Dawn Images

It started with a passing comment about a cricket match. There was nothing political or provocative about it, instead it was the simple sharing of an opinion about a match. But within hours, the responses began to pile up. These commenters were not disagreeing or debating with the woman who had posted the comment. They responded with sexualised threats that were detailed and graphic, and they would not stop.

The comments weren’t just insults, they were written in a way that would linger on in her mind and make her think several times before ever posting again.

Another woman, a rights activist, described how her photographs were lifted, edited, and circulated with explicit captions. A third woman spoke about being warned not to attend public events because “women like her” would be dealt with there. While the language of the threats keeps changing, the core message to these women stays the same — if you don’t step back, you will be pushed back.

These accounts are taken from the #ThinkBeforeYouSpeak campaign run by Pakistan’s Uks Research Centre in partnership with the World Association for Christian Communication (WACC). Over the course of a year, the project, titled SafeWords, Safer Worlds, collected testimonies, tracked abusive language online, and tried to map out how misogyny works in Pakistan’s digital spaces. What emerged was that online digital abuse of women is not a series of isolated incidents and is instead a phenomenon with a consistent pattern.

For over two decades, Uks Research Centre has worked on gender and media in Pakistan. With the WACC’s support, this project allowed Uks to go from documenting women’s representation to documenting abusive language used against women.

One of the outputs of the project is a compendium of gendered abusive terms in Urdu and Punjabi. This is not a simple glossary in the conventional sense — each entry includes the meaning of the slur, the usage of the word, and, most importantly, variations in spelling, including Roman Urdu and distorted forms that are commonly used online. This level of detail is important and to understand that it is necessary to talk about the research methodology.

During the research, Uks used Meltwater, a media monitoring and analytics tool usually used by corporations to track the sentiment around their products and services. In Uks’ case, the tool was used to track how abusive terms actually appear on social media. What was found was that the same word can appear in dozens of forms. One word is easily altered using symbols to replace letters, by removing vowels, or by stretching spellings to evade social media filters. So, in addition to being widespread, the abuse is also adaptive.

The compendium lists words and also shows how they mutate in real time. This makes it a practical resource for anyone trying to build or improve moderation systems in South Asian languages, especially on social media. The findings are depressing, as the majority of insults commonly directed at women fall into very clear categories. There are accusations of sexual immorality, there are references to incest that target female family members, there is a reduction of women to their body parts, and there are attacks on ‘family background’.

This is certainly not a coincidence. It is reflective of a society in which a woman’s identity is tied to her sexuality and her family’s honour. The online abuse of women is simply reflective of society’s hierarchies in the digital space.

Beyond data collection, the project also collected anecdotal evidence through a social media campaign and focus group discussions across 15 cities in Pakistan. The project gathered direct testimonies from women and also some men about their online experiences. Unfortunately, the responses were consistent, almost predictable.

By and large, women described how they routinely modified their behaviour. “Apart from having safety concerns in the real world, the virtual world is also a cruel place,” one journalist said. “Women who express their strong opinions online are frequently subjected to hate comments… They end up limiting their exposure.”

In practice, limiting exposure, meant sharing fewer posts, avoiding posting pictures, self-censoring while posting opinions, and avoiding risks.

Another respondent spoke for many other women when she explained how this works. “I feel I have to think two or three times… ‘what will people think’.”

Other respondents also spoke about the cumulative impact of abuse. In many cases, the trolling is not about one message — it often snowballs into dozens of comments, sometimes even hundreds. “This has become an unfortunate reality as combatting trolls online is an exhausting process that most women… avoid in a bid to protect their sanity,” said one respondent. As a result of this exhaustion many women disengage and start withdrawing from online conversations.

There is also a link between online abuse and offline violence. Many respondents talked about doxxing, stalking, and the morphing of images. Several women pointed to the growing threat posed by AI tools that can generate fake images and audio clips.

In Pakistan, where an individual’s reputation is often linked to family honour, the risk is not a theory. As one participant recalled, morphed images have led to extreme consequences, which include violence within families.

It also seems that women in public roles, such as journalists, activists, and politicians, face even more intense targeting. “Women journalists are especially targeted… to gag their voice,” one respondent noted.

Importantly, some male participants also talked about the imbalance. One man noted that women’s opinions online are taken “less seriously” and receive more condescending responses.

One of the goals of the SafeWords, Safer Worlds project was to move beyond awareness and engage institutions responsible for regulating digital spaces. The Federal Investigation Agency (FIA), which handles cybercrime complaints, did not engage meaningfully with the project, despite repeated attempts by the Uks team. This was reflective of the pattern that was reported by users about the complaint process being slow, difficult to navigate and generally discouraging.

On the other hand, the Pakistan Telecommunication Authority (PTA) was more responsive. The PTA participated in discussions, reviewed the compendium, and issued a public service message on Women’s Day acknowledging online violence against women.

Most global social media platforms rely on moderation systems trained primarily on English-language data. As a result, abusive content in Urdu, Punjabi and, by default, Hindi, usually goes undetected, especially when it is deliberately misspelled.

The compendium directly addresses this gap and provides a dataset that platforms can integrate into their moderation systems. At present, there is no requirement for them to do so.

The project also encountered a challenge stemming from the ban of X in Pakistan during the campaign period. The ban restricted data collection on a platform where a lot of the abuse occurs and made it difficult to track engagement trends consistently.

The SafeWord, Safer World project generated significant public engagement and the #ThinkBeforeYouSpeak campaign reached several users across Facebook and Instagram. It also produced 1o podcast episodes featuring journalists, activists, and digital rights experts, and conducted focus group discussions in several Pakistani cities, including smaller urban centres where such conversations are less common. More importantly, it created a research-based compendium of abusive terms in Urdu and Punjabi, which is grounded in actual usage data.

It is easy and lazy to think of this as a problem of “online behaviour”. The fact is that the language that was documented in the compendium reflects offline realities. It draws on cultural ideas about shame and honour and it is not the internet that has created these attitudes. As a result, many of us exist in a digital environment where participation is imbalanced, where women can speak, but the cost of speaking is high. The evidence from the project makes that clear.

The project’s findings give rise to several questions. Will social media platforms incorporate local-language data into their moderation systems? Will regulatory bodies move beyond paying lip service to gendered abuse in the digital space to actual enforcement of the law? Will reporting mechanisms be simplified and made accessible?

Or will the burden continue to fall on women to block abusers and report them? Or to withdraw from online spaces completely?

There are no easy answers, but the SafeWords, Safer Worlds project does offer detailed evidence that is difficult to ignore. What we now need is accountability.

This article has been produced in collaboration with Uks