases/2024-73
