# When journalism is turned off: Preliminary findings on the effects of Meta's news ban in Canada
Sara Parker, Saewon Park, Zeynep Pehlivan, Alexei Abrahams, Mika Desblancs, Taylor Owen, Jennie Phillips, and Aengus Bridgman
All authors of this preprint are members of The Media Ecosystem Observatory, an interdisciplinary collaboration between McGill University and the University of Toronto PEARL that studies information ecosystem health. This report relies on digital trace data collected by the McGill team as well as survey data collected by the University of Toronto team.
April 2024
## Executive Summary
This document presents preliminary findings of an ongoing research project on the effects of Meta's decision to end news availability on its platforms for Canadian users. It examines the impact of this "news ban" on Facebook, specifically focusing on how it affected engagement with Canadian news outlets on Facebook and behaviour within politically-oriented Canadian Facebook Groups.
### Our key findings:
- Even six months after the ban, a large number of Canadians (approximately 33%) still say they use Meta's flagship social media platforms Facebook or Instagram for access to Canadian political and current affairs information.
- The Facebook Pages of national news outlets lost approximately 64% of their Facebook engagement following the end of news availability for Canadian users. Local news outlets lost approximately 85%. Almost half of all local news outlets stopped posting on Facebook entirely in the four months following the ban.
- Engagement with politically relevant pages and groups has remained unchanged since the ban, suggesting politically-oriented users have not reduced their Facebook usage.
- Members of politically-oriented Facebook Groups have circumvented the ban by posting screenshots of Canadian news articles. Although there are fewer screenshots of news post-ban than there were links to news articles pre-ban, the total engagement with news content in these Groups has remained consistent.
Overall, we observe that the end of news availability on Facebook has resulted in a significant decrease in engagement for Canadian news outlets, but has not substantially changed the behavior of Canadian Facebook users.
## Introduction
In August 2023, the news disappeared from Facebook and Instagram in Canada. In July, the Canadian government passed the Online News Act, which requires "large digital intermediaries" to enter into negotiation with Canadian news outlets to compensate them for posting news content on their platforms.1 Similar to, but evolving considerably from, the Australian News Media Bargaining Code, the Canadian Online News Act aims to address a market imbalance in the digital ad market that has affected the new industry by obliging companies like Google and Meta to enter into negotiation with news publishers for licensing agreements to their content. While Google negotiated an exemption agreement with the Canadian government through the creation of a \$100 million per year fund for Canadian journalism, Meta disputed the Act entirely, claiming that it was based on a "fundamentally flawed premise" and maintaining that they can only reasonably comply with the Online News Act by ending "the availability of news content in Canada."2 On August 9, 2023, they did.
This study is part of an ongoing assessment of the effects of Meta's decision to end news availability for Canadian users. We refer to this event as 'the ban', as it effectively banned Canadian users from seeing, sharing, or engaging with news content on Meta's platforms.The following analysis focuses on how the ban affected the visibility and dissemination of Canadian news content on Facebook, with specific attention paid to engagement with Canadian news outlet Facebook Pages and activity on Facebook Groups focused on Canadian politics. In particular, we are able to assess how important Canadian news was for Facebook: if it was, we should see substantial shifts in engagement with news content and activity within Facebook Groups who previously shared it frequently.
We expected the disappearance of news on Meta platforms to have caused a major shock to the Canadian information ecosystem – that is, the diverse relationships between content producers (e.g. news outlets), information consumers (Canadians), and the means through which information is shared and seen (e.g. social media platforms). Facebook and Instagram are key parts of the Canadian information ecosystem, being the first and third most popular platforms Canadians say they use to get the news. Figure 1 shows data from a nationally representative survey sample of Canadians (n = 2127, conducted March 7-10, 2024) taken six months after the ban went into effect.3 Surprisingly, 33% of Canadians still use Facebook (25%) and/or Instagram (12%) to get news about politics and current events several times a month, even though they cannot see content from news outlets on either platform.
1 House of Commons, "An Act respecting online communications platforms that make news content available to persons in Canada." Statues of Canada 2023, June 22 2023.
https://www.parl.ca/DocumentViewer/en/44-1/bill/C-18/royal-assent
2 Clegg, Nick, "Meta's Position on Canada's Online News Act", Meta Canada Newsroom, May 8 2023. https://about.fb.com/news/2023/05/metas-position-on-canadas-online-news-act/
3 The survey was conducted by the University of Toronto PEARL lab under the leadership of Peter Loewen and Shane Littrell.
Figure 1: Social media apps used by Canadians to get news about politics and current affairs - Percentage by platform.
This remarkable stability in Meta platform users' continued use of the platforms for politics and current affairs anticipates the findings from the detailed investigation into engagement and posting behaviour of Canadians. We find that the ban has significantly impacted Canadian news outlets but had little impact on Canadian user behaviour. Consistent with the ban's goal, we find a precipitous decline in engagement with Canadian news and consequently the posting of news content by Canadian news outlets. The effect is particularly acute for local news outlets, while some news outlets with national or international scope have been able to make a partial recovery after a few months. Additionally, posting by and engagement with alternative sources of information about Canadian current affairs appears unmoved by the ban. We further find that Groups focused on Canadian politics enjoy the same frequency of posting and diversity of engagement after the ban as before. While link sharing declines, we document a complementary uptick in the sharing of screenshots of Canadian news in a sample of these political Groups, and confirm by reviewing a number of such posts where users deliberately circumvented the link-sharing ban by posting screenshots. Although the screenshots do not compensate for the total loss of link sharing, the screenshots themselves garner the same total amount of engagement as news links previously had.
Taken together, these findings suggest that, on Facebook, Meta has deprived users of the affordance of news sharing without suffering any loss in engagement of their user base. Users continue to discuss Canadian politics and their discourse continues to be informed by Canadian news, in part by their ability to circumvent the ban through workarounds. These workarounds could put Meta back into scope of C-18. Meanwhile, Meta's ban depletes the click-through rate to Canadian news websites, depriving outlets of web traffic at a time when the profitability of Canadian news is already waning. Our findings highlight the impact of the ban on Canadian news outlets, as well as underscore the significant role digital platforms like Facebook play in mediating access to information and shaping political conversation.
## Data and methodology
For this analysis we collected Facebook data from June 1, 2023 to December 31, 2023, pre- and post-dating the start of the ban. Since we are primarily interested in how the ban affected engagement with Canadian news outlets, we collected posts from Canadian news outlets as well as from Pages and Groups who frequently shared and engaged with Canadian news.
For our main data analysis, we identified 987 Canadian news outlet Facebook Pages: 60 national-level news outlets and 794 local-level news outlets. We began by reviewing Media Cloud's "national" and "state-local" dataset of Canadian news organizations, and removed all inactive or non-Canadian outlets to make our own dataset of Canadian news outlets. We then visited each news outlet URL, searched the website for a "facebook.com" domain, and recorded the full Facebook link. These links were manually reviewed and validated. We then manually searched for the Facebook Pages of the news outlets that did not have a facebook.com domain linked on their home page.
We further collected data on non-news Facebook Pages that frequently post about Canadian current events. We compiled 183 Pages of Canadian federal and provincial politicians, political and advocacy organizations, commentators, and content creators using two methods. First, we examined "facebook.com" links that were shared within Canadian politics-oriented Facebook Groups. In Facebook post data, a "facebook.com" link indicates that the post is a reshared post from another user from outside the Group. We reviewed the Pages that were linked to the most frequently, and manually verified if they were a) Canadian and b) primarily posted about Canadian current events. Second, we reviewed the Pages that appeared when searching #canpoli or #canadianpoli on Facebook, and added those that were Canadian and primarily posted about Canadian current events.
Finally, we wanted to know how Canadian users responded to the ban in general, so we analyzed a dataset of 589 politics and local community Groups. We used Meta's Crowdtangle API to identify public Groups that had shared a link to a Canadian news site more than ten times between June 1 and August 1, 2023. This yielded a dataset of 914 Facebook Groups. We collected all the descriptions of the groups and then classified each Group using zero-shot classification into one of the following categories: politics, local community, heritage, support group, and leisure interest. We isolated and manually verified only the Groups labeled as "politics" or "local community" to ensure that they were Canadian and correctly labeled.
We collected all posts from these 987 news outlets, 183 non-news Pages, and 589 groups from June 1 to December 31, 2023 using Meta's Crowdtangle API.
### Methodology
We rely on a pre-post means comparison for our analysis, focusing on post volume, engagement rates, and frequency of different types of post, then comparing their daily or weekly averages before and after the ban. Specifically, we examine the overall posting behaviour and engagement trends for Canadian news outlets and "influencer" Pages and within political and community Groups. Additionally, in a sample of Groups, we identify the total number of posters per day, the frequency of links to known misinformation sites, and the frequency of screenshots of Canadian news articles before and after the ban.
## Results
### Posts by Canadian news outlets
We look at the overall posting behavior and engagement trends of Canadian news outlets' Facebook Pages to cheek a) if news outlets are continuing to post, despite being blocked from Canadian users/their primary audience and b) if Canadian news outlets may still be generating engagement from sharing content on Facebook.
*Research Question 1: Did Canadian news outlets reduce their posting frequency after the ban?*
*RQ2: Did engagement with Canadian news outlet Facebook Pages drop significantly following the ban?*
Figure 2: Total number of news outlet posts on Facebook - Daily by news outlet category.
Figure 2 shows the daily total posting volume of Canadian news outlets. We observe that Canadian local news outlets have posted less than half as often on Facebook after the ban (47% of pre-ban volume), while national news outlets have lost over a third of their pre-ban content volume (63% of pre-ban volume). Local news outlets have abandoned their Facebook Pages *en masse*: before the end of September 2023, almost half of local news outlets had stopped posting on Facebook entirely.
Figure 3: Total number of reactions, comments, and shares on posts by Canadian news outlets - Daily by news outlet category.
Figure 3 shows total engagement with Canadian news outlets, again separated by local versus national outlets. Notably, national news outlets have recovered 36% of their Facebook engagement since the ban, while local news outlets continue to receive only 15% of their original engagement. Additional analysis is necessary to identify the circumstances in which Canadian news outlets continue to be engaged with. We observe that some news outlets' publish content relevant to non-Canadian audiences, particularly by posting entertainment or sports-related news. We have also identified a small number of news outlets that were omitted from the ban and are not blocked to Canadian users, thereby allowing their Pages to remain visible.
### Posts by influencers
After the disappearance of news from Canadian Facebook feeds, some users – particularly those who rely on Facebook for their news – may have been motivated to engage more with other types of Pages that share content related to Canadian current events and politics, such as politicians, commentators, and content creators. We refer to this group of non-news-outlet Pages as "influencers", as they represent some of the sources and creators we believe that Canadian users would turn to if they actively sought out information and discussion about Canadian politics on Facebook after they could no longer view content from news outlets. We examined these influencers' overall posting and engagement trends before and after the ban to assess if the ban on August 9 had any effect on a) the behavior of the influencers themselves and b) engagement with their Pages.
*RQ3: Did Canadian political "influencers" begin posting more frequently after the ban because they expected Canadian users to seek out alternate sources of information and discussion of current events?*
*RQ4: Did the Facebook Pages of Canadian political "influencers" receive more engagement after the ban?*
Figure 4: Total number of influencer posts on Facebook - Weekly
Figure 4 shows the total number of influencer posts per week before and after the ban. Although we see variance in the volume of posts per week, it does not appear to be correlated to the ban. We again see a significant drop in posts at the beginning of October 2023; further analysis is needed to identify the cause of this decline and the subsequent peak in posts at the beginning of November.
Figure 5: Total number of reactions on influencer posts – Weekly
Figure 5 shows the weekly total engagement received by posts by influencers. There is high daily variance in engagement with the influencer pages, but the ban had no discernable impact on influencer engagement. Instead of the inverse relationship we predicted, engagement with influencers appears distinct from engagement with news.
While the ban's null effect on influencers may be due to a lack of interest from Canadian Facebook users in current events, there are many other factors that could be complicating the relationship between the disappearance of news content and engagement with other sources of news and information. Canadians may be turning to influencers focused on other countries for their political content, which are not included in our analysis. Facebook itself may also be suppressing the visibility of news or political content, consistent with their algorithm's prioritization of posts from friends and family. 4 In other words, these findings do not necessarily suggest that Canadians are no longer seeking out political information or news on Facebook. Rather, more research is needed to determine how Canadians engage with news on Facebook outside of interaction with news outlets themselves.
Media Ecosystem Observatory 9
4 Zuckerberg, Mark. Facebook post: "One of our big focus areas for 2018…" January 11, 2018. https://www.facebook.com/zuck/posts/10104413015393571
### Group activity
In a country as geographically diverse as Canada, online political communities are particularly important as a means to not only associate but engage directly with people with similar views. For a rural NDP supporter in Alberta, for instance, it is much easier to turn to a pro-Albertan NDP Facebook Group to discuss policy proposals than drive ten hours to Edmonton or Calgary to attend the occasional party supporters mixer. Online groups also allow constituents across the country to connect with one another on political issues – a connection that is rarely fostered by hyper-regional Canadian politics.
We turn to these Groups to better understand how Canadians engage with news on Facebook. In total, we analyze over 1.5 million posts from June to December 2023 from 244 "political" and 345 "local community" Groups. These Groups were all identified as frequent sharers of Canadian news content, but we distinguish between political and community Groups because we expect the engagement with news to differ between, for example, heated debates about Justin Trudeau's latest scandal and (equally heated) debates about expected construction on the local Main Street.
First, we measure the volume of daily posts and average reactions to determine if these Groups saw a lower level of activity after the ban. If we assume conversation about national and local politics is largely centered around or inspired by current events, we expected that the level of activity in these Groups would decline after the ban as users were no longer able to share news links about which conversation could occur.
*RQ5: Did the level of activity in Canadian political and local community Groups decline after the ban?*
Figure 6: Total number of posts in Canadian Facebook Groups - Daily by Group category.
Figure 7: Average number of reactions per post in Groups - Daily by Group category.
Figure 6 shows the total number of posts per day in our Facebook Groups, while Figure 7 shows the average reactions per post. Surprisingly, in both political and community Groups, there was no change in post volume or engagement after the ban. Either news was not all that important in these Groups to begin with, or users within the Groups identified ways to circumvent the ban and discuss news content. Evidently, Canadians still have something to talk about in these Groups, regardless of whether or not they learn about it on Facebook or from other sources of news.
#### Diversity of posters
To verify that the majority of users in Facebook Groups have indeed remained active after the ban, we looked at the diversity of posters in Groups before and after the ban. Prior to beginning this project, we spent some time reviewing the sample of 60 Groups (40 political and 20 community) to get a better idea of what we expected to see in our large-scale analysis. During this review, we noticed that many of the posts in some of these Groups came from 1 or 2 dedicated posters who often posted multiple times a day, compared to more "casual" posters who only posted occasionally. We suspected that, perhaps in some Groups, many users had left Facebook or stopped posting in political/local community Groups in response to the ban, causing some highly motivated/dedicated users to supplement the loss by posting more. If this happened, then it would have obscured the decline in post volume. We therefore examine the total number of posters per day to check if Group members are operating as usual, or if a handful of highly motivated users started posting frequently after the ban, in line with our observation.
*RQ6: Following the ban, did the number of members per Group who post content decline?* .
We first collected the URLs of all posts from our sample of 60 Facebook groups posted between June 1st 2023 and December 31st 2023 using Facebook's Crowdtangle API. This resulted in a total of 125,050 post links. For each URL, we then scraped the poster's name, ID, and profile link using a custom scraper. For each Group, we counted the number of unique posters per day and simply summed the number of unique posters for each Group per day.
Figure 8: Average number of unique posters in sample of Facebook Groups - Weekly by Group category.
Figure 8 shows the average number of unique posters per week in our sample of Facebook Groups. We see no significant change in the number of posters in these Groups over time, adding to the growing body of evidence that Canadian Facebook users continued to engage in discussion on the platform, even after they were blocked from sharing or seeing news content.
#### Misinformation links
Seeing as politically-engaged Canadian Facebook users remain active on the platform, we were curious if they had begun filling the vacuum left by news content with less reliable content. We therefore examine the frequency of sharing misinformation in Facebook Groups.
*RQ7: In the absence of official news sources on Facebook, did Group members begin to share more content from sources of misinformation after the ban?*
Automatically identifying misinformation, disinformation, and misleading content is an exceptionally challenging technical task that is yet to be mastered by social media platforms themselves, so we did not expect to succeed in identifying all false content in Facebook Groups. Instead, we relied upon a dataset of known misinformation links that were frequently used among Canadian social media users.5 We pulled all the posts from June 1 to December 31, 2023 from our dataset of 244 political Facebook Groups, then extracted all the domains of the links shared from those posts. We then matched our list of misinformation domains to the extracted domains, and calculated the percentage of misinformation links among all links shared before and after the ban – the assumption being that, if we saw a rise in links to misinformation sites, that would likely be a sign that other forms of misinformation were being shared as well.
Surprisingly, the number of misinformation links in political and local community Groups *decreased* after the ban.
#### Political Groups:
- Prior to the ban: 2.8% of links (5612 out of 198,587 links) were misinformation links
- After the ban: 1.4% of links (5306 out of 379,202 links) were misinformation links
This may be because of a 'blanket' reaction to the ban from Facebook users. The proportion of misinformation links among all links has decreased because, potentially, users recognize that they cannot share links to news outlets and think about these sources of misinformation as news. As seen in the following section, the volume of posts in political Groups containing any kind of link decreases within two months of the ban, so users may have begun to turn away from sharing information about current events and politics via links and are sharing other types of content instead.
### Sharing news without links
When examining ways Facebook users were circumventing the ban, we noticed that some users were posting screenshots of news articles in political and local community Groups. We wanted to know if this screenshot method was being used across Canadian Facebook Groups, particularly
5 This list was generously provided by the Global Disinformation Index (GDI).
in those that frequently shared Canadian news stories prior to the ban. The screenshot method, while not generating more engagement or viewership for news outlets themselves, allows for news stories to be distributed and discussed among Canadian Facebook users
*RQ8: In the absence of other means of sharing Canadian news on Facebook, did highly motivated users in Facebook Groups circumvent the ban by sharing screenshots of Canadian news stories?*
We downloaded all the shared images over six months (from June 1 to December 31, 2023) from our sample of 60 Groups, yielding a total of 45,033 images. We employed a three-layer classification process to identify news screenshots from these images. Each layer consists of a filtering using a Contrastive Language–Image Pre-training (CLIP) algorithm for image classification. CLIP leverages a large-scale dataset of images and textual descriptions to train a model that understands and interprets visual concepts in a manner akin to natural language. Unlike traditional image classification models that rely on pre-defined labels and extensive manual annotation, CLIP learns visual concepts from natural language descriptions. When given a set of input classes with descriptive labels, CLIP can accurately match images to these classes by finding similarities between the image content and the textual descriptions. CLIP performs well in tasks where it hasn't been explicitly trained, as long as the input classes and descriptions are well-defined and descriptive.
First, we used CLIP to identify images related to Canadian news, political figures, and current events (n = 31,361 or 70% of shared images). Next, of those images related to Canadian news, we used CLIP to identify screenshots that display news articles or headlines from media sources (n = 9,374 or 21% of shared images). From those screenshots, we used CLIP to identify seven distinct categories: news, news articles, posters, advertisements, announcements, entertainment, and a catch-all group for any images that did not fit the previous classifications.6
We observe a dramatic increase in posts containing screenshots of Canadian news stories in the post-ban period. Before the ban, only approximately 19 posts per week featured a screenshot of Canadian news. After the ban, the number of posts with Canadian news screenshots triples to an average of 68 posts per week. This increase in news screenshots after the ban does not correlate with an overall increase in non-news screenshot images, suggesting that the increase in news screenshots is indeed a result of the news ban. News screenshots are prevalent in our sample of political Groups as seen in Figure 9, which shows the volume of posts with Canadian news links and posts with screenshots of Canadian news.
6 Although we considered eliminating the second step to streamline the process, we found that doing so significantly compromised the accuracy of our results. This intermediate step was crucial for effectively distinguishing between the nuanced types of content within our dataset, thereby enhancing the overall precision of our classification approach. This filtration was essential for maintaining the relevance and quality of our dataset.
Figure 9: Total number of posts with a photo or link from sample of 40 Political Groups - Weekly by post type.
In the post-ban period there were approximately 192 fewer posts per week that linked to Canadian news stories. Despite screenshot posts being much less likely to occur than news articles, the total engagement with content directly engaging with Canadian news content (linking in the pre-ban period or screenshotting in the post-ban period) has remained steady. Screenshots of news stories in political Groups, although fewer in number, have matched engagement previously generated by news links.
Figure 10: Total reactions on posts with Canadian news links vs. Canadian news screenshots in sample of political Facebook Groups – Weekly by post type.
We observe a similar, but smaller, trend in local community Groups. However, as the vast majority of posts in these Groups are not related to news at all, we do not include it in this analysis.
Although our sample size of Groups is modest (only 40 political Groups), these results are promising for the state of Canadian political discourse on Facebook. Users who had previously engaged with and discussed news content on Facebook are evidently still interested in seeing news on the platform, and therefore continue to share and like news content when possible.
## Discussion
We find little evidence that Facebook usage has been impacted by the ban. After the ban took effect, the collapse of Canadian news content production and engagement on Facebook did not appear to substantially affect users themselves. The Facebook Pages of Canadian political influencers experienced neither an uptick nor a downturn in engagement. Facebook Groups dedicated to discussing Canadian politics did not experience an exodus of members, nor a decline in frequency of posting. And while news link sharing became impossible, users quickly adapted by posting screenshots of news articles. Meta's ban on news sharing in Canada does not appear to have had an adverse effect on the number and attentiveness of active Facebook users in Canada.
However, the ban undoubtedly had a major impact on Canadian news. Local news outlets have been particularly affected by the ban: while large, national news outlets were less reliant on Facebook for visibility and able to recoup some of their Facebook engagement regardless, hundreds of local news outlets have left the platform entirely, effectively gutting the visibility of local news content. In fact, the loss of local news on Facebook has already had profound consequences for Canadians. Soon after the ban, wildfires began raging in northern Canada. In previous years, local news organizations had relied on Facebook to share essential information about the spread of forest fires and safety orders; in 2023, they found themselves blocked.7 Although government accounts were still visible to Canadian users, they proved to be insufficient in countering misinformation and confusion and effectively delivering information to residents of small, rural communities.8
The loss of news content notwithstanding, online political communities have demonstrated remarkable resilience in the wake of the news ban. Not only have members of political Groups adopted alternative strategies to share news (e.g. screenshots), but they have remained exceptionally engaged with news content. Although less news content is shared, the engagement with each post containing news content is significantly higher than before the ban, indicating a strong attachment to news on Facebook among a specific community of users. This response to the news ban is seen as an example of inertia in online political communities. Despite the introduction of a major change to the type of content they are able to share, members of political Facebook Groups have continued to share the prohibited content regardless, while maintaining the same volume of political conversation. Politically engaged users continue to engage in the same behavior, but have simply adapted their strategies to changes in their environment. We expect to see a similar inertia in other online political communities across the social media platforms we study.
## Future work
Control group and statistical treatment
Currently we employ Facebook influencers as a control group against which to evaluate user behaviour on Facebook. This is an imperfect control because engagement to their content could be subject to either a substitution effect (discussed above) or to a decline given users. Those two forces could be acting simultaneously, thus obscuring changes in behaviour among Canadian
8 Ibid. 7 Lindeman, Tracey. "Meta's News Block Causes Chaos as Canada Burns." WIRED Magazine. 28 August 2023. https://www.wired.com/story/meta-facebook-instagram-news-block-canada-wildfire/
users. A large and perhaps international control group would allow us to better evaluate why national news outlets experienced such a major decline from August to September, and then began recovering by late October. Additionally, while we do not observe major changes in user behavior besides the adoption of screenshots as a way of sharing news, we would like to run similar tests on a sample of non-Canadian political Facebook Groups and non-Canadian political influencers to ensure the validity of our study.
### Other platforms
We also intend to explore the dynamics of news dissemination and engagement on other platforms. News content on Instagram is also subject to Meta's decision to end news availability on its platforms. The decline of political discussion may be more dramatic on that platform given a) the lack of groups or defined communities and b) the higher production costs associated with the multimedia (i.e., visual and audio) content. To fully understand the impact of Meta's decision on news organizations and political conversation in Canada, we will extend this analysis to Instagram.
We also hope to look at the role of news on non-Meta platforms. Although Facebook is the platform most employed by Canadians to stay informed, YouTube, Twitter, and TikTok follow closely behind in usage for news. We will analyze the role that news plays in Canadian political communities on these platforms, as well as if there were any changes in engagement, post volume, or behavior in relation to news following Meta's news ban. In other words, we would like to gain a more comprehensive idea of the value of news to online spaces in a post-ban world.
## Bibliography
- Clegg, Nick. "Meta's Position on Canada's Online News Act." Meta (blog), May 8, 2023. https://about.fb.com/news/2023/05/metas-position-on-canadas-online-news-act/.
- Government Bill (House of Commons) C-18 (44-1) Royal Assent Online News Act Parliament of Canada, Pub. L. No. C–18 (2023).
- https://www.parl.ca/DocumentViewer/en/44-1/bill/C-18/royal-assent.
- Lindeman, Tracey. "Meta's News Block Causes Chaos as Canada Burns." Wired. Accessed April 2, 2024.
- https://www.wired.com/story/meta-facebook-instagram-news-block-canada-wildfire/.
- Radford, Alec, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, et al. "Learning Transferable Visual Models From Natural Language Supervision." In Proceedings of the 38th International Conference on Machine Learning, 8748–63. PMLR, 2021. https://proceedings.mlr.press/v139/radford21a.html.
- Zuckerberg, Mark. Facebook post: "One of our big focus areas for 2018…" January 11, 2018. https://www.facebook.com/zuck/posts/10104413015393571