Article: Emotional Tone and User Trust
Reviewed: November 2026 · Method version 2 · How these reviews work
Each factual claim in the article was checked against the source it cites, then challenged by reviewers whose job was to find the strongest evidence against it: replication attempts, meta-analyses, later papers that cite the original work, and research from other fields that study the same mechanism. A separate reviewer weighed both sides, and a script re-checked every quoted source. Where a claim was overstated or misattributed, the article was corrected. The research below supports the article as it now reads.
On Christmas Eve 2014 Eric Meyer wrote that Facebook’s Year in Review, whose default caption was “It’s been a great year! Thanks for being a part of it,” showed him “Eric, here’s what your year looked like!” over a photo of his daughter, who had died that year.
No change needed.
Meyer called it “inadvertent algorithmic cruelty,” did not think anyone meant harm, and noted the code worked for most people.
No change needed.
The copy had been written for one emotional state and delivered to everyone, with no obvious way to make it stop.
No change needed.
Meyer: “The design is for the ideal user, the happy, upbeat, good-life user. It doesn’t take other use cases into account.”
No change needed.
Mailchimp’s style guide says you have the same voice all the time but your tone changes; tone adjusts to the moment, including the emotional state of the person addressed.
No change needed.
In a 2016 Nielsen Norman Group study, 100 participants rated pairs of nearly identical web texts that differed only in tone.
No change needed.
Perceived trustworthiness explained about 52% of the variation in whether people would recommend the organization; friendliness added only about 8% more.
NN/g’s finding that trustworthiness predicts willingness to recommend far more than friendliness fits consumer research showing perceived competence drives purchase intent.
Warmer tones usually did well in the study, but trust carried far more weight, so warmth pays off only when it doesn’t cost credibility.
Experiments outside UX find informality raises perceived goodwill but can lower perceived competence in professional contexts, which supports the article’s caution while NN/g’s data show warmth usually helped.
Anaïs Gretry and colleagues ran three experiments comparing brands posting on social media in an informal or a formal style. Informality raised trust when people already knew the brand and lowered it when they didn’t. The authors’ explanation is social norms: people expect brands to follow them, so casual talk seems appropriate from a familiar brand and inappropriate from an unfamiliar one.
Later work on chatbots and workplace messages supports the idea that informality works best when it fits the relationship, though one replication did not confirm the ‘appropriateness’ explanation and a chatbot replication found no familiarity effect.
Sela and colleagues found that “we” instead of “you and [the brand]” could make people like a real brand more or less depending on whether the implied closeness matched the expected relationship; customers and non-customers reacted differently, and perceived trustworthiness carried the effect.
A 2025 meta-analysis finds pronoun effects are small and depend on context, consistent with Sela’s fit account.
In the NN/g study a playful car-insurance text was rated friendlier than its serious twin but less trustworthy; “the friendliness and irreverence actually undermined users’ perceptions of trustworthiness and professionalism.”
No change needed.
A casual bank text beat its formal version on both friendliness and trustworthiness.
No change needed.
Taken together, the research points to a practical test that reaches beyond social posts, though no study has tested it directly: would this sound right coming from a company the reader met yesterday, on the worst day they could be having?
Service-recovery experiments find formal language from human agents helps most when failures are severe, which supports writing for the reader’s worst day.
Playful register, from a brand she barely knows, about money. Gretry’s finding on social media posts suggests this costs trust, and studies of joking replies to customer complaints point the same way.
Experiments on humorous company replies to complaints found humor lowered perceived trustworthiness, extending Gretry’s social-media finding toward failure moments.
A joke that guesses at her finances. Cards fail for many reasons, often dull ones such as an expired date or a bank’s fraud check, and a stranger has no business guessing which.
Payment processors list expired cards and fraud screening among common decline reasons, alongside insufficient funds.
Mailchimp’s guide says forced humor can be worse than none at all.
No change needed.
In bad-news moments (errors, declines, outages), avoid humor and emoji aimed at the person having the problem and write in the most serious tone the voice allows.
Experiments on company replies to complaints find humor lowers perceived trustworthiness, and humor aimed at the customer does the most damage; gentle humor can work with onlookers.
Anniversaries, memories and year-end reviews need a way to opt out, because the moment is not happy for everyone.
No change needed.
Start warm but plain, and let informality grow with familiarity. Save “we” and other closeness cues for when the relationship can carry them.
Research on pronouns finds that closeness cues like “we” help only when they match the relationship people expect; the effects are small and vary by setting.
NN/g’s video explains the four scales used in the study (humor, formality, respectfulness, enthusiasm); the exercise rates each screen 1 to 5 on those four dimensions.
No change needed.
Tone works as a relationship signal. It tells readers how well a company thinks it knows them, and readers tend to trust companies whose estimate matches their own.
Studies of pronouns, informal brand posts and informal messages all find that implied closeness helps when it fits the relationship and can backfire when it doesn’t.
These edits were made on November 12, 2026 after the review.
Claim reviewed: Warmth helps, but only while it doesn’t cost credibility.
Now reads: Warmer tones usually did well in the study, but trust carried far more weight, so warmth pays off only when it doesn’t cost credibility.
Why: The NN/g study mostly found that warmer tones helped: in three of four pairs the friendlier text was rated as trustworthy or more so, and only the playful insurance text lost trust. Other research does show informality can lower perceived competence in professional settings, so the trade-off is real, but the sentence makes warmth sound riskier than the cited study found. Now cites The double-edged sword of informality, J. Social and Personal Relationships (2025).
Claim reviewed: Gretry and colleagues found informal brand posts raised trust when people already knew the brand and lowered it when they didn’t, because people expect brands to follow social norms, so casual talk seems appropriate only from a familiar brand.
Now reads: Anaïs Gretry and colleagues ran three experiments comparing brands posting on social media in an informal or a formal style. Informality raised trust when people already knew the brand and lowered it when they didn’t. The authors’ explanation is social norms: people expect brands to follow them, so casual talk seems appropriate from a familiar brand and inappropriate from an unfamiliar one.
Why: Gretry and colleagues report exactly this pattern and explanation, from three experiments on brand social media posts. The finding has not been directly replicated, and a large replication of a related workplace-email study did not confirm “appropriateness” as the reason, so the mechanism is best presented as the authors’ explanation rather than settled fact.
Claim reviewed: Taken together, the research suggests a test that applies well beyond social posts: would this sound right coming from a company the reader met yesterday, on the worst day they could be having?
Now reads: Taken together, the research points to a practical test that reaches beyond social posts, though no study has tested it directly: would this sound right coming from a company the reader met yesterday, on the worst day they could be having?
Why: The test is a sensible synthesis: the cited studies show familiarity matters and NN/g advises considering the reader’s emotional state, and service-failure research finds formal language helps human agents more when failures are severe. But no study tested a “worst day” standard, and the Gretry study covered only social media posts, so this is the article’s heuristic rather than a research finding.
Claim reviewed: A playful headline (“Oops! Your card got rejected” with an emoji) from a barely known brand, about money, costs trust, as Gretry’s finding suggests.
Now reads: Playful register, from a brand she barely knows, about money. Gretry’s finding on social media posts suggests this costs trust, and studies of joking replies to customer complaints point the same way.
Why: Gretry’s informal markers were exclamation points and emoji, and the effect on trust reversed for unfamiliar brands, so the link to this headline is reasonable. Their study was about social media posts, not payment errors; studies of joking replies after service failures mostly find a trust cost, though results vary with the kind of humor. Now cites The bright and dark sides of humorous response to online customer complaint, European J. Marketing (2020).
Claim reviewed: Cards fail for dull reasons, such as an expired date or a bank’s fraud check.
Now reads: A joke that guesses at her finances. Cards fail for many reasons, often dull ones such as an expired date or a bank’s fraud check, and a stranger has no business guessing which.
Why: Stripe’s documentation confirms that expiration dates and fraud screening are common reasons for declines. It also lists insufficient funds, so not every decline is dull; the article’s real point, that the company can’t know the reason and shouldn’t guess, still holds. Now cites Stripe documentation, Declines.
Claim reviewed: At first contact, start warm but plain and let informality grow with familiarity; use “you” and the brand name until the relationship is real.
Now reads: Start warm but plain, and let informality grow with familiarity. Save “we” and other closeness cues for when the relationship can carry them.
Why: Letting informality grow with familiarity follows from Gretry’s results. The pronoun advice is a stretch: Sela found that the effect of “we” depends on fit with the relationship, but did not test a rule that “you” plus the brand name is the safe default, and a 2025 meta-analysis finds pronoun effects are small and vary by setting.
Claim reviewed: Tone works as a relationship signal: it tells readers how well a company thinks it knows them, and readers trust the companies whose estimate matches their own.
Now reads: Tone works as a relationship signal. It tells readers how well a company thinks it knows them, and readers tend to trust companies whose estimate matches their own.
Why: This restates Gretry and Sela as a general rule. Both studies, and newer work on chatbots and informal messages, support the idea that tone works best when its implied closeness fits the relationship. But the evidence comes from specific settings, and one replication did not confirm the appropriateness explanation, so “tend to” is more accurate than a flat law.