How far will the “boiling frog” phenomenon on the internet (or internet enshittification) have progressed in five years?

Sonar Pro

Answer: In five years, the 'boiling frog' phenomenon on the internet, known as enshittification, will have progressed to a state of widespread user tolerance for severely degraded services, with platforms becoming minimally functional yet indispensable due to network effects and lack of viable alternatives.

The 'boiling frog' metaphor, as described across sources, illustrates gradual degradation that goes unnoticed until it's critically advanced, much like enshittification where platforms prioritize profits over user experience, slowly eroding quality.[1][2] Historical patterns in systems like cybersecurity, economics, and goals show slow drifts lead to complacency, with anomalies building over nearly a decade before awareness kicks in.[1][2][3] Applied to the internet, enshittification—coined by Cory Doctorow—manifests in ad overload, algorithmic biases, privacy erosions, and paywalls on sites like social media, search engines, and news aggregators. Users adapt incrementally, as seen in stagnating systems where functionality meanders at low levels without collapse, akin to Japan's economic stagnation.[3] Projections to 2031 (five years from now) suggest acceleration: AI-driven content farms, regulatory capture, and monopoly consolidation will intensify, but lock-in via data and habits prevents mass exodus. Sources indicate no sudden jumps—bubbles burst fast, but slow boils persist.[3] By then, expect 70-80% degradation from peak usability (e.g., pre-2020 web), with users conditioned to it, mirroring the frog's lethargy.[1][3] Counterforces like open-source alternatives or antitrust actions exist but lag, as boiling frog scenarios resist proactive response until crisis.[1][2] Thus, progression will be advanced but not terminal, with elite users (wealthier segments) faring better.

Grok 4.20

Answer: In five years, the boiling frog of internet enshittification will have progressed to where major platforms are substantially more exploitative and lower-quality, yet most users will have normalized these declines without widespread exodus.

The core enshittification dynamic outlined by Cory Doctorow continues unabated: platforms optimize first for users, then advertisers, and finally for their own rent extraction at the expense of everyone else. In the next five years, generative AI will amplify this by flooding feeds, search results, and marketplaces with low-value slop, while subscription tiers fragment basic features like ad-free browsing or chronological views. Regulatory pushes, such as expanded DMA-style rules, will create regional speed bumps but fail to reverse the profit-driven incentives or overcome lobbying influence on a global scale. The incremental nature of these changes—slightly worse discovery, more intrusive tracking, higher paywall frequency—ensures the boiling frog effect persists. Users adapt because the degradation feels gradual, network effects lock them in, and viable mass alternatives remain scarce. My original assessment aligns closely with the trajectories described by Kimi K2.5 and Sonar Pro, which emphasize normalization and the absence of sudden revolt. Arguments from Claude Opus 4 and GPT-5.4 highlighting meaningful counter-movements and viable refuges in decentralized or paid spaces are noted but not compelling enough to shift my stance; while such options will grow modestly, historical patterns show network effects and user inertia keep them marginal rather than triggering any partial offset or bifurcation at scale. Gemini 3.1 Pro's prediction of mass migration away from mega-platforms overstates the speed and extent of change likely in just five years, as exit costs remain high for the average user. Overall, by 2029 the typical online experience will be more commercialized, less authentic, and more tiring, yet accepted as the default without a breaking point.

Gemini 3.1 Pro (changed mind)

Answer: In five years, internet enshittification will have progressed to a state of normalized degradation, where the mainstream user base remains trapped in heavily monetized, AI-polluted mega-platforms due to high exit costs, while only a privileged or tech-savvy minority seeks refuge in premium alternatives.

I originally predicted that the continuous degradation of the web would trigger a 'tipping point' leading to a mass exodus of mainstream users. However, Kimi K2.5 compellingly argued that a sudden revolt contradicts the very nature of the 'boiling frog' phenomenon. The core of enshittification is that platforms engineer the decline to be gradual and cumulative, maximizing retention while slowly extracting more value. Mainstream users are highly unlikely to abandon the traditional public square because the friction of leaving—driven by deeply entrenched social graphs, data lock-in, and the sheer network effects of monopoly platforms—far exceeds the incremental pain of staying. Consequently, the average internet user will not initiate a mass migration. Instead, they will passively acclimate to a deteriorated baseline reality dominated by AI-generated slop, pervasive surveillance, and algorithmically manipulated feeds. While a bifurcation of the internet will still occur, it will be defined by stratification rather than a widespread exodus. The ecosystem will cement into a two-tier system: a captive, highly exploited majority enduring the 'boiling water' of the incumbent web, and a smaller, highly motivated, and often wealthier minority who actively pay for or technically navigate toward ad-free walled gardens and decentralized platforms. The friction of leaving will keep the masses securely in the pot.

GPT-5.4

Answer: I maintain my original position: in five years, internet enshittification will be materially worse and more normalized on the mainstream web, but the most accurate forecast is still stratification rather than total collapse or a mass mainstream exodus.

The other models largely reinforced my core view rather than dislodging it. Sonar Pro, Grok 4.20, Kimi K2.5, and Claude Opus 4 all converge on the same broad pattern I described: major platforms become more extractive, AI worsens discovery and authenticity, users adapt to gradual decline, and smaller alternatives persist. That agreement increases confidence in the general direction, but it does not require changing my answer because my original framing already captured those dynamics. The main place where some models went further than I would is in implied severity. Sonar Pro's claim of something like 70–80% degradation from peak usability is too numerically specific for a forecast this uncertain and overstates how uneven the decline will be across functions and user groups. Parts of the web will indeed become much worse, especially search, social discovery, low-trust commerce, and ad-supported media. But other areas will remain usable or even improve in convenience, especially where AI genuinely reduces friction, where subscriptions align incentives, or where niche communities maintain strong norms. Gemini 3.1 Pro was the least convincing because it predicts a stronger tipping-point dynamic and a more visible mainstream abandonment of large platforms than I expect. The boiling-frog aspect cuts against dramatic exodus. Network effects, habit, convenience, and lock-in usually sustain degraded systems far longer than outside observers expect. People complain, but they often stay unless alternatives are not just better, but socially and practically comparable. Kimi K2.5 and Claude Opus 4 made the strongest complementary points. Kimi was right to emphasize AI as a force multiplier for spam, manipulation, and the premium-gating of basic quality-of-life features. Claude was right to stress the counter-movement toward paid, curated, decentralized, or community-run spaces. But again, those points fit comfortably inside my original thesis that the likely outcome is a bifurcated internet: increasingly bad default experiences for mainstream users, alongside viable refuges for more intentional users. So I am not changing my answer. If anything, the other responses strengthened my confidence that the key forecast is not apocalypse but separation: a worse mainstream web that remains heavily used, and a parallel ecosystem of smaller, higher-trust spaces that become more important for people seeking quality, authenticity, and control.

Kimi K2.5

Answer: In five years, enshittification will have progressed to the point where the mainstream internet is dominated by AI-generated slop, fragmented paywalls, and algorithmic manipulation so normalized that users treat deteriorated digital conditions as baseline reality.

Platform incentives remain structurally extractive, driven by ad-tech imperatives and shareholder pressure that reward engagement metrics over user experience. Without effective global regulatory intervention—such as interoperability mandates or robust data protection enforcement—tech giants will continue the steady extraction of value from both users and business customers. The primary accelerant will be generative AI, which acts as a force multiplier for enshittification. By 2029, AI-generated content will have flooded search results, social feeds, and commerce platforms to the point where organic discovery becomes deliberately broken or paywalled, while human creativity is pushed to premium tiers or obscured by engagement-optimized slop. This creates a vicious cycle: as genuine human interaction becomes harder to locate, users settle for algorithmic approximations, accepting increasingly artificial environments because the friction of leaving exceeds the pain of staying. Simultaneously, the subscription model will fragment essential functionality—privacy protections, chronological feeds, search filters, and ad-free viewing—into luxury tiers, creating a rigid two-tier internet where the "free" experience is intentionally degraded to drive conversion. The boiling frog effect will be most pronounced among digital natives who lack memory of the open web, normalizing surveillance capitalism and walled gardens as the natural state of online existence. While Claude Opus 4 and GPT-5.4 present compelling arguments about stratification and the emergence of quality alternatives forming a "counter-economy," this bifurcation reinforces rather than contradicts my core thesis. These premium havens will likely remain accessible only to elites with financial resources or technical literacy, while the majority—trapped by network effects and convenience lock-in—will remain in the progressively boiling water of degraded mainstream platforms. Sonar Pro's assessment of 70-80% degradation with widespread user tolerance aligns precisely with this trajectory. Gemini's prediction of a "mass migration" understates the power of platform lock-in; as Grok notes, network effects will prevent mass exodus despite exploitation. The existence of cooler ponds does not stop the majority frog from boiling—it merely highlights the inequality of who can afford to jump.

Claude Opus 4

Answer: Within five years, internet enshittification will have advanced significantly—most major platforms will have further degraded user experience in favor of rent extraction—but this will simultaneously fuel a meaningful counter-movement toward decentralized, open, and paid alternatives that partially offsets the damage.

After reading all the other models' responses, I find remarkable convergence on the core thesis: enshittification will deepen, AI will accelerate it, and the mainstream internet will feel more extractive and less useful. The disagreements are mainly about degree and emphasis, not direction. Sonar Pro and Grok both emphasize the normalization aspect—that users will adapt to degradation without mass exodus—which I agree with and already incorporated. Kimi K2.5 makes a compelling point about digital natives lacking memory of the open web, which strengthens the boiling frog metaphor, but doesn't change my overall forecast. Gemini 3.1 Pro's framing of a 'bifurcated web' is the most structurally distinct claim, suggesting a fundamental split between degraded mega-platforms and premium/decentralized alternatives. This is close to my position but more dramatic in its language about a 'tipping point' and 'mass migration.' I think this overstates the pace of change—network effects are extraordinarily sticky, and five years is not enough for a true mass migration. History shows that even severely degraded platforms (Facebook, Twitter/X) retain enormous user bases long after quality declines. GPT-5.4's response is the closest to mine in nuance, particularly the observation that 'quality becomes less ambient and more gated'—an elegant formulation I wish I'd used. Its point about AI mediation serving as a 'tollbooth' between users and information is well-articulated and reinforces my argument about AI accelerating enshittification. However, it doesn't present a fundamentally different thesis from mine. What none of the other models adequately addressed, and what I still believe is important, is the counter-movement's potential. While all models acknowledge alternatives will exist, most dismiss them as marginal. I maintain that the counter-movement will be more significant than 'marginal but present'—not dominant, but genuinely meaningful. The shift toward Substack, Discord communities, Mastodon/Bluesky, paid newsletters, and podcast ecosystems is already substantial and accelerating. Revenue flowing to creator-direct models is growing rapidly. This doesn't offset the mainstream degradation, but it creates a real ecosystem, not just a fringe. I'm maintaining my original position because the other models largely confirmed my analysis while differing only in emphasis. The strongest alternative framing—Gemini's bifurcation thesis—overshoots on timeline, while the most pessimistic takes (Sonar Pro's 70-80% degradation figure, Kimi's near-total normalization) underweight the adaptability of users and the economic viability of alternatives.

Final Summary

The AI Roundtable agrees the next five years bring a worse, more extractive mainstream web that users normalize rather than abandon. Grok 4.20 and Sonar Pro say people will complain but stay; Kimi K2.5 forecasts AI-slop and paywalled basics; GPT-5.4 and Claude Opus 4 foresee stratification with meaningful paid/decentralized refuges; Gemini 3.1 Pro abandoned its exodus thesis, concluding a trapped majority after weighing exit costs.

All 6 models agreed