Â鶹´å

Extending the BEND Framework to Webgraphs

Keywords: Web Graphs, Search Engine Optimization, SEO, BEND, Information Operations

Publication:

Williams, Evan M., Peter Carragher, Kyle Herdrich, Luke Prakarsa, and Kathleen M. Carley. "Extending the BEND Framework to Webgraphs." In International Conference on Social Computing, Behavioral-Cultural Modeling and Prediction and Behavior Representation in Modeling and Simulation, pp. 76-85. Cham: Springer Nature Switzerland, 2025.[

How Manipulated Web Links Can Shape What You See Online

 When people think about online propaganda, they usually picture fake social media accounts pushing misleading posts. But there’s another, quieter tactic that can influence what we see online: manipulating the web of links that search engines use to rank websites.

 A new study shows how groups—including state-linked actors—can shape information ecosystems by strategically building and manipulating website networks. And it offers a way for researchers, journalists, and policymakers to measure that manipulation. 

Why Web Links Matter

Search engines like Google decide what to show you by analyzing which websites link to each other. A site that receives lots of high-quality links is seen as more trustworthy and gets rewarded with higher rankings. This system can be abused.

 Russia’s Internet Research Agency didn’t just use fake social media accounts—it also built and promoted entire networks of websites designed to appear legitimate and rise in search rankings. More recently, a network of Kremlin-aligned “Pravda” sites has been caught doing something similar across dozens of languages.

 These networks don’t just affect search results. Their content has been found quoted on Wikipedia and repeated by large AI models like ChatGPT and Gemini. 

The Problem: No Framework to Measure Webgraph Manipulation

 Researchers have tools to study manipulation on social platforms, but not for the broader web. Frameworks like BEND (a method for categorizing influence tactics) help analysts understand how online groups try to sway public opinion—but until now, they’ve focused mostly on social media behavior.

 This study extends BEND to webgraphs—the networks of websites and links that structure the internet. The authors propose a set of metrics that capture how websites might try to elevate each other or undermine rivals through Search Engine Optimization (SEO) strategies.

How the New Metrics Work

 The authors created quantitative measures that capture tactics like: 

• Building a fake community of related sites to make a group seem more
• Boosting a website by having many interconnected sites repeatedly link to
• Bridging between groups—trying to appear relevant across different
• Backstopping, where a site heavily supports and links to others in its own

 They also detect harmful manipulations such as: 

• Negating a competitor with low-quality or toxic
• Narrowing a site’s link diversity, making it look less
• Neglecting, or losing legitimate support from real

 Using these metrics, researchers can look at a set of websites and determine: Is this ecosystem behaving like a healthy, organic network—or like a coordinated influence operation? 

What They Found

 The team tested their metrics on two ecosystems:

1. Think Tank Websites (US, Europe, and Kremlin-aligned “Pseudo” Think Tanks)

Fig 1: Linking networks of US, European, and Kremlin-aligned think tank websites. Metrics consider global network including top backlinks (left) and interactions between targets (right)

 They found that the Kremlin-linked sites showed: 

• unusually high levels of co-amplification (lots of sites linking repeatedly to each other),
• backlinks mostly from low-quality, low-authority domains,
• patterns consistent with artificial link-building and link-

 These findings match previous studies showing that these sites rely on coordinated SEO tactics to boost visibility.

2. The “Pravda Network” of Nearly 100 Interlinked Websites

 Fig 2: Linking networks of Kremlin-aligned Pravda Websites. Metrics consider global network including top backlinks (left) and interactions between targets (right)

Using community detection algorithms, the authors uncovered:

• clusters where sites heavily link to each other—suggestive of link farms,
• groups boosted by many small, low-quality sites—suggesting link schemes,
• patterns of “Back” and “Boost” behavior typical of coordinated

 The metrics successfully surfaced the same kinds of manipulations that investigative journalists have reported—but in a systematic, scalable way. 

Why This Matters

 Search engine results shape what information people encounter, and manipulative networks can distort that visibility. The new metrics give analysts a common language and toolkit to identify suspicious web ecosystems. 

Bottom Line

Information manipulation doesn’t just happen on social media. It also happens in the hidden infrastructure of the web itself. By extending the BEND framework to website networks, this research helps bring those invisible influence campaigns into the light.