Every major platform says it wants a "safe" or "positive" community. Almost none design for it. Engagement-ranked feeds reward whatever provokes a reaction, and outrage, mockery, and cruelty provoke reactions more reliably than kindness does. The result is a two-decade natural experiment: what happens to human behavior when a billion people are handed a megaphone, a mostly anonymous crowd, and an algorithm that rewards conflict?
The data on the answer is no longer thin. Independent watchdog groups, academic researchers, and the platforms' own transparency reports now give a fairly clear picture of where kindness survives online and where it doesn't. This post makes the case for why that matters, then scores the ten largest platforms — Instagram, TikTok, Facebook, X, YouTube, Reddit, LinkedIn, Snapchat, Pinterest, and Discord — on their actual ability to preserve kindness, using real user-reported data, academic research, and independent safety audits rather than each company's marketing copy.
Why kindness is a platform metric, not just a values statement
Three things happen when a platform lets cruelty go unchecked, and all three are measurable.
People leave, or they leave part of themselves at the door. Pew Research's most recent survey of teen platform use found that roughly three-quarters of teenagers on TikTok, Instagram, and Snapchat see harassment and bullying as a real problem for people their age, and a meaningful share have experienced it directly — about 30% on Snapchat and roughly 20% each on Instagram and TikTok reported at least one form of harassment, from name-calling to threats (Pew Research Center, 2026). Harassment doesn't just hurt the person targeted; it teaches everyone watching to self-censor, which quietly drains the authenticity that makes a platform worth using in the first place.
Unkindness compounds at scale in ways moderation struggles to catch. A 2025 study in Scientific Reports that tracked 123 million Reddit comments over 14 years found something counterintuitive: before the mid-2010s, users tended to get less toxic the longer they stayed on the platform, as community norms socialized them. After roughly 2013–2015, that pattern reversed — the longer high-activity users stuck around, the more toxic they became (Yin et al., Scientific Reports, 2025). Whatever shifted in platform design and incentives over the last decade, it stopped rewarding the behavior that used to keep people civil.
Cruelty is now a governance and business risk, not just a user-experience problem. GLAAD's 2026 Social Media Safety Index — the sixth annual, independently audited scorecard of how major platforms handle hate and harassment — found that every platform it tracks except one saw its safety score drop to a historic low, largely because companies rolled back moderation policies and hate-speech protections that had been in place for years (GLAAD Social Media Safety Index, 2026). Advertisers, regulators, and users are all watching those scores now, which means kindness has quietly become a line item.
None of this means every platform needs to feel like a greeting card. It means the platforms that treat kindness as core infrastructure — not a content policy nobody reads — are the ones more likely to keep people, and to keep them well.
Methodology: how we scored each platform
Rather than inventing a number from vibes, each platform below gets a Kindness Preservation Score (1–100), built as a composite of four inputs, weighted roughly evenly:
- Independent safety audits — primarily GLAAD's Social Media Safety Index, which grades platforms on published hate-and-harassment policy strength and enforcement transparency (available for Instagram, Facebook, Threads, TikTok, YouTube, and X).
- Direct user-reported experience — survey data from Pew Research, TBD Marketing, and platform transparency reports on how often people say they were harassed, bullied, or made to feel unsafe.
- Independent academic and investigative research — peer-reviewed toxicity studies and journalistic ranking work that isn't funded by the platform being studied.
- Design and moderation structure — whether the platform's core mechanics (algorithmic amplification vs. curated discovery, public quoting vs. private messaging, centralized vs. community moderation) tend to reward or dampen unkind behavior.
Where a platform lacks an independent audit (Reddit, LinkedIn, Snapchat, Pinterest, Discord), the score leans more heavily on academic research, transparency-report data, and cross-platform comparative studies. This is an editorial synthesis, not an official industry metric — no single number can capture something as contextual as "kindness" — but every figure cited below is a real, sourced statistic rather than an estimate.
The Kindness Preservation Score: 10 platforms ranked
| Rank | Platform | Score (1–100) | What the data shows |
|---|---|---|---|
| 1 | 78 | Least outrage-driven design; no reply/quote culture | |
| 2 | 58 | Lowest ambient toxicity, but a real DM harassment problem | |
| 3 | TikTok | 54 | Best-in-class safety policy on paper, still algorithm-driven |
| 4 | 51 | Bimodal — deeply kind and deeply toxic, by subreddit | |
| 5 | Snapchat | 49 | Highest teen-reported harassment rate of the ephemeral apps |
| 6 | Discord | 47 | Culture varies entirely by server; enforcement scale is large |
| 7 | 43 | Comparison culture plus a declining safety score | |
| 8 | YouTube | 41 | Steepest safety-score drop of any platform in 2026 |
| 9 | 37 | Largest population, highest absolute abuse volume | |
| 10 | X (Twitter) | 22 | Lowest independent safety score; toxicity is the design |
1. Pinterest — 78
Pinterest is the outlier because its core mechanics don't reward conflict. There's no reply-chain, no quote-and-dunk, and the platform has restricted political and health misinformation more aggressively than most peers. Pinterest's own research, run with outside partners including UC Berkeley, Nielsen, and Morning Consult, found the platform "more likely than its competitors to always leave its users feeling positive after spending time" there (Pinterest, "A More Positive Internet? It's Possible."). That research is company-commissioned, so it deserves a discount for bias — but it's also the only platform on this list whose business model doesn't depend on maximizing time spent arguing, which independently supports the finding.
2. LinkedIn — 58
LinkedIn's reputation as the "civil" platform is only half true. Cross-platform comparative research consistently ranks it as having the lowest ambient toxicity, with most negativity limited to workplace one-upmanship and spam rather than harassment (VisionMedia Interactive, 2025). But a 2024 survey by TBD Marketing found 65% of respondents had received harassing, negative, or troll messages on LinkedIn, and 44% had experienced unwanted sexual comments or advances via DM or comment — concentrated heavily against women (TBD Marketing, LinkedIn Bullying & Harassment Survey, 2024). That survey's sample was small and skewed toward legal professionals, so it shouldn't be read as universal — but it's a real signal that LinkedIn's public feed being calm doesn't mean its private messages are.
3. TikTok — 54
TikTok scored highest of any platform GLAAD evaluated for hate-and-harassment protections, and is one of only two major platforms with explicit policies against both targeted misgendering and deadnaming, alongside a ban on promoting conversion therapy content (GLAAD Social Media Safety Index, TikTok, 2025). On the ground, Pew found about 20% of teen users report direct harassment, lower than Snapchat, with roughly seven in ten describing their overall experience as mostly positive (Pew Research Center, 2026). The discount here is structural: TikTok's algorithm optimizes purely for watch time, and separate research has tied heavy use to worse self-reported mental health outcomes — a different axis from kindness, but one that shapes how vulnerable users experience the platform.
4. Reddit — 51
Reddit is the hardest platform to score with one number because it isn't really one platform — it's thousands of independently moderated communities with wildly different norms. Comparative research describes it as having "mixed outcomes," with genuinely supportive communities sitting alongside subreddits that foster open hostility, particularly around politics and culture-war topics (VisionMedia Interactive, 2025). The large-scale Scientific Reports study of 123 million Reddit comments found that, unlike Wikipedia, Reddit's most active users trend toward more toxic behavior over time, not less (Yin et al., 2025) — a sign that Reddit's karma-and-visibility incentives can reward exactly the wrong behavior in its highest-engagement corners.
5. Snapchat — 49
Snapchat's ephemeral, mostly private messaging design should, in theory, reduce public pile-ons — and it does limit them. But it also removes the paper trail that makes harassment easier to report and enforce against. That trade-off shows up in the numbers: Pew found roughly 30% of teen Snapchat users report experiencing at least one form of harassment, the highest of the three platforms it studied, even though about seven in ten still describe their overall experience as positive and 44% say the app actively helps their friendships (Pew Research Center, 2026).
6. Discord — 47
Discord doesn't have an independent third-party safety audit on the scale of GLAAD's, but its own transparency report is unusually granular. In the first half of 2024 alone, Discord took enforcement action against 92,295 distinct accounts for harassment and bullying and another 67,969 for hateful conduct, disabling tens of thousands of accounts and removing thousands of servers (Discord Transparency Report, January–June 2024). That scale cuts both ways: it shows a real, persistent harassment problem, but also a platform actively enforcing against it rather than ignoring it. Because Discord is organized around individually owned servers, the actual kindness of your experience depends almost entirely on which servers you're in — a well-run community server can feel dramatically safer than the platform average.
7. Instagram — 43
Instagram's GLAAD safety score fell to 41 out of 100 in the 2026 index, down four points from the year before, part of a broader retreat by parent company Meta from LGBTQ-inclusive hate-speech protections (GLAAD Social Media Safety Index, 2026). Pew's teen data puts direct harassment at around 20%, comparable to TikTok, but Instagram's core design — public comment threads under highly curated, appearance-driven content — creates more opportunity for comparison and cruelty than Snapchat's largely private messaging model, even at similar harassment rates.
8. YouTube — 41
YouTube posted the steepest decline of any platform in the 2026 Social Media Safety Index, dropping 11 points to a score of 30 out of 100, which GLAAD attributed to the removal of protective language from its community guidelines (GLAAD Social Media Safety Index, 2026). We score it a few points above that raw figure because YouTube's comment sections, while historically some of the internet's most toxic spaces, sit alongside an enormous long-tail of small, tightly knit creator communities where moderation and tone are set locally — closer in spirit to Discord or Reddit than to a single monolithic feed.
9. Facebook — 37
Facebook combines the largest user base on this list with some of the worst documented abuse. Its GLAAD score sits at 40 out of 100, down five points year over year (GLAAD Social Media Safety Index, 2026). A 2025 Global Witness survey specifically of climate defenders found 62% had experienced online abuse on Facebook, the highest rate among mainstream platforms in that study (cited in VisionMedia Interactive, 2025) — a population under unusual strain, so that figure shouldn't be read as representative of the average user's experience, but it illustrates how bad things can get for people already being targeted. Facebook's group structure means, like Discord and Reddit, individual experience varies enormously — but its scale means both the best and worst corners are larger than almost anywhere else.
10. X (Twitter) — 22
X scores lowest on every independent measure we found. GLAAD's 2026 index put its overall safety score at 29 out of 100 — the lowest of any platform evaluated — citing a weakened misgendering and deadnaming policy and a process that effectively requires harassment targets to self-report before the platform will act (GLAAD Social Media Safety Index, X, 2025; 2026 update). Comparative toxicity research consistently ranks X as the most hostile of the major platforms, and one study specifically found that even good-faith corrections through X's Community Notes feature tend to trigger more anger, disgust, and outrage in the replies that follow, rather than defusing conflict (cited in VisionMedia Interactive, 2025). X's core design — public, quotable, reply-chain-driven, and optimized for velocity — is arguably the closest thing on this list to cruelty-as-a-feature.
One platform you won't find in the table above is Kai. That's deliberate: Kai isn't a public feed competing for outrage-driven attention, it's community space built around safe, judgement-free expression.
It doesn't have the years of independent safety audits, transparency reports, and academic study that made the other ten scoreable with real data.
What it does have is a design philosophy that runs in the opposite direction from most of this list — safety checks that screen for crisis language before any response is generated, no algorithmic feed rewarding conflict, and no incentive to keep people scrolling through content that makes them feel worse.
It's early, and it would be dishonest to hand it a number next to platforms with a decade of published research behind them. But it's worth naming as the kind of thing this whole post is arguing for: a platform where kindness is a design requirement from day one, not a policy rolled back once growth targets take over.
What this actually tells us
Two patterns hold up across every source cited here. First, design predicts kindness better than stated values do. Platforms built around public reply chains and outrage-optimized ranking (X, Facebook, and to a lesser extent Instagram and YouTube) score worse than platforms built around curation or private messaging (Pinterest, LinkedIn, Snapchat), almost regardless of what each company's community guidelines say on paper. Second, stated policy must be more than window dressing — GLAAD's finding that platform scores fell across the board specifically because companies rolled back existing protections shows that when platforms choose to weaken moderation, harassment measurably increases. Kindness online isn't an accident of good user behavior; it's downstream of specific, reversible decisions about what a platform amplifies and what it's willing to enforce against.
For any platform — including new ones — that's the actual lesson: kindness has to be built into the mechanics, not bolted on as a policy page nobody clicks.
Methodology note: Kindness Preservation Scores are an editorial synthesis by the author, built from GLAAD's Social Media Safety Index, Pew Research Center survey data, peer-reviewed academic research, platform transparency reports, and independent comparative journalism. They are not an official industry metric, and figures drawn from small or self-selected surveys (noted inline) should be read as directional rather than universally representative.