Research indicates that hate speech, online or offline, can significantly affect the individuals and groups it targets. Such harms may include psychological distress, trauma, and fear, which can discourage people from speaking freely and contribute to self-censorship. Studies have also identified an association between online hate speech and hate crimes. Evidence of these harms does not, in itself, establish that prohibiting hate speech is an effective or proportionate response, particularly given the cost to freedom of expression. Hate speech restrictions are open to misuse as a way of suppressing dissenting or unpopular viewpoints, and have been invoked against the very minority and vulnerable groups such regulation is intended to protect. Legal and policy responses have increasingly emphasized content removal, stronger platform moderation, and automated enforcement. Yet these approaches carry significant free speech costs. Regulatory pressure can encourage precautionary over-removal, while automated moderation continues to struggle with context, irony, dialect, and culturally specific forms of expression.
Counterspeech offers another way of responding, one that can challenge harmful expression while leaving room for public contestation. Its potential also reaches beyond persuading the person responsible for the original message. Counterspeech can influence those watching the exchange, including the so-called “movable middle,” support those targeted by hostility, and undercut the impression that hateful or false claims are widely accepted.
The evidence, however, is mixed. At The Future of Free Speech, we have explored this in practice, training more than 400 people worldwide and translating our counterspeech toolkits on hate speech and disinformation into more than 11 languages. This October, we will continue this conversation with Mina Dennert at our Global Free Speech Summit.
This piece examines the promise of counterspeech alongside its limits. Rather than asking whether counterspeech works in the abstract, it asks a more useful question: what forms of counterspeech work, for whom, in which contexts, and toward what ends?
What Do We Mean by Counterspeech?
Counterspeech encompasses responses to hateful or harmful expression that seek to challenge its influence through further expression. The Dangerous Speech Project describes it as “any direct response to hateful or harmful speech which seeks to undermine it.” Its methods include factual correction, principled disagreement, empathy, humor, alternative narratives, amplification, and expressions of solidarity. Counterspeech can arise spontaneously from individual users or through organized and coordinated campaigns.
Benesch and colleagues identify four patterns of interaction: one-to-one, one-to-many, many-to-one, and many-to-many. Their research identifies strategies ranging from correcting factual claims and pointing out hypocrisy to denouncing hateful or dangerous expression.
Buerger’s research found that most counterspeakers she interviewed regarded spectators as their principal audience. Many considered attempts to change committed hateful speakers a poor use of their time. Their attention instead turned to those reading the discussion, who often considerably outnumber those actively participating in it—the “movable middle” among them, people whose views are neither extreme nor entrenched and who may therefore be receptive to competing arguments.
Effectiveness therefore has to be assessed against the purpose of the intervention. A response may leave the original speaker unmoved while correcting false information for those following the discussion, encouraging others to intervene, challenging an apparent consensus, or supporting those targeted by hostility.
Does Counterspeech Work?
Studies have examined different platforms, audiences, and forms of intervention, using different measures of success. One example comes from research by Miškolci, Kováčová and Rigová on anti-Roma speech on Facebook in Slovakia. The study examined 60 Facebook discussions containing more than 7,500 comments and tested interventions based on factual information and personal experience. The interventions did little to alter the behavior of those expressing anti-Roma attitudes. They did, however, draw other participants with pro-Roma views into the discussion. The significance lies less in converting those expressing hostility than in changing who participates and which perspectives become visible.
Audience composition also matters. Schieb and Preuss, using a computational simulation model, found that two factors mattered the most: the relative size of the groups involved, and the influence counterspeakers could exert on undecided participants.
Scale presents a further difficulty. Individual dialogue is poorly suited to coordinated harassment, organized propaganda campaigns, automated accounts, or networks whose participants have no interest in genuine exchange. Counterspeakers confronting organized disinformation have adapted their objectives accordingly. The Lithuanian Elves, for example, do not attempt to persuade coordinated trolls associated with Russian disinformation operations. Their efforts focus on ordinary readers, placing corrective information into the discussion so that those encountering propaganda can recognize it as false.
How counterspeech is delivered also affects the response it receives. Bartlett and Krasodomski-Jones, examining content that challenged extremist and populist right-wing material on Facebook, found that form and tone mattered. Posts framed as questions generated particularly high levels of interaction, and humorous and satirical content also drew strong engagement. Their research called for more constructive counterspeech and more specific arguments. Frenett and Dow reached a related conclusion in their pilot study of one-to-one online interventions. Antagonistic messages failed to generate responses, while a more casual, sentimental, and personalized approach was considerably more successful in initiating and sustaining engagement.
Large-scale research provides further evidence of the potential of organized counterspeech. Garland and colleagues analyzed 131,366 political conversations on German Twitter over four years, examining interactions involving the far-right network Reconquista Germanica and the organized counterspeech group Reconquista Internet. Following the emergence of Reconquista Internet, the relative frequency of counterspeech increased while hate speech decreased, and organized counterspeech appeared better able to steer subsequent conversations toward further counterspeech. The authors avoid causal claims but conclude that organized counterspeech may help curb hateful rhetoric and produce a more balanced public discussion. The paper was subsequently corrected to revise several dataset figures, including the number of conversations analysed, without changing its substantive conclusions.
There are costs for the people doing the responding as well. Counterspeech often relies on individuals volunteering their time and exposing themselves to hostile online environments. Research with young people has identified uncertainty about what to say, concern that intervention could make the situation worse, and anxiety about the reactions of others as barriers to speaking up. Collective initiatives can alleviate some of this hesitation, as the examples below suggest.
The limits extend beyond the online conversation itself. Counterspeech cannot address structural discrimination, power asymmetries, or offline violence on its own, nor can it reliably alter the behavior of committed ideologues or provide an adequate response to every form of coordinated harassment. These problems may require legal, institutional, technological, or social responses beyond speech.
Taken together, the research suggests that effectiveness depends heavily on what an intervention is trying to achieve. Persuading a committed speaker sets a particularly demanding standard. Influencing undecided readers, encouraging supportive voices to participate, shifting the tone of a discussion, or keeping hateful expression from going uncontested are different measures of success. The evidence also points toward the importance of tone, messenger, audience, and organization.
From Evidence to Practice
Real-world examples show how varied these interventions can be. They range from thousands of people coordinating responses in a Facebook thread, to sustained conversations between individuals, to campaigns that deliberately take online hatred into public spaces.
One of the most developed examples of collective counterspeech is #iamhere, founded by Swedish journalist Mina Dennert in 2016. Dennert began responding to xenophobic comments on Facebook and soon recruited others to join her. The initiative developed into an international network in which participants collectively enter comment threads containing hate or misinformation, post factual and civil responses, and support constructive comments already present in the discussion. Members follow rules that emphasize factual accuracy, respectful language, and avoiding personal attacks. They are also encouraged to amplify constructive contributions rather than repeatedly engaging with hateful comments in ways that could increase their visibility.
The collective dimension serves another purpose. Buerger found that counterspeakers often felt hesitant or vulnerable when responding to hostility alone, and that acting as part of a group gave some members greater confidence to participate and reduced the isolation.
A very different example comes from Megan Phelps-Roper, who grew up in the Westboro Baptist Church and promoted its messages online. On Twitter she came into sustained contact with people who challenged her beliefs. Some responded with insults; others engaged through argument and conversation. Particularly significant were people who questioned Westboro’s interpretation of scripture from within a religious framework that she recognized. More personal exchanges gradually complicated her understanding of the people she had been taught to condemn. She left the church in 2012 and has subsequently described how these interactions led her to question its teachings.
The example is unusual and should not be taken as evidence that respectful engagement routinely transforms deeply entrenched beliefs. It illustrates the importance of engagement that the recipient regards as credible, and shows how counterspeech can work over time rather than through a single corrective message.
Brazil’s “Mirrors of Racism” campaign used amplification in a much more public way. In 2015, after journalist Maria Júlia Coutinho was subjected to racist abuse online, the Black women’s civil rights organization Criola partnered with advertising agency W3haus to reproduce racist social media comments on large billboards in the neighborhoods where those who posted them lived. The billboards carried the message, “Virtual racism, real consequences.” Criola’s Lúcia Xavier explained that the strategy sought to take racism out of the online environment and expose it in the real world.
Humor offers another strategy. German journalist Hasnain Kazim, after receiving sustained xenophobic and anti-Muslim abuse, began responding to many of the messages sent to him, often through deliberate exaggeration. For example, when asked whether he ate pork, he replied that he ate only elephant and camel. Kazim has reported that some correspondents reconsidered their language, explained what lay behind their anger, or apologized. He has also described the considerable emotional cost of responding repeatedly to abuse, eventually abandoning the attempt to answer every message.
The Capacity to Speak Back
Counterspeech deserves a larger place in debates about online hate speech and disinformation. The evidence shows its potential to influence wider audiences, encourage bystanders to participate, support those targeted by hostility, and, in some cases, change the tenor of online discussions. Its effectiveness, however, varies considerably according to context, audience, messenger, tone, and objective.
Its limits should remain part of the conversation. Some speakers have little interest in dialogue; some interventions risk amplifying the content they challenge; and coordinated disinformation, harassment, and structural discrimination require responses beyond counterspeech.
For too long, debates about harmful online expression have concentrated on what should be removed, who should remove it, and how quickly. The harder question is how to strengthen the ability of citizens and communities to challenge what they encounter.
A resilient culture of free speech depends on protecting our ability to speak and strengthening our capacity to speak back.
Natalie Alkiviadou is a Senior Research Fellow at The Future of Free Speech. Her research interests lie in the freedom of expression, the far-right, hate speech, hate crime, and non-discrimination.
This is very interesting! Does your research at all extend into other forms of counter-expression e.g. shame, ostracism, and even stronger social consequences (disassociation, termination, etc.)?
While we certainly must continue protecting speech from state censorship, don't we need self-limiting principles on our "free speech culture" so the most extreme, hateful, bigoted, dehumanizing and eliminationist speech does not appear to be and then become socially acceptable within polite society and our mainstream politics?
Should certain speech and viewpoints remain taboo to openly express? What happens if they don't?