Media is dead.
Long live media.
Media institutions are shrinking while demand for reliable information keeps growing. This scan follows the evidence out to 2036.
Awestruck Futures is the foresight practice from Awestruck Labs. Each edition takes a single question about the next decade and runs it through a structured methodology: signals, drivers, proto-forecasts, and alternative futures.
Media is dead. Long live media.
Asking "Is media dead?" confuses the institutions that delivered media with the job media does.
The institutions that organized media for the past 70 years, including general-interest publishers, mid-size cable networks and regional broadcasters, are in terminal decline. In 2025 alone, 136 US newspapers closed and 17,000 entertainment and media jobs were cut. The Paramount-Skydance-WBD merger created a $111 billion entity while local TV newsrooms disappeared overnight.
The function those institutions served, informing people so they can make decisions, is in higher demand than ever.
Retail went through the same pattern: Amazon took the scale end, DTC brands took the niche end, and the middle hollowed out. In media, three to four mega-entertainment entities are consolidating at the top while thousands of specialized micro-operations become viable at the bottom, leaving little room in between.
The Attention Defection
The audience is leaving of its own accord.
40% of people globally now avoid the news on purpose, mostly because of what it costs them emotionally: 39% cite its effect on their mood, 31% feel overwhelmed and 20% feel powerless to act on what they learn.
Meanwhile, a new category of attention competitor has emerged that few media strategists are modeling. AI companion apps recorded 220 million cumulative downloads by mid-2025, and 72% of US teenagers have tried one, with 52% using one regularly. People use them as relationship substitutes, and they absorb hours that used to go to scrolling feeds.
UC Irvine's Gloria Mark found people now average 47 seconds on any one screen before switching, down from about two and a half minutes in 2004. At that speed, every screen is a triage decision.
Discord absorbs 94 minutes daily from its user base. "Parasocial" was named Cambridge Dictionary's Word of the Year. The floor of available attention is lower than any media business model assumes, and the competitors for that attention include entities that talk back.
The Epistemological Crisis
The information environment is becoming untrustworthy from several directions at once.
AI generates content at zero marginal cost. About half of new English-language web articles are now primarily AI-generated (a peak of 50.9% in late 2025), though most of them barely surface in Google or ChatGPT. A single false blog post was enough to get a Google AI Overview to cite it as fact, because repetition at scale now passes for consensus.
Screenshots destroy provenance. Samsung shipped the first phone with embedded content credentials in January 2025, but a single screenshot eliminates all verification data. Detection tools are failing. Across 56 studies and 86,000+ participants, people spotted deepfakes 55.5% of the time, statistically no better than a coin flip.
The cost of evaluating whether something is true now exceeds the value of consuming it for most casual information needs. Read that way, the 40% news avoidance rate is a rational response.
The Paradox
The opportunity comes from model collapse.
Model collapse, documented in Nature, means AI systems trained on AI-generated content degrade irreversibly. The technology companies that spent a decade devaluing human-produced content now need it to survive. Verified, human-created content becomes essential training substrate. The institutions dying today are the essential infrastructure for the technology killing them.
Whoever supplies that verified content will build the next decade of media.
The data that anchors this scan
These numbers show where the legacy system is collapsing, where attention has already moved and where the money is consolidating.
AI companion app downloads by mid-2025. 88% year-over-year growth. Average sessions 10x higher than general media apps.
Average time on any one screen before switching, down from about 2.5 minutes in 2004. Teens toggle even faster.
of Google searches end without a click. Rises to 83% when AI Overviews appear. Publisher referral traffic dropped 33% in one year.
peak share of new English-language web articles that were primarily AI-generated (late 2025). Most of them barely surface in Google or ChatGPT.
of global news consumers now actively avoid news. Joint-highest ever recorded. US: 42%. UK: 46%.
annual cost to watch all NFL games across 10 services. FCC opened formal inquiry. Nearly 9,000 public comments filed.
human accuracy spotting deepfakes across 56 studies, statistically no better than a coin flip. 46 US states have enacted synthetic media legislation.
nonprofit newsrooms in the US, up from near-zero twenty years ago. $400M in philanthropic funding mobilized through Press Forward.
Five working forecasts
Each forecast is a structural claim about how the next decade reshapes the business of media, based on where the signals point.
The 47-second attention window is real, and it applies to commodity information: general news, entertainment, opinion, anything where nothing forces depth. A parallel economy exists for high-stakes, specialized information where the consumer has no choice but to go deep, such as a parent researching a child's diagnosis or a professional tracking regulatory changes. In those domains, attention is unlimited, urgent and willing to pay.
By 2030, media splits along this axis. Shallow-commodity content must deliver value in seconds, monetized by advertising or free with platform subsidy. Deep-specialized content earns sustained attention through irreplaceability, monetized by subscription, membership, or institutional license. The shallow layer is enormous and nearly free. The deep layer is fragmented, high-value, and structurally resistant to AI substitution because the consumer's stakes demand verified human judgment.
The 47-second rule governs optional information, and essential information runs on different economics entirely.
Where this goes by 2036
Each of these futures is internally consistent and plausible given the signals.
The media industry built between 1950 and 2020 ran on a specific economic arrangement: institutions aggregated audiences, sold access to advertisers, and used the revenue to fund content. That arrangement has broken: advertising migrated to platforms, audiences fragmented, AI flooded the content supply, and the institutional middle hollowed out.
Several models replace it, operating at the same time.
Independent creators and journalists with owned audiences form cooperatives. These are mutual credibility networks, where membership signals quality and shared infrastructure lowers operating cost. These cooperatives sell direct subscriptions, run collective ad sales, and negotiate AI content licensing deals as a bloc.
Brands, finding that owned audience is their most defensible marketing asset, begin sponsoring or renting custom newsrooms. The brand funds a dedicated editorial operation that produces real content under a credibility framework it could never build alone.
AI companies, facing model collapse, license human-verified content at scale. This creates a recurring revenue stream for anyone producing original, verifiable work. The content the technology industry spent a decade devaluing becomes essential infrastructure for its core product.
The media landscape becomes a barbell: 3-4 mega-entertainment entities at the top, thousands of specialized micro-operations at the bottom, nothing viable in the middle.
Who wins: Creators with owned audiences. Cooperative networks with brand sponsors. AI content licensors. Brands that build or rent newsroom operations. Platforms that provide infrastructure.
Who loses: Mid-size general-interest publishers, SEO-dependent content operations and legacy newsrooms that keep their old advertiser relationships.
Wild card: Whether platforms that host cooperatives eventually extract enough rent to recreate the same dynamics that killed legacy media.
The through-line across all four: Across all four, the function media served (sense-making, accountability, shared reference) keeps growing while the institutions that delivered it keep shrinking.
Three forks between now and 2036
A single trunk in 2026 diverges at three decision forks into four endpoints in 2036. The shape of media in ten years depends on which forks resolve which way.
Do trust institutions form?
- YesUpper branch · Futures A or B
- NoLower branch · Futures C or D
Do creator cooperatives, verification networks, content-licensing collectives, or new editorial institutions emerge fast enough to establish credibility infrastructure before the information commons degrades? The 407 nonprofit newsrooms, Press Forward's $400M, the creator cooperative experiments, Lookout Local's 68% open rates are all early signals. The question is whether these remain experiments or become infrastructure.
Does the web stay open?
- OpenFuture A · The New Guilds (40%)
- ClosedFuture B · The Walled Gardens (25%)
If trust institutions form, the next fork is structural. Do publishers keep content on the open web (with new revenue models like AI licensing, brand sponsorship, cooperative ad sales) or do they pull content behind walls entirely? The 33% referral traffic decline and 65% zero-click rate push toward closure. The brand-as-sponsor and AI-licensing revenue models pull toward openness.
Who funds verification?
- AI companiesFuture D · The Platform Pivot (15%)
- NobodyFuture C · The Dark Age (20%)
If trust institutions do not form organically, the question becomes whether any entity with sufficient capital steps in. Model collapse gives AI companies a structural economic incentive: they need human-verified content or their products degrade. If they act on this incentive, verification gets funded as infrastructure. If they do not (or act too slowly), the information commons continues to degrade.
Five strategic questions
The institutional middle is past fixing, and the function media serves is looking for new structures to live in.
Do you build an influencer-led newsroom?
Recruit or develop creators with owned audiences. Give them editorial infrastructure. Let them carry the credibility relationship your masthead no longer can. CNN, Yahoo, The Washington Post, and Bustle Digital Group are already building curated creator networks as internal operating models. The boundary between "journalist" and "creator" is dissolving. Publishers who treat this as a talent retention problem may accidentally build the next media model.
Do you restructure your advertiser relationships?
Move from selling ad impressions against your content to offering brands a sponsored newsroom: dedicated editorial capacity that produces real content under a credibility framework the brand cannot build alone. Advertisers are buying a trusted audience relationship, and you have the credibility infrastructure to provide one.
Do you license your content differently?
Your archive and ongoing human-verified content production is training substrate that AI companies need to avoid model collapse. Price it as a revenue line. Publishers negotiating licensing deals are treating their archives as intellectual property. This is a 3-5 year window before the market matures, and the publishers who move first set the terms.
Do you close the SEO door?
Every piece of content you index for search is content an AI Overview can summarize without sending you a reader. Google referral traffic dropped 33% in one year. Zero-click searches hit 65%. The math on open distribution may no longer work. Closing it costs you visibility on the open web and gives you a direct audience relationship you control.
Do you specialize into the deep-attention economy?
The 47-second window governs commodity information, while high-stakes verticals (health, legal, financial, regulatory) operate on entirely different attention economics. A parent researching a child's diagnosis does not toggle after 47 seconds. A professional tracking regulatory changes that affect their livelihood does not need to be entertained. Readers in those verticals will pay for depth, which makes specialization a positioning move.
The brand-as-media-company opportunity
The media crisis is also a brand problem, and it contains a brand opportunity most companies have not yet named.
For decades, brands rented access to audiences assembled by media companies: buy the placement, borrow the publisher's audience and hope the message lands. That intermediary layer is collapsing as advertising revenue moves to platforms, publisher audiences fragment and the referral web breaks.
Owned audience is now the most defensible marketing asset. Brands that recognized this early are already building internal content studios that outperform external agencies. A brand can produce video, podcasts, newsletters, and social content at scale and still not own the audience relationship if all distribution runs through platform algorithms the brand does not control.
Content marketing produces material for a brand's sales funnel. A brand-as-media-company funds a genuine editorial operation that serves an audience's information needs, and in doing so builds a trust relationship the brand owns outright.
Examples would include a financial services company funding a personal finance newsroom staffed by independent journalists, a healthcare company running a patient information service with clinical editorial standards, or a technology company sponsoring a research publication in its category. The brand pays for it, the editorial team supplies the credibility, and the audience gets useful information and a trust relationship with the brand that no ad placement could create.
Strategic questions for brand leaders
What information does your audience need that no one is providing well?
The gap between what your customers need to know and what existing media serves them is your editorial opportunity.
Can you fund editorial capacity with genuine independence?
The credibility holds only if the editorial operation stays independent of sales: the brand funds it and keeps its hands off the editorial output. This is the sponsored-newsroom model, and it requires executive conviction that trust compounds over time.
Are you building audience you own or renting audience you do not?
Every dollar spent on platform advertising builds the platform's audience data. Every dollar spent on an owned editorial operation builds a direct relationship with an audience that comes back because the content serves them.
What is your content licensing position?
If your brand produces original research, proprietary data, or expert analysis, that content has value as AI training substrate. License it. The same content that serves your audience also generates licensing revenue from AI companies facing model collapse.
Twelve drivers shaping the next decade
Each driver below has a velocity, a trajectory and a note on what it does to the media system.
About half of new English-language web articles are primarily AI-generated (a 50.9% peak in late 2025). Deepfake production grew 900% YoY. Human deepfake detection averages 55.5% across 56 studies, no better than chance.
Volume overwhelms the ecosystem's ability to distinguish human from machine. Produces model collapse, epistemological corruption, and a paradoxical premium on verified human content.
42 signals across five clusters
Each cluster groups signals that point to the same kind of change.
- 20
17,000+ entertainment and media jobs cut in 2025 (18% increase over 2024). Restructuring and consolidation. Jobs not returning in previous form.
Strong - 21
Spotify, iHeart, and Amazon all cut podcast staff while independent creators grow. Corporate podcast investment era (2019-2023) over. Format survives, money flows to individuals.
Strong - 22
47% of consumers feel overwhelmed by subscriptions. Deloitte: 41% say content not worth the price. Unbundling thesis reversed. Rebundling concentrates power in the largest catalogs.
Strong - 26
136 US newspapers closed in 2025 (two per week). 213 news desert counties. 50 million Americans with limited access to local news. Rate not slowing.
Strong - 29
Substack reaches $1.1B valuation, writers earn $450M. 5M paid subscriptions. Writer revenue exceeds many traditional publishers' total.
Strong - 33
WRTV Indianapolis newsroom disappeared overnight. Broadcast consolidation mirrors print with a 5-10 year lag.
Moderate - 36
YouTube consolidates digital video at 2.7B monthly users. Individual creators rival broadcast networks in reach. YouTube now functions as the cable bundle.
Strong - 41
Paramount-Skydance acquires WBD for $111B. Largest media merger in history. Top of the pyramid consolidating into 3-4 entities.
Strong
Questions worth asking
The institutional middle is disappearing, and these are the practical questions that follow.
Restructuring, and unevenly. The commercial local newspaper model is finished in most markets. What replaces it is a mix: AI-assembled hyperlocal newsletters (Patch, Lookout Local), nonprofit newsrooms (407 organizations and counting), state-funded consortia, philanthropic capital through Press Forward, and Facebook groups operating as volunteer-run civic infrastructure.
Where the signals came from
Awestruck Futures
What this means for your marketing
- 01
Owned audience is now the most defensible marketing asset as rented publisher access collapses.
- 02
Treat AI companies as a customer, not just a traffic threat: verified, specialized content is becoming something they pay to license.
Want a forecast like this for your category?
We build studies like this on questions our partners need answered before the market catches up.
Awestruck Futures is the foresight practice from Awestruck Labs. Each installment takes a single question about the next 10, 20, or 30 years and maps it through structured methodology: signals, drivers, proto-forecasts, and alternative futures.
Awestruck Futures · Media 2036
42 signals · 12 drivers · 4 scenarios · 10-year horizon