Panel Orientation: Platform News Products

Alphabet: Google News

Interface Orientation: [✓]Desktop [✓]Mobile
Content Recommendation: [✓]Algorithmic [*]Human (Special Coverage)
Content Guidelines: [✓]Exists
Customized COVID-19 Policy:[✓]Present

Background. Google News primarily uses algorithms to select and rank news content. These algorithms choose what all users see in the Top Stories, Headlines, and Full Coverage sections, and they personalize content for each user in sections such as For You.

Members of the Google News staff are able to add temporary topics for special circumstances, like events or elections. They also choose news topics and publications to highlight in the Newsstand tab of the Google News App.

According to the platform, the essential purpose of the product is “to help everyone understand the world by connecting people with high quality news from a variety of perspectives.”

Google News states it aims to “help you make sense of the news” by:

• “Using technology to connect you to information”
• “Providing access to context and multiple perspectives”
• “Elevating trustworthy information” 
• “Fighting misrepresentative and deceptive practice”

Google News also says it aims to “help sustain a healthy news ecosystem” by:

• “Building and strengthening audiences for high-quality news”;
• “Collaborating with the news industry to drive innovation.”

Ranking Factors & Process

Google News offers a number of general news categories that are available for every user. The content available in Top Stories is chosen algorithmically and only varies depending on language and regional settings. 

Content is also personalized in the For You sections based on user-specified interest about topics, location, and publishers. Depending on account settings, Google News can suggest content based on user activity. 

Google News describes its ranking process as one in which its algorithms weigh a host of factors. It identifies its primary factors as:

• Relevance
• Interests
• Location
• Prominence
• Authoritativeness
• Freshness
• Usability

Ranking in Google News is determined by the following factors: 

• Relevance (Relevance to search terms)
• Location (Google uses location to help find content relevant to a user’s area, such as Local stories)
• Prominence (Noteworthy news events as signaled by trending articles or stories, as well as coverage being prominently featured by publishers)
• Authoritativeness (Authoritativeness signals are designed to help surface pages that “demonstrate expertise, authoritativeness and trustworthiness on a given topic”)
• Freshness (Freshness refers to when the article was published and when it is more useful to have up-to-date content over an older article)

In addition to the factors above, the For You tab is ranked based on: 

• Usability (usability assesses page load times, how easy it is to view content on the site, if the site is formatted for different devices, etc.)
• Interests (depending on account settings, individual interests may help determine results in sections such as Discover or For You in Google News)

Google News emphasizes that it “elevates news from authoritative sources” and that it “requires publishers to be transparent and accountable in order to be represented in news results” in accordance with its content policies. Google News also states that it “uses technology to sort massive amounts of content to connect you with important, relevant, and useful news.” [Emphasis added]

Content Labels. Google News uses labels to help classify and sort information. Publishers are advised to apply labels to their entire site as well as different parts of their site. Labels are also algorithmically applied to a publisher’s site.

The labels that publishers can apply themselves are: 

• Opinion
• Satire
• User-generated
• Press release
• Blog

Information found in this section is adapted from the following documents: How Google News stories are selected, Building news experiences, Help you make sense of the news, Surfacing useful and relevant content, Presenting news in helpful ways, Ranking within Google News, Organizing news from around the web, What does each label mean?

External Audits. While Google News may continually be refining its approach and results, external audits of Google News and Google Search have taken place in recent years.

• One 2017 academic study did not find support for the filter-bubble hypothesis, but did find a “general bias in that Google News over-represents certain news outlets and under-represents other, highly frequented, news outlets.” https://doi.org/10.1080/21670811.2017.1338145
• Another academic study in 2019 found similarly that recommendations were rather homogeneous but that 69% of all recommendations were to five news organizations. https://doi.org/10.1016/j.chb.2018.07.043
• A study of Google News in 2018 (published in 2019) by the Economist of results based from Kansas found no evidence of overall ideological bias; however, personalizations such as search history or certain keywords are likely to have an impact. The implication instead was that “Google’s main form of favouritism is to boost viral articles.” https://www.economist.com/graphic-detail/2019/06/08/google-rewards-reputable-reporting-not-left-wing-politics

Apple News

Interface Orientation: [✓]Desktop [✓]Mobile
Content Recommendation: [✓]Algorithmic [✓]Human
Content Guidelines: [✓]Exists
Customized COVID-19 Policy: [✓]Present

Background. Apple News is an algorithmically and human-determined ranking and recommendation system for news content. It can be accessed via desktop app at apple.com/apple-news or as the “Apple News” mobile app. The homepage presents Top Stories in the Today section. Scrolling through Apple’s Today section, past Top Stories, takes users past a number of other sections (e.g., Politics, Business) that either populate by default or that the user selected to follow. Each section displays previews for a certain number of stories as the user scrolls past.

At the bottom of the app, there are three main sections: Today, News+ (which allows users to dive deeper into certain publications), Audio, and Following (which allows users to follow certain topics or beats).

Content on the service is also curated by topic. A set number of sections show up by default; other sections allow users to find news in areas they have explicitly selected to follow in the app (See the Apple News+ description).

Ranking Factors & Process

Apple editorial staff chooses five stories to be featured in the Top Stories section of Apple News. The team reportedly selects from a pool of between 100 and 200 pitches per day. The Apple editorial team also curates a selection of featured stories. 

Algorithms are used to select content for the Trending Stories section. News content is also personalized using algorithms when users follow, unfollow, or block categories. 

Information found in this section is adapted from the following documents: Apple News, View news stories chosen just for you on iPhone, Auditing News Curation Systems: A Case Study Examining Algorithmic and Editorial Logic in Apple News, Apple News’s Radical Approach: Humans Over Machines.

Facebook News (News Tab)

Interface Orientation: [*]Limited Desktop [✓]Mobile
Content Recommendation: [✓]Algorithmic [✓]Human
Content Guidelines: [✓]Exists
Customized COVID-19 Policy: [✓]Present

Background. Facebook News is a news product that is available for both desktop and mobile users. A blend of human and algorithmic curation is implemented to rank and recommend content to users. The platform uses machine learning models to personalize a users’ news experience, choosing content based on diverse factors including previous engagement. In addition, Facebook’s human curation team chooses content based on publicly editorial guidelines. According to their public-facing materials, the platform news product aims to deliver content that is “informative, reliable and relevant to your community”.  This feature is different from Facebook’s News Feed which ranks and recommends activity from the pages, people and groups that a user follows.

News Sections currently include: 

• Home
• New For You
• COVID 19
• Your Local News
• Science & Tech
• Health 
• Entertainment 
• Business
• Sports

Selection of News Pages and Publishers. To be included as a news source on Facebook news, publishers must meet Facebook’s News Page Index Registration Guidelines which includes: 

• Being active during the past 90 days
• Status as a news provider with a multi-person editorial staff
• Reporting on current events or timely information
• Citations or links to factual sources
• Date and time stamps on published content
• Pages that primarily publish content that is neither aggregated nor user-generated
• Provides transparent information about writers or editors

Publishers must also meet integrity standards which look at factors such as:

• Misinformation
• Community standards violations (e.g., hate speech)
• Clickbait, engagement bait, and scraped content

Facebook divides publishers into four main categories: 

• General News 
• Topical News
• Diverse News 
• Local News

Ranking & Recommendation

Facebook News prioritizes original reporting, diverse perspectives, and relevant stories. The news platform product only includes publishers who register and meet the aforementioned news page index registration requirements. A machine learning model identifies news content using some of the same guidelines from the news page index. 

Algorithmically Curated Content. Facebook News uses machine learning algorithms to personalize the news experience for each user. The ML model evaluates articles based on thousands of “signals” (including publisher location, byline, topic, previous interaction etc.) and the chosen articles are given a score using a predictive algorithm similar to the one that ranks content in the news feed. The score is intended to measure the likelihood that a reader will care about the content, ultimately upranking or downranking content according to the score. Users can further personalize the content they receive by hiding articles and  blocking publishers.  

To personalize Facebook News to include local news, users can set a primary location in the settings function.

Human-Curated Content. A team of journalists curates content to give an overview of the day’s news and the biggest headlines of the day.

According to Facebook, topics are chosen based on: 

• Prevalence
• Interest
• Impact
• Diversity

They state that they prioritize stories based on: 

• Facts
• Diverse voices
• Original reporting
• On-the-record sourcing
• Timeliness
• Depth and context
• Fairness
• Local reporting

Information Centers. Human curation teams are also responsible for updating information on the COVID-19 Information Center and the Climate Science Information Center. 

Microsoft: Microsoft News

Interface Orientation: [✓]Desktop [✗]Not Mobile
Content Recommendation: [✓]Algorithmic [✓]Human
Content Guidelines: [✗]No longer exists*
Customized COVID-19 Policy: [✗]Not found

Background. Microsoft News is a human and machine-determined ranking and recommendation system which curates content for MSN.com, The Edge browser start page, the Microsoft News app, and more. It can be accessed via web at www.msn.com/en-us/news or as the “Microsoft News” mobile app.

The Edge start page presents Headlines and content on the service is curated by topic. A set number of sections show up by default; other sections allow users to find news in areas they have explicitly selected to follow in the app; because it allows for individual sign-in, the products may reflect personal interest given their statement that “user choice and personalization are essential to the experience.”  

Rather than a mission statement, one can find an updated Microsoft News belief statement that includes: 

• “A free, well-funded press plays a critical role in society”
• “In delivering unparalleled breadth and depth of high-quality journalism”
• “That trustworthy and diverse perspectives matter”
• “User choice and personalization are essential to the experience”

Microsoft News no longer has a set of defined Editorial Standards to which it refers (2020 version). Instead, its principles can be discovered in a FAQ for Microsoft News.

Ranking Factors & Process

According to its FAQ, “Microsoft News believes in the power of combining human and machine curation.” Microsoft News content is curated by hundreds of editors in combination with algorithms that personalize based on user preferences. Microsoft News publishes content from over 1200 publishers that send “more than 100,000 unique pieces of content” everyday. Then:

“Our AI scans the content as it arrives, processes to understand dimensions like freshness, category, topic type, opinion content and potential popularity and then presents it for our editors. Our algorithms suggest appropriate photos to pair with content to help bring stories to life. Editors then curate the top stories throughout the day, across a variety of topics, so our readers get the latest news from the best sources.”

(from About us)

Microsoft News states that it labels content in an effort to clearly flag to readers when a news article contains advertising, such as sponsored studies or native advertising. According to Microsoft’s FAQ, this content is marked with an Ad label.

Factual Content. If Microsoft discovers there are “factual or any other significant errors impacting the quality of the content, we immediately notify the publisher and request corrections.”

Bias and Diversity. Stating the challenge of balancing conflicting viewpoints, Microsoft News states that they attempt to provide a range of perspectives in their coverage: “we believe that trustworthy and diverse perspectives matter, and we are constantly working to develop content experiences that will allow us to feature all sides of a story.”

Further Automation. In 2020, Microsoft announced that AI would replace the work of Microsoft externally contracted news producers. According to the Seattle Times, the following are examples of the work that the human contractors performed:

• Identifying trending news stories from dozens of publishing partners
• Helping to optimize the content by rewriting headlines
• Adding better accompanying photographs or slide show to stories
• Planning content
• Maintaining the editorial calendars of partner news websites and assigning content to them

Twitter Explore — News

Interface Orientation: [✓]Desktop [✓]Mobile
Content Recommendation: [✓]Algorithmic [*]Limited Human (Twitter Moments, COVID-19)
Content Guidelines: [*]Limited documentation
Customized COVID-19 Policy: [✓]Present

Background. Twitter’s Explore Tab and News Tab is an algorithmically determined ranking and recommendation system for news content. It can be accessed via web at Explore (https://twitter.com/explore) or on the Twitter mobile app. Twitter provides little information about how the Explore tab is driven in general including its News section: “Explore tabs may vary based on your location and settings.”

To reach the news section, the user has to navigate to the Explore section on desktop or the search icon on the Twitter app. The main section that then appears is For You. Then the user can navigate to five other sections: COVID-19, Trending, News, Sports, or Entertainment.  

According to a Twitter help desk technician, News in the What’s Happening section or explore page is driven by an algorithm not only based upon “who you follow, your interests, and your location,” but are also based on trending topics and Twitter Moments. The difference between news and a trending topic however was not clarified (2 July 2021). 

Twitter has a number of policies (“The Twitter Rules”) that sometimes involve human curation and can affect what content appears on its platform.  Some policies of interest include:

• Synthetic and Manipulated Media policy
• COVID-19 misleading information policy
  • Coronavirus: Staying safe and informed on Twitter
• Civic integrity policy
• Defending and respecting the rights of people using our service
• About public interest exceptions on Twitter

We highlight a few elements of these policies further below. In 2020, some safety principles on Twitter highlighted the tension with freedom of expression. The page at this URL no longer exists, with the earlier content absorbed into the above policies. An updated version of an ongoing help desk “Safety and security” page continues to address issues such as hacking, privacy, and abuse.  

Ranking Factors & Process 

Algorithmic Recommendation for Trends. News recommendation is driven by an algorithm on Twitter, and is linked to the way they approach trends. By default, trends are oriented to who a user follows, what interests they have expressed, and where the user is based. It is also possible to see what trends exist outside of personalization; instructions are on the FAQ page.  Rules exist for Trends, as governed by the platform policies. Twitter notes that for trends, “in some cases, we may also consider the newsworthiness of the content, or if it is in the public interest when evaluating potential violations. In these cases, the content might continue to trend on our platform.”

Human Curation with Twitter Moments. Twitter Moments are intended to surface “the best of what’s happening on Twitter.”  Moments can be algorithmically generated or human curated by a team. The team’s goal is to provide “insight and context” on conversations happening on the platform in 5 languages across 16 markets.  

Curators are trained according to a policy that focuses on standards of impartiality, accuracy, corrections, and other content-related standards regarding adult, sensitive, and illegal material:

• Impartiality (“On topics of public debate, we seek to represent a variety of viewpoints whenever feasible.”)
• Accuracy (“Our goal is to highlight quality Tweets that represent accurate information.”)
• Corrections (“If we become aware that we have highlighted inaccurate information, we will update the Moment with a visible correction and issue an updated Tweet. In rare cases, we may delete the Moment and post a retraction using the relevant country Twitter account.”)
• Standards (“Profanity, violence, nudity and other types of potentially sensitive content are avoided except when it’s necessary to highlight a newsworthy conversation”)

Twitter Moment curators are also expected to avoid conflicts of interest. Moments may also be curated by individuals or Twitter partners who must also meet curatorial standards. 

Policies: Misleading Information. According to the platform, “You should be able to find reliable information on Twitter.”  Decisions about adding labels and warnings for “misleading information” such as synthetic tweets and manipulated media — as well as potential removal of those tweets — are assessed using the following questions: 

• “Is the media synthetic or manipulated?” 
• “Is the media shared in a deceptive manner?”
• “Is the content likely to impact public safety or cause serious harm?”

Twitter takes varying actions when tweets fall into these categories: 

In order to make some of these determinations, Twitter may use their own technology or work with third parties. There is little additional information about the process that Twitter uses to identify tweets that fall into these categories.

Policies: COVID-19. According to Twitter’s COVID-19 misinformation policy, users “may not use Twitter’s services to share false or misleading information about COVID-19 which may lead to harm.” Twitter still uses a blend of automated and human processes to address the spread of harmful messages related to COVID-19. Automated processes surface content that is likely to cause harm, and human teams manually review reports of content that need additional context. Twitter has updated their COVID-19 tabwith “credible” information, including “Why vaccines are safe,” and “What COVID-19 vaccine efficacy means.”

External Audits. Because Twitter states that it relies heavily on personalized algorithms for recommendation of Topic and News content, with little guidance on what qualifies as News, it can be helpful to understand how Twitter algorithms perform. 

A recent 2021 study created archetypal sock puppets to understand how Twitter handled news recommendations generally. Among its discoveries: over time, algorithmic recommendation in Twitter increased the overall diversity of sources, but it also tended to increase partisan messaging while decreasing messages with direct links to news articles in feed results.

Information found in this section is adapted from the following documents: Twitter Explore, Visit Explore to see what’s happening The Twitter Rules Safety on Twitter Twitter Moments guidelines and principles, Updating our approach to misleading information, Building rules in public: Our approach to synthetic & manipulated media, COVID-19 Misleading Information Policy,Coronavirus: Staying safe and informed on Twitter, Twitter Trends FAQ. It also contains reference to an email from Twitter Support (ref:_00DA0K0A8._5004w2B8X7A:ref).

Yahoo News

Interface Orientation: [✓]Desktop [✗]No Mobile
Content Recommendation: [✓]Algorithmic [✓]Human
Content Guidelines: [✗]Unable to find
Customized COVID-19 Policy: [✗]Unable to find

Background. Content on Yahoo News is both human curated and algorithmically-determined. It can be accessed at https://news.yahoo.com/. The site has a number of default sections including: 

• News Home
• US
• World
• Politics
• Health
• Science
• Originals

The Originals section includes a dropdown menu with subsections:

• 360
• Skullduggery
• Conspiracyland 

Researchers were unable to find any information related to Yahoo’s ranking and recommendation processes. 

Information found in this section is adapted from the homepage of Yahoo News: https://news.yahoo.com/. NewsQ researchers were unable to find public-facing information about ranking factors and process for the platform to date. We will update this section when possible.