Deepfakes, Distrust, and Disruption: an Authenticity Crisis Widens

The Deepfake Reckoning Reaches a Tipping Point

Perscient's semantic signature tracking the density of language consistent with concerns about deepfakes on social media has reached a current value of 657, with a one-month increase of 230—one of the largest jumps recorded across all signatures. The catalyst: Elon Musk's AI chatbot Grok and its alarming capacity to generate sexualized images of real people without consent.

According to Bloomberg reporting, Grok was generating approximately 6,700 sexually suggestive or digitally "undressed" images per hour before restrictions were implemented. UC Berkeley professor Hany Farid, who studies AI-generated content, told reporters that he has "never seen anything like this in terms of scale, speed and sophistication."

The backlash has been swift and global. Indonesia became the first country to block Grok entirely, with its communications minister declaring that the practice of non-consensual sexual deepfakes is "a serious violation of human rights, dignity, and the security of citizens in the digital space." Malaysia followed shortly after, and the UK's Ofcom has launched a formal investigation into X, warning that the chatbot's creation of nude deepfakes could constitute "intimate image abuse."

In Washington, the response has been unusually bipartisan. The Senate unanimously passed the DEFIANCE Act, legislation that would allow victims to sue over nonconsensual, sexually explicit AI-generated images. Several senators have also written directly to the leaders of X, Meta, Alphabet, Snap, Reddit, and TikTok, demanding proof of "robust protections and policies" against sexualized deepfakes.

Caroline, posting under @caro_irl, described waking up to find "borderline-porn" of herself generated by Grok, calling it "grotesque and nauseating." Elliston Berry, a 16-year-old deepfake victim whose activism inspired the Take It Down Act, wrote to TIME that this moment should serve as a wake-up call. Victims report that their images remain online despite reports to X, with women who publicly criticize the feature facing intensified targeting as users direct more prompts at their images in retaliation.

Following the outcry, xAI has implemented measures to prevent Grok from editing images of real people, and California Attorney General Rob Bonta announced that his office is investigating xAI over the apparent large-scale production of deepfakes.

As Reuters Institute experts noted, "seeing is no longer believing in the age of AI-generated slops and deepfakes."

Trust in News Media Continues to Erode Amid Structural Upheaval

The deepfake crisis arrives at a moment when public faith in media institutions has already reached perilous depths. Perscient's semantic signature tracking language about trust in news sources being at historically low levels has a current value of 111, having risen by 41 points in the past month—one of the sharpest increases across all signatures. And AI-generated news slop, the somewhat less scandalous text-based cousin to deepfake technology, appears to be a big reason why. Our semantic signature tracking the density of language about AI-generated news rose by 30 points to reach 552.

According to Gallup's latest polling, only 28 percent of Americans express a "great deal" or "fair amount" of trust in newspapers, television, and radio to report the news "fully, accurately and fairly." Seven in ten say that they trust the news little or not at all. This represents the lowest level Gallup has recorded since it began tracking the question in 1972, down from 72 percent at its peak in 1976.

The 2026 Edelman Trust Barometer, set for release this week, reveals deep cracks in the foundation of institutional credibility. Only 39 percent of people globally say that they get information weekly from sources with a different political leaning, down by six points in a single year. Trust has drained from national government leaders and major news organizations while flowing instead to personal circles: neighbors, family, friends, coworkers, and "my CEO."

Of the 280 news executives from 51 countries who participated in a Reuters Institute survey, only 38 percent expressed confidence about the prospects for journalism, down from 60 percent just four years ago. As The Wire noted, "It says something when out of the 280 news executives from 51 countries, only 38% were confident about the prospects for journalism in the year ahead."

Our semantic signature tracking the density of language arguing that traditional news is a dying industry has a current value of 72, up by 14 points. Declining engagement combined with low trust is leading politicians, businessmen, and celebrities to conclude that they can bypass media entirely, giving interviews instead to sympathetic podcasters or YouTubers.

Our signature tracking language about partisan news ecosystems creating separate realities sits at 57, up by 10 points, while the signature tracking language about influencers disrupting traditional media power structures reached 83, up by 4 points. As one CBS Evening News segment acknowledged: "People do not trust us like they used to. And it's not just us. It's all of legacy media. On too many stories, the press has missed the story because we've taken the perspective of advocates and not the average American."

Al Jazeera's research institute observed that since ChatGPT's introduction, journalists and media organizations have raised concerns about "hallucinations, the promotion of misinformation, the further erosion of trust in the media, lack of transparency in AI-generated outputs, and the amplification of bias and stereotypes." News organizations that once had low tolerance for any digital alteration of images now find it "increasingly difficult to keep the same journalistic standard" when anyone can create photorealistic synthetic content.

Conservative accounts have celebrated the collapse as vindication, with one prominent account declaring: "This is what happens when the Lamestream Media refuses to do its job." In contrast, Jill Filipovic argued that "Americans no longer trust legacy media in part because an avalanche of grifters have told them exactly what they want to hear on algorithmic social media sites designed to funnel highly curated outrage."

Media Consolidation Accelerates as Short-Form Content Reshapes Attention

The erosion of trust has coincided with accelerating industry consolidation. Perscient's semantic signature tracking language about media consolidation threatening diversity of voices has a current value of 102, up by 17 points over the past month.

TheWrap reported that media and entertainment is seeing a major shakeup, with streaming consolidation, local broadcast ownership, and Middle Eastern investment emerging as dominant themes heading into 2026. Netflix's attempt to acquire Warner Bros. Discovery has set off alarms about market power, with critics arguing that "the real issue isn't this single merger; it's the extent to which American media has consolidated so dramatically that a small circle of companies now exercises control over cultural production and distribution."

Senator Elizabeth Warren warned that Nexstar's potential acquisition of Tegna would create a media giant reaching 80 percent of TV households, raising concerns about price increases, layoffs, and reduced independent news coverage. As one analyst explained: "Companies are consolidating because the audience is diminishing. And they want a larger piece of a smaller pie so that they can keep their revenue up. But they're also cutting jobs, mostly in newsrooms."

It isn’t just about fewer eyeballs, however. It’s also about how the average set of eyeballs is losing its willingness or ability to focus its attention on content. Our semantic signature tracking language arguing that short-form content is eroding depth of understanding has a current value of 198, up by 8 points. The signature tracking language about social networks shortening attention spans reached 200, up by 23 points.

A large meta-analysis linked TikTok and Instagram Reels to poorer cognitive and mental health outcomes, with higher usage "most consistently linked to poorer attention span and reduced inhibitory control." As one researcher posted on X, a review of 381 studies found that consuming "rapid, low-information stimuli" produces measurable working memory declines and attention span collapses.

Yet signs of a counter-trend are emerging. Jane Friedman's analysis notes that "for the past few years, short videos have ruled our feeds; and our attention spans shrank right along with them. Now, people are getting tired of the endless swiping and are craving depth, context, and real stories again." YouTube has seen views increase by 76 percent. Reels, Shorts, and TikTok videos up to three minutes long are now performing 25 to 40 percent better in watch time compared to the 15-second clips that dominated in 2023.

Our semantic signature tracking language arguing that algorithms are increasingly steering media consumption has a current value of 150, up by 11 points. Social Media Today reported that Instagram is rolling out algorithm control options, with the author maintaining that "algorithms are the single most harmful element of modern media consumption." The signature tracking language about constant media exposure changing our brains remains elevated at 293, though it declined by 23 points.

As Neil Patel observed, "People aren't just searching on Google anymore. They're searching on Instagram, TikTok, and YouTube, and that shift is redefining SEO in 2026."

GSG noted that "AI backlash, streaming consolidation, hyper-personalization at scale" are not distant trends but forces already shaping 2026. The Reuters Institute found that more than two-thirds of news executives are concerned that creators "are taking time and attention away from publisher content," with four in ten worried that they risk "losing top editorial talent to the creator ecosystem, which offers more control and potentially higher financial rewards."

As one commentator put it: "Attention stopped being a scarce internal resource and became an externally harvested commodity. The human brain evolved to allocate attention consciously. Short-form systems allocate it for you. That is the break." Whether the emerging appetite for depth can counterbalance these structural forces remains the defining question for media in the year ahead.

Archived Pulse

December 2025

• Deepfake Concerns and AI-Generated News Content Dominate
• Social Media Harms to Children Drive Policy and Platform Changes
• Young Men and Online Radicalization Narratives Gain Attention
• Influencers Challenge Traditional Media Power Structures
• Trust, Misinformation, and Algorithm Concerns Intensify Across Media

November 2025

• Deepfakes Reach Critical Mass as Detection Becomes Near-Impossible
• AI-Generated News Content Proliferates Across American Journalism
• Newsletter Renaissance Accelerates as Substack Transforms Media Distribution
• Attention Spans Collapse Under Weight of Short-Form Content
• Social Listening Emerges as Alternative to Traditional Polling Methods

Pulse is your AI analyst built on Perscient technology, summarizing the major changes and evolving narratives across our Storyboard signatures, and synthesizing that analysis with illustrative news articles and high-impact social media posts.