This Storyboard - which we call our "stain" chart - shows you at a glance how strong or weak a given narrative is right now relative to its history.

For each narrative or "semantic signature" listed on the left of the chart, we have a series of blue dots on the right, each of which represents a specific weekly density or volume of that narrative. reading from within the date range that we are covering. The yellow arrow is the most recent reading, so it's just like the "YOU ARE HERE" spot on a map. The x-axis scale shows the range of index values.  If a dot is at 100, that means that story is 100% more present in media than usual. If it’s at 0, it means it’s at its normal level.

The light blue shaded box covers the middle 50% of readings across the date range, so you can see quickly if the current reading is typical (inside the blue box), depressed (left of the blue box), or elevated (to the right of the blue box).

If you hover over a specific blue dot, you will see the specific date and measurement that the dot represents.

AI and Deepfake Content Fears Shape October Media Narratives

Deepfakes Reach Critical Mass as Detection Becomes Near-Impossible

The ability to distinguish authentic video content from AI-generated fabrications has reached a crisis point, and it’s showing up in the conversations we are having about media. Perscient's semantic signature tracking concerns about deepfakes on social media platforms rose to a z-score of 5.47, approaching its all-time high of 5.65, as language expressing alarm about AI-generated fake videos intensified by 0.26 from the previous week.

Some fabricated content has become so common that it is practically a meme on some social media sites. For example, during Hurricane Melissa in late October 2025, AI-generated disaster videos flooded social media sites. As with prior hurricanes, the most viral content included sharks swimming in a Jamaican hotel pool. Yet other examples included fabricated images of Kingston airport allegedly ravaged by the storm, which circulated widely as the devastating storm made landfall. The fabricated content competed with legitimate hurricane coverage, creating confusion during an actual disaster.

The trajectory of deepfake incidents suggests exponential growth. Deepfake incidents in the first quarter of 2025 already exceeded all of 2024 by 19 percent, with 179 incidents recorded compared to 150 for the entire previous year. Women journalists have been particularly targeted, with Women Press Freedom documenting seven major deepfake incidents in 2023, rising to 20 in 2024, and more than 150 cases in the first nine months of 2025. Broadcast journalists around the world are being targeted in an explosion of deepfake scams used to peddle products online.

The human capacity to detect these fabrications has essentially collapsed. Testing revealed that only 0.1 percent of participants could correctly identify all fake and real media shown, while 70 percent of people doubt their ability to distinguish real from fake voices. Social media commentators have begun sounding alarms about the implications. One observer warned that "in the near future, open-source models will make it easy to generate fake videos of anyone using public footage" that will be nearly impossible to detect, while another noted that "within the next year or two, anyone who hates you will be able to generate any kind of defamatory video of you" that cannot be proven fake.

Tech companies have pledged to add tamperproof markers to AI-generated videos, with social platforms promising to make those markers visible to users. However, testing by The Washington Post found that Facebook, TikTok and other major platforms do not use the tech industry standard touted as a way to flag fake content. Even when watermarks are included, observers note they can be easily removed, and open weights AI video models without guardrails are coming.

AI-Generated News Content Proliferates Across American Journalism

Perscient's semantic signature measuring language indicating discussion of concerns about the proliferation of AI-generated news content reached a z-score of 4.29, reflecting growing awareness that artificial intelligence has become a significant producer of journalism. New research released in October 2025 found that more than 9 percent of all news coverage in U.S. newspapers contained AI-generated text, a finding that surprised many in the industry.

The distribution of AI use varies dramatically by publication size and section. AI use on news pages of major daily papers with circulation over 100,000 was limited to 1.7 percent of articles, while nearly one in ten articles at smaller papers contained significant AI content. Opinion pages showed even higher adoption rates, with AI-generated content appearing in The New York Times, The Wall Street Journal, and The Washington Post at 4.5 percent overall, significantly more common than the 0.7 percent on news pages.

The global scope of AI-generated news operations has expanded rapidly. NewsGuard identified 2,089 undisclosed AI-generated news and information websites spanning 16 languages, with generative AI tools deployed to turbocharge misinformation operations. These unreliable outlets operate with little to no human oversight, representing a boon to content farms and misinformation purveyors.

Traditional news organizations have begun experimenting with AI presenters. Britain's Channel 4 launched an AI-generated news presenter named "Arti" in late October 2025 to read dispatches on social media channels, marking a first in UK television history. Meanwhile, former local news websites like the Farmingdale Observer abruptly pivoted to AI-generated viral content in 2024, with articles showing up in Google Discover and getting help from Google's algorithms.

The phenomenon has sparked concern among legislators. Senator Amy Klobuchar stated that "AI-generated news articles are drowning out articles written by actual humans. We need clear rules of the road for AI to prevent disinformation and to ensure journalists are compensated for their work." Others see different implications, with one observer arguing that AI content represents "the future of hyper-local media" because "traditional local news is dying because the economics don't work."

Newsletter Renaissance Accelerates as Substack Transforms Media Distribution

Perhaps as a result of the dilution of traditional forms of media, Perscient's semantic signature tracking language about newsletters bypassing traditional media distribution rose by 0.69 to a z-score of 4.79, reflecting the platform's transformation into a significant force in media economics. At least 52 Substack accounts now earn a minimum of $500,000 annually, up from 27 accounts in February 2023, with the top ten accounts collectively earning $40 million a year.

The platform's growth metrics demonstrate its accelerating momentum. As of March 2025, there were more than 5 million paid subscriptions on Substack, up from 4 million just four months earlier, with more than 50,000 writers earning money from the platform. Substack added 500,000 new paid subscriptions and 32 million new free subscriptions in just three months, heavily driven by Notes and the app.

Prominent journalists continue migrating to the platform. Recent moves included The Times' David Aaronovitch and Henry Winter, former MSNBC contributor Mehdi Hasan, The Guardian's Jim Waterson, and Washington Post columnist Jen Rubin, whose new Substack already has tens of thousands of paying subscribers. The platform has evolved beyond its original newsletter focus, with Notes becoming the dominant growth engine. One creator made $4,546 from a single note, demonstrating the platform's capacity for viral monetization.

The success stories have proliferated on social media. One writer celebrated breaking through $80,000 of annual revenue after starting with zero expectations, while another detailed how a small newsletter for Cleveland grew to 20,000 subscribers with a 50 percent open rate and $3,000 monthly revenue, on pace for $100,000 the following year.

The platform's transformation of media economics has drawn analysis from observers who see it as part of a broader shift. One commentator argued that "Substack is eating the audience of longform writing. Journalists can just start their own stack, build their own audience. Legacy media responds by putting everything behind paywalls, which destroys their reach, shrinks their audience, sending them into a death spiral." However, others have expressed concern about the loss of editorial oversight, noting that while "there is outrageously good stuff being published on Substack," it's "all done without editors," with journalism "increasingly replaced by opinion pieces."

Attention Spans Collapse Under Weight of Short-Form Content

Perscient's semantic signature measuring language about social networks shortening attention spans intensified by 0.36 to a z-score of 4.35, reaching its all-time high. The average social media user's focus on a single post declined from 12.1 seconds in 2015 to 8.25 seconds in 2025, while teen users now toggle between apps every 44 seconds, compared to 2.5 minutes a decade ago.

Research published in BMC Pediatrics found that users who primarily consume short-form content under 30 seconds show a 27 percent reduction in sustained attention during task-based activities, with 61 percent of users aged 18-34 affected by "scroll fatigue." The study revealed that frequent exposure to rapid, fragmented content on platforms like TikTok and Snapchat overstimulates cognitive processes, leading to decreased working memory capacity and impaired cognitive control.

Behavioral data confirms the shift in consumption patterns. In a 2025 study, 52 percent of respondents admitted they skip videos longer than 60 seconds, even if the topic interests them. Generation Z has an average attention span of approximately 8 seconds compared to Millennials' 12 seconds, though 53.1 percent of U.S. Gen Zers also regularly engage with long-form content, suggesting a more complex relationship with different content formats.

The cognitive implications extend beyond simple attention metrics. New research from a Neurology study shows thinking and memory problems have nearly doubled among U.S. adults under 40, with younger brains particularly affected. A JAMA study found that "preteens using increasing amounts of social media perform poorer in reading, vocabulary and memory tests in early adolescence compared with those who use no or little social media."

Cultural commentators have begun articulating concern about the broader implications. One observer noted that "you don't lose yourself in the same way with a 15-second social media reels as you do with a 300-page book. Reading a book means stepping into someone else's mind and walking beside them for hours." Another warned that "there is too much entertainment. This constant need to be amused has had catastrophic effects on our collective focus, our literacy, our education, our politics."

Social Listening Emerges as Alternative to Traditional Polling Methods

Perscient's semantic signature tracking language about social listening replacing traditional polling strengthened by 2.67 to a z-score of 2.52, reflecting growing commercial and analytical interest in social media monitoring as a research methodology. The global social media listening market is projected to grow from $9.61 billion in 2025 to $18.43 billion by 2030, with 89.4 percent of surveyed companies using listening to analyze social data, making it the number one methodology for this purpose.

The adoption of social listening reflects changing priorities in market research and cultural analysis. Nearly 88 percent of agencies and over 68 percent of in-house brands perceive social listening as important for achieving business objectives, with 79 percent of marketers believing data analytics skills, including social listening, are critical for success in 2025. Companies using social listening effectively achieve up to 10 percent faster revenue growth compared to those that do not.

Investment in the space continues to grow, with over 30 percent of people in 2025 using two social listening tools. The primary objective for analyzing social data has shifted toward cultural and trend analysis, with AI-driven assistants now able to sift through millions of social mentions in seconds to pinpoint what matters, categorizing conversations by topic, sentiment, and cultural context.

The platforms being monitored have expanded beyond traditional social networks. Brands now monitor Reddit, Twitch, and niche forums for real customer insights, with conversations in these spaces often more honest and detailed than public posts. Research from PyMC Labs and Colgate demonstrated that GPT-4o and Gemini could predict purchase intent at 90 percent reliability compared to actual human surveys, using a method called Semantic Analysis with zero focus groups or survey panels.

The expansion of audio data as a source for analysis represents another frontier. The market for environmental audio data is projected to grow from $0.35 billion in 2023 to $23 billion by 2032, with voice AI alone projected to surpass $100 billion. Companies are using global audio datasets to train accent-aware speech models, helping machines understand human speech across accents, languages, and conditions, reflecting the broadening definition of what constitutes analyzable social data.

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.