×
Methodology
Silenced maps publicly reported incidents in Pakistan where online speech (social media posts, tweets,
videos, WhatsApp messages, podcasts and vlogs) led to real-world consequences. This page explains how
reports are collected, filtered and checked, and what the database cannot tell you.
1. Scope
The database covers incidents reported from January 2016 onwards, the year the Prevention of
Electronic Crimes Act (PECA) came into force. It tracks:
arrests, detentions and First Information Reports (FIRs);
mob violence and lynchings;
custodial torture;
enforced disappearances and abductions;
deaths, death sentences and other human rights violations
An incident is included only when the report links it to something the person said or shared online.
2. Sources
Automated scrapers read the archives of these outlets back to 2016, then check for new stories every
month:
Pakistani English press: Dawn, The Express Tribune, The News International, Samaa TV, Geo News, Dunya NewsPakistani Urdu press: Daily Jang, Daily ExpressInternational: BBC Urdu, Al Jazeera, The Guardian, Human Rights Watch, Amnesty International
International outlets are included because local media face censorship and self-censorship on these
subjects. Some incidents are reported only from outside Pakistan. The scrapers respect each site's
robots.txt and limit how often they make requests. How far back each archive goes varies by outlet.
3. Filtering
Every article goes through automated, rule-based keyword filtering in English and Urdu. An article is
kept only if it mentions all of the following:
a digital-speech context: social media, Facebook, X/Twitter, WhatsApp, TikTok, YouTube, a post,
tweet, video or podcast;a law or allegation: PECA, blasphemy (s.295/298 PPC), "digital terrorism", "anti-state
narrative", sedition, "fake news", the Anti-Terrorism Act and similar; anda consequence: an arrest, FIR, detention, torture, mob attack, disappearance or death.
The incident type, category, location and any law sections cited are then extracted by the same rules.
No generative AI is used to write or summarise entries; each summary quotes the source article's own
sentences: the one describing the online post and the one describing the consequence. Where no single
sentence ties the two together, the summary reads "Source reports [consequence] linked to online
activity:" followed by the article's sentence about the online activity.
4. Verification
Every source link is checked automatically. An incident is published only if its link returns a
successful HTTP response (status 200 or another 2xx code, after following redirects).
Links that return an error (such as 403, 404, 410 or a server error) are marked Broken Link and
hidden from the map, but they are not deleted . They stay in the database and are checked again
every month. If a link starts working again, the incident reappears automatically. Temporary failures
such as timeouts are simply retried on the next run.
"Verified" means the source link works. It does not mean the allegation has been fact-checked.
5. Categories
Religious Blasphemy (s.295/298 PPC), Ahmadi and minority cases, desecration, sectarian content.
Political Treason, sedition, "anti-state" speech, criticism of the military or government, journalists, political parties.
Social Harassment, blackmail, "obscenity", modesty and morality cases, gender-related cases.
Classification is keyword-based and approximate.
6. Limitations
Reported incidents only. The database captures incidents that reached the news, not every
incident. Many cases, especially outside major cities or involving minorities, are never reported.
Year-to-year changes can reflect media attention as much as real trends.Allegations are not verified. Inclusion means a credible outlet reported the incident. It does
not mean the underlying allegation is true or that anyone is guilty.Automated extraction makes mistakes. Keyword rules can miss relevant stories, include
irrelevant ones, or misread a location or category. When several outlets report the same incident
(same city and type of consequence, within two days, with closely matching text) it is shown once,
with every outlet's report linked. Matching is automatic, so an incident can still occasionally
appear twice, or two similar incidents be combined.Archive coverage varies. Some outlets' online archives do not reach back to 2016.
7. Shutdown map
The Shutdown map (button in the header) is a separate record of internet blackouts, mobile and
cellular service suspensions, and blocks on platforms such as X (Twitter), TikTok, YouTube or WhatsApp
in Pakistan. It is built from the same news sources plus two specialist trackers:
NetBlocks reports and Access Now's #KeepItOn shutdown dataset and statements.
An article counts as a shutdown when its headline or opening sentences report one in Pakistan, in
English or Urdu (for example "mobile internet suspended", "internet blackout", "X blocked in Pakistan",
"انٹرنیٹ سروس معطل"). Reports about other countries, including Indian-held Kashmir, and outages caused by
power cuts or cable faults are left out. Articles that already count as speech incidents stay on the
main map, so the two maps never share an entry.
The shaded area shows how far the shutdown reached, as the report describes it: the whole country, a
province, or a circle around a city. Darker shading means more shutdowns were reported in that area.
Duration, the services cut and the reason are read from the report's own wording; "Not stated" means
the report did not say. When a report names cities in three or more provinces, it is shown as
nationwide.
8. Offline Violence map
The Offline Violence map (button in the header, or /#violence) records physical security
incidents in Pakistan reported by the same news sources: bomb blasts (including suicide attacks, IEDs,
grenade and rocket attacks), militant attacks (on checkposts, police stations, convoys and polio teams,
ambushes, kidnappings), targeted killings, and security operations (intelligence-based operations,
CTD raids, encounters, search and clearing operations, arrests of militants and explosives recovered).
An article counts when its headline reports violence and its headline or opening sentences use one of
these terms, in English or Urdu (for example "suicide attack", "IBO", "targeted killing", "خودکش حملہ",
"سیکیورٹی فورسز کی کارروائی"), about a place in Pakistan. Accidents (gas cylinder or factory explosions), court
cases, anniversaries, statistics round-ups, opinion pieces and events in other countries are left out.
Words that also describe ordinary crime ("explosion", "police raid", "encounter") count only when the
report is about terrorism, and killings the report ties to robberies or personal disputes are left out.
Articles that already count as speech incidents or shutdowns stay on those maps, so no entry is on two maps.
Red pins are attacks (circle: blast, triangle: militant attack, diamond: targeted killing); slate squares
are security operations. The type comes from the headline first. Fatalities are the deaths the report
itself states in its opening sentences, counting everyone killed (attackers included); later reports often
give a higher toll, so treat it as a minimum. Militant group lists the groups the report names:
naming a group is not proof it was responsible. The state's term "Fitna al-Khawarij" is counted as the TTP;
"Fitna al-Hindustan", which the state uses for Baloch separatist groups as a whole, is kept as its own label.
9. Gov Takedowns
The Gov Takedowns view (button in the header) shows, for every six months, how many requests
Pakistani authorities (courts, the PTA, ministries, police) sent to four platforms to remove or block
content, and what the platforms did. The figures come from each platform's own transparency report,
downloaded as published:
Meta (Facebook and Instagram): content restrictions based on local law. Meta publishes only the
number of items it restricted in Pakistan, not the number of requests or a compliance rate.Google (all products, including YouTube): government removal requests, items named, and items
removed. Up to 2018 the compliance rate is the share of requests Google fully or partly complied with;
from 2019 it is the share of named items removed.TikTok : government removal requests, the content and accounts named, and what TikTok restricted
in Pakistan for local law or removed under its Community Guidelines. Figures in this format start in
the first half of 2022.X (Twitter): legal demands to remove content, accounts named, and the share of demands where X
withheld content or acted under its own rules. X has published no figures for Pakistan since the
second half of 2021: its later reports name only a few large countries.
Platforms count differently (a request can name one post or thousands of accounts), so compare a
platform with itself over time rather than platforms with each other. A missing figure means the platform
did not publish it. These numbers are not checked independently.
Below the charts, What the PTA told courts and Parliament gives the other side: the PTA's own
count of links it reported to each platform and how many were blocked, as Pakistani news reports quote
its court reports and Senate or National Assembly briefings. Text-matching rules (not AI) read these
figures from reports that name the PTA and a court or Parliament, and each row shows the sentence it
came from. The PTA's blocking rate is not the same measure as a platform's compliance rate, and its
totals usually run from when it began counting. Newspapers sometimes misprint the PTA's numbers, so
check the quote and its source. Only reports whose link still works are shown.
Always read the original source before relying on an entry. To report an error, email
radicalismreport@gmail.com .
×
About Silenced
Mission
Silenced is an independent, non-partisan monitoring platform that documents the real-world consequences
of online speech in Pakistan. We bring scattered news reports of arrests, FIRs, violence, torture, and
enforced disappearances together in one searchable, mappable record. By documenting these cases, Silenced
aims to highlight the shrinking space for digital dissent and the chilling effect these prosecutions have
on public discourse, allowing journalists, researchers, lawyers, and the public to see patterns that single
news stories cannot show.
Academic Guidance
This portal is guided by a PhD graduate of Concordia University in Montréal. Originally from Pakistan, he
has close knowledge of the local context. His doctoral research examines how online platforms,
particularly social media, are misused in political and religious affairs in Pakistan, and how this usage
has transformed the country's religio-political and social life.
Data & Transparency
We believe in open access to information. The data mapped here is scraped monthly from local and
international sources and is available for download directly from our platform for independent research
and journalistic use.
Disclaimer
Inclusion in this database does not verify the underlying allegations or prove guilt. Every entry is a
pointer to a published news report; the original source remains the final authority. Silenced is a neutral
observer and does not endorse any political party or religious group. Please note that this database is
limited to what is reported by the media; incidents in areas with limited press freedom may be
underrepresented.
Contact
Email radicalismreport@gmail.com to:
Request a correction.
Report a broken or incorrect link.
Submit a missing incident (please include a link to a published report).
How to Cite
If you use this data in your research, reporting, or publications, please cite this platform as:
Silenced (2026). Tracking Consequences of Online Speech in Pakistan. Retrieved from
https://silenced.observer
Silenced