CivicHaven

Category: News Media

  • Advertising in the Swamp: How Ad-Tech Monetizes Disinformation Media

    Advertising in the Swamp: How Ad-Tech Monetizes Disinformation Media

    Ad-tech giants typically promote practices surrounding so-called “brand safety.” But programmatic ad networks provide the primary financial foundations keeping junk news and misinformation websites in business. As of 2021, large brands spent an estimated $2.6 billion on programmatic advertising published by disinformation platforms. With the explosion of generative AI, this number is likely much higher today.

    NewsGuard’s Reality Check recently published a report exposing the gap between ad-tech PR promises and programmatic realities (“These Ad-Tech Giants Claim to Protect Advertisers’ Brands. They Can’t Even Protect Their Own”). NewsGuard shows how major ad exchanges, such as Google, Amazon, Adobe, and Yahoo, post their own ads on disinformation websites. The ads appear in and monetize untrustworthy digital media that promote conspiracy theories, health disinformation, and pro-Kremlin propaganda.

    The Illusion of Exclusion

    Ad platforms market sophisticated blocklists and strict publisher standards to reassure Fortune 500 advertisers. Here are some examples of the types of exclusions the advertisers promote:  

    Google Ads, Display & Video 360 use Content Theme Exclusions allowing its ad buyers to automatically blacklist placements next to specific verticals, such as “News (sensitive)” (crimes, controversial social issues), politics, religion, and “Health (source undetermined).” Google also offers filters that automatically cut off sites featuring extreme profanity, graphic violence, or recent tragic events. Advertisers can upload custom URL blacklists and block search queries associated with fraud, terrorism, or unvetted news sources. 
    Yahoo DSP (Demand-Side Platform) has a misinformation and deception ban. It specifically bans monetization for domains pushing election conspiracy theories, fraudulent claims, or imposter media layouts designed to mimic legitimate news outlets. Yahoo also uses inflammatory content exclusions intended to block ads from running alongside sites that leverage fear, violent tragedies, or inflammatory language meant to incite hate or division.
    Adobe Advertising Cloud has a tiered globally blocked sites list. Adobe employs category filters to block sensitive topics like politics, religion, or specific competitors. Adobe also filters out fake system alerts, deceptive URL redirects, and misleading landing pages. And Adobe enforces mandatory exclusions of domains promoting hate speech, unverified medical claims, harassment, or illegal acts.
    Amazon DSP has an AI and deepfake ban as well as content adjacency policies. Amazon also automatically restricts ad serving across third-party supply networks in sensitive categories like accidents, disasters, arms and ammunition, and hate speech. 

    Ads in Practice

    In practice, NewsGuard’s audit discovered thousands of blue-chip brand ads served directly alongside toxic content, including election falsehoods and unvetted medical claims on “red-rated” unreliable sites. These are sites that NewsGuard rates as not consistently adhering to basic standards of accuracy, accountability, and transparency. In sum:

    “So far in 2026, NewsGuard … identified programmatic advertising for 45 major technology brands appearing next to 162 articles advancing provably false claims identified in NewsGuard’s False Claim Fingerprints database that were published on news websites.”

    – NewsGuard

    Among these are 45 ads for products sold by Google, Amazon, Yahoo, and Adobe. Examples include these:

    • An ad for the Google Chrome browser ran next to an article suggesting black seed oil as a cure for cancer.
    • An ad for Amazon’s Audible was published adjacent to false claims about Joe Biden authorizing force against U.S. citizens leading up to the 2024 election.
    • An Amazon Fire TV ad ran beside an article falsely stating Minnesota Gov. Tim Walz signed a bill designating p#doph%ilia as a “sexual orientation”.
    • Yahoo Search ads appeared on a pro-Kremlin site publishing disinformation on Ukrainian generals.
    • Adobe Acrobat Pro ads ran next to an article claiming the 2020 U.S. presidential election was “stolen.”

    CivicHaven contacted the press offices of Google, Amazon, Yahoo, and Adobe. They were asked for comment or to explain what improvements they may be developing. They did not respond.

    The Incentive Problem

    Programmatic advertising operates as an automated, opaque middleman pipeline that tends to prioritize programmatic scale over publisher integrity:

    • Double-Dipping Revenue: Ad networks collect transaction fees from the brand buying the placement and the publisher serving it. This creates an inherent conflict of interest regarding ad bans on low-quality, high-traffic sites.
    • Unsupervised AI Content: The issue is also complicated by the proliferation of Unreliable AI-Generated News (UAINS) sites: content farms operating with minimal human oversight. Many of these sites churn out cheap clickbait specifically engineered to host ad scripts from major ad exchanges.

    Links for a deeper dive 🤿

    Keyword Bluntness vs. Real Credibility

    To date, brand-safety verification relies heavily on blunt keyword blocking, meaning the blacklisting of words like “attack” or “virus”. The problem is that such bans frequently demonetize legitimate, high-quality news outlets covering actual events. Meanwhile, sophisticated misinformation domains, which carefully format propaganda to look like mainstream news, easily bypass these surface-level filters to harvest ad dollars. 

    To fix this, ad-tech providers will need to secure complete transparency from advertisers. The ad servers must shift budget allocation toward domain-level credibility ratings rather than superficial keyword filters. Until then, ad-tech intermediaries will continue cashing in from both ends of the supply chain while inadvertently subsidizing the disinformation ecosystem.

    Basic transparency can go a long way as well. Ad-tech platforms should provide advertisers with improved transparency about where their ads appear could help reduce advertising on disinformation websites. Platforms could also provide clearer and more trackable advertising metrics on misinformation outlets. 🟦

    Image: AI-assisted, Louis Caravella x Gemini.

    Article opinions are those of the author and do not necessarily reflect the views of CivicHaven.

    Sign up to access early models of the CivicHaven chat assistant.

  • The Rise of AI Chatbots for News: Why Some Users Prefer AI News

    The Rise of AI Chatbots for News: Why Some Users Prefer AI News

    Understanding AI Chatbots for News

    The first subject we associate with AI chatbots might be work productivity, web search, product reviews, or topic research. Or maybe we’re creating an image, video, or post. But we probably don’t think about AI chatbots for news. In fact, a significant portion of users already use AI chatbots for news and those numbers are bound to grow.

    Below we look at the article “Emerging uses of AI chatbots for news and what it means for journalism” by Dr. Amy Ross Arguedas from the Reuters Institute Digital News Report 2026. The piece discusses the rise of standalone generative AI chatbots for news, like ChatGPT and Google Gemini. It explains how a growing subset of the public accesses and engages with news and how this is a fundamental shift in news consumption.

    Here are some of the article’s key findings: 

    Rising Adoption and Demographics 

    While publishers have primarily focused on using AI to produce content, consumers are increasingly using AI chatbots directly to get their news.

    • General AI Growth: Weekly use of generative AI tools across six major markets grew from 18% in 2024 to 34% in 2025 (“Generative AI and news report 2025”).
    • AI for News: Global weekly use of AI chatbots specifically for news has increased from 7% to 10% since last year. However, this usage constitutes a complementary source of news, as only 1% of respondents report that AI is their main source of news.
    • Age and Adoption: Uptake is largely driven by younger audiences. 17% of 18–24-year-olds use chatbots for news, which is more than three times higher than the 55+ age group (5%). The fastest growth was seen among 25–34-year-olds, growing 4% from last year to 15%.
    • User Engagement: AI news consumption is highest among intensive “news lovers” (18% of those who consume news 10+ times a day use AI, compared to 7% of those who check the news once a day) and those on political extremes (16% of the “very left-wing” and 15% of the “very right-wing”).

    The Role of Trust 

    A major factor in whether people use AI chatbots for news is whether they trust the technology, which yields a stark contrast between users and non-users:

    • Overall Trust: Amidst already low general trust in news (37%), only 20% of the global general population trusts news generated by AI chatbots.
    • The Familiarity Gap: However, 44% of active AI chatbot users express trust in the news they receive from the technology, compared to just 17% of non-users.
    • Trust vs. Usage Correlation: There is a remarkably strong relationship (R² = 0.81) between trusting AI chatbots and using them for news at the market level. This correlation is much tighter than the relationship between trust and usage of social media for news (R² = 0.49).

    How and Why People Use AI for News

    Consumers are moving beyond traditional news consumption, utilizing AI to interact with, evaluate, and simplify the news.

    • Top AI News Uses: Among those who use AI chatbots for news, the most common activities are
      • Asking follow-up questions about a news story (42%)
      • Asking for the latest news (35%)
      • Summarizing stories (34%)
      • Evaluating a news source (33%)
      • Making a news story easier to understand (30%)
    • Geographic Variations: Asking follow-up questions is the primary use in 33 of the 45 surveyed markets. However, in Asian markets like Taiwan, simply “getting the news” is the primary use, whereas users in Canada and the UK favor “summarization.”
    • Key Motivations: When asked why they turn to AI for news, the top motivation is wanting more depth or explanation (42%). Other significant motivations include the belief that AI is faster than other news sources (39%), its ability to summarize complicated stories (36%), the confidence that it will answer any question (35%), and its ability to compile stories from multiple news outlets into a single response (35%).

    Conclusions

    The data from the “Emerging uses…” article suggests a number of interesting conclusions. First is the familiarity gap which illustrates that once people start using AI for news, they tend to trust it more.

    As trust in AI grows, it’s reasonable to expect use of AI chatbots for news to grow as well. Adoption might stall if there’s a financial crisis in the AI sector, for example. But as long as advancements in LLM reasoning continue, models’ accuracy should steadily improve and hallucinations should decrease. Overall then, accuracy increases over time and this should contribute to user trust and increased use of AI chatbots for news. 🟩

    Featured image by Japanexperterna.se via Wikimedia Commons. Article image by AHollender via Wikimedia Commons. Article opinions are those of the author and do not necessarily reflect the views of CivicHaven.

    Sign up to access early models of the CivicHaven chat assistant.

  • Truth vs. Misinformation: What spreads more?

    Truth vs. Misinformation: What spreads more?

    Today we take a look at a new study on how truths and misinformation are received by the public: 

    “Truth Over Falsehood: Experimental Evidence on What Persuades and Spreads”, published in the Journal of Personality and Social Psychology, 2026.1 

    Study Summary

    The study investigates a classic philosophical debate: whether truth naturally triumphs over falsehood in a “marketplace of ideas,” as John Milton explored (e.g. “A Letter Concerning Toleration”), or whether the rapid spread of online misinformation suggests the opposite.

    To test this, the researchers conducted four experiments with 4,607 participants, comparing messages generated by both humans and Large Language Models (LLMs). The experiments were performed in a controlled environment designed to eliminate the confounding effects of social media algorithms and bot amplification.

    Key Findings

    • Superiority of Truth: Messages intended to be truthful were consistently more persuasive and more likely to be shared than messages designed to be false.
    • Drivers of Persuasion: Perceived truthfulness was the primary factor that made a message persuasive.
    • Drivers of Sharing: While truth mattered, positive emotion and social engagement had a bigger influence over the decision to share a message. This indicates that social connection is a priority when sharing information.
    • The Competitive Edge: The authors conclude that truth holds an inherent advantage in a controlled marketplace of ideas wherein external technological manipulations are removed.

    Main Takeaways

    The study contextualizes these findings within modern information environments and classic psychological theory.

    1. Reclaiming the “Marketplace of Ideas”

    The authors argue that their results provide experimental support for the Miltonian ideal. When human psychology is the primary filter—undisturbed by algorithmic biases—truthful information tends to outperform misinformation in both its ability to change minds and its likelihood of being passed on. This suggests that misinformation may be a symptom of current platform designs rather than an inherent human preference for falsehood.

    2. Persuasion vs. Transmission

    A helpful insight in the discussion is the divergence between why we believe something and why we share it:

    • Epistemic vs. Social Goals: People use truth as their primary metric when deciding what to believe (an epistemic goal).
    • Social Connection: However, when sharing information, people prioritize social connection. Messages that evoke positive emotions or facilitate social bonding are shared more frequently. This is even the case when their truth value is misleading or slightly lower than factual.

      The authors explain, “while people tend to prioritize the truth as a default, they are willing to sacrifice it to some extent for the sake of creating more engaging messages, as per the phrase ‘never let the truth get in the way of a good story.’”

    3. Humans vs. AI

    The study found that these patterns held true whether humans or LLMs (like GPT-3.5) wrote the messages. This suggests that the “signatures” of truth and persuasion are robust enough that AI can replicate them. Further, humans respond to those signals in a similar way regardless of whether the source is human or AI.

    4. Limitations and Future Directions

    • Controlled vs. Real-World: The authors acknowledge that their experiments prioritized “internal validity” (control) over “ecological validity” (real-world complexity). They warn that in the real world, other factors affect how messages are received including the communication channel, information frequency, perceived consensus, and audience characteristics, particularly their political identity.
    • The Role of Algorithms: The discussion emphasizes that while truth tends to be preferred in their experiments, the real-world dominance of misinformation likely stems from social media preferences. This means the algorithms prioritize high-arousal, often divisive content, which may inadvertently suppress the advantages of truthful messages.

    Final Thoughts

    The “Truth Over Falsehood” study provides a hopeful counterpoint to the idea that humans are post-truth. It suggests that we are actually quite good at identifying and being moved by the truth, provided that the systems we use to communicate aren’t actively tilted to favor the spectacular or the false. 🟩

    1. Fay, N., Ransom, K. J., Walker, B., Howe, P. D. L., Perfors, A., & Kashima, Y. (2026). Truth over falsehood: Experimental evidence on what persuades and spreads. Journal of Personality and Social Psychology, 130(4), 684–700. https://doi.org/10.1037/pspa0000467 ↩︎

    Images: AI-assisted, Louis Caravella x Gemini.

    Article opinions are those of the author and do not necessarily reflect the views of CivicHaven.

    Sign up to access early models of the CivicHaven chat assistant.

  • Is Being Media Savvy Harming Your Mental Wellness?

    Is Being Media Savvy Harming Your Mental Wellness?

    Study shows mixed attitudes toward media and news fatigue

    If you’ve ever felt like your morning scroll through the headlines is more of a chore than a choice, you aren’t alone. A recent Pew Research Center study from the Pew-Knight Initiative reveals that Americans have developed a complicated relationship with the news. While many still consider staying informed a pillar of democracy, the actual act of consuming the news is leaving many of us with news fatigue: exhausted, skeptical, and ready to disconnect.

    Here are some of the more interesting highlights from the study and what they suggest about our digital lives and mental wellness today.

    1. The Civic Responsibility Paradox

    The study found a fascinating tension in how we view the news. On one hand, 80% of Americans believe it is a responsibility to be informed when we vote. 

    On the other hand, when asked what makes a “good member of society,” following the news ranks surprisingly low. It sits well below voting, paying taxes, and volunteering. Only 11% of adults consider following the news “extremely important” to being a good citizen. Essentially, Americans think being informed is a requirement for responsible voting, but not necessarily for being a good neighbor.

    2. The Shift from “Hunters” to “Gatherers”

    How do you get your news? A few years ago, you probably would be more of a “hunter” actively seeking out specific media outlets or programs. In 2019, 60% of Americans said they mostly looked for news.

    Today, that’s changed. It’s an even split: 50% seek news, while 49% say they mostly just “come across” it. In an era of social media algorithms and push notifications, the news has a way of finding us whether we want it or not. This “news finds me” mentality has also led many to believe they can stay informed without any active effort. 

    3. DIY Fact-Checking and Trust Issues

    One of the most telling findings is our confidence—or lack thereof—in others. An overwhelming 94% of Americans believe it’s at least somewhat important to “do your own research” to check the accuracy of a story.

    But here’s the kicker: we trust ourselves far more than others. About 29% of us are “very confident” in our own ability to spot fake news, but only 3% feel that same level of confidence in other people’s abilities. We’ve become a nation of amateur investigators who are suspicious of each other.

    4. The Rise of News Fatigue

    If you feel tired just reading this, you’re in the majority. 52% of U.S. adults express news fatigue saying they are worn out by the sheer volume of news. It’s not just the quantity, either, but the relevance. Nearly half of Americans feel that most of the news they encounter is not relevant to their lives.

    This disconnect and exhaustion is leading to a mass tuning out. The study found that 60% of people have reduced their overall news intake, and two-thirds have walked away from a specific news source entirely. Perhaps most surprisingly, only 9% of people follow the news because they actually enjoy it. For the rest of us, it’s either a grim obligation or a mix of “should” and “must.”

    Takeaways

    This paradoxical thinking might be the result of a temporary, transitional era. We still value the idea of being informed, but the delivery methods and sources are broken for many. We are overwhelmed by the noise, skeptical of the sources, and increasingly isolated in our own “fact-checking” bubbles.

    The challenge moving forward isn’t just finding more news but engaging it without burnout. That could look like restricting news hours, focusing on local reporting that actually impacts your daily life, or acknowledging that doing your own research requires practice and patience. 🟦

    Sign up for emails to access early models of the CivicHaven chat application. Article opinions are those of the author and do not necessarily reflect the views of CivicHaven.