AI Is Breaking The Contracts That The Internet Is Built On

High-value AI opportunities hide behind old workflows, processes, ideas, concepts, and assumptions. We must immerse ourselves in new workflows to understand how AI will disrupt products and businesses. In my last article, I explained how AI is rendering legacy business models obsolete, but that’s a micro force.

AI’s disruptions are also undermining the contracts that the internet is built on, causing macro disruptions. I will focus on internet content and social media, but variations of those themes are impacting other internet institutions. While old product categories like search, social media, and websites are being disrupted, new ones are being created.

Ready to learn more about AI product strategy? My next engaging and interactive instructor-led AI Product Management cohort starts soon. Learn more and enroll here.

I Saw A Decade Of Progress Revealed In 20 Minutes

I gave my daughter access to ChatGPT three weeks ago, and it has since become her most frequently used app in terms of screen time. It surpassed social media and game apps. I watched her play games with ChatGPT, and it reset my view of LLMs’ place in the new product landscape. The games were more interactive and personalized to her interests.

They are primitive today, but that’s a failure at the implementation layer, not a technical weakness. Higher gaming complexity is right around the corner.

Google’s Veo 3 extends this into the realms of image, audio, and video content. Simple games can be rendered in real time, and let’s face it, most game apps are very simplistic variations on a theme. Think of Candy Crush puzzles, trivia, and card games. This video shows a more complex version than what I have in mind, but it’s proof of capability. If Veo 3 can generate realistic game walkthroughs, the experience most game apps deliver can be generated today, although costs are definitely a limiting factor for the most advanced capabilities.

What if the game isn’t going the way I want it to? The model will modify it for you. NVIDIA has already built its 50x series chips to support generating parts of the game on the fly. Soon, many of the games we play will be generated in real-time, rather than running mostly on coded logic. Player prompts, preferences, and behaviors will replace some parts of that logic.

What if I want to play with someone else who isn’t available? LLMs will soon be able to simulate their gameplay and chat style. The ways AI can change the gaming experience are only limited by our imaginations and how completely we can reinvent the game.

AI is a new operating system. I said that three years ago, and we’re finally seeing how that will play out. Computers won’t run Windows or Linux. They’ll run GPT and Claude. The LLM enables a single pane of glass where users can access anything. Our keyboards and mice are obsolete. Voice and context are the new interface devices.

Breaking Contracts & Incentivization Structures

Google’s new AI search paradigm pushes links to external sites way down the page. Users no longer need to click through to a website to find a detailed answer, shop, or get access to resources. Google can monetize the content that website owners used to. Many websites are seeing their traffic plummet, and revenues are declining with it.

What’s the point of building a website if no one goes there? Why create quality content just to have another company monetize it? An increasing share of the internet is getting put behind a paywall, and out of generative search’s reach.

What if I don’t want to wait 2 years for season 3 of ‘The Last of Us’? With a walkthrough of the games and the first 2 seasons, the model can generate the rest of the show. However, all of this breaks the contracts that underpin the content creation world.

Imagine this running on the sidebar of your ChatGPT window. It’s a crude generative social media feed. With a few cosmetic and functional touches, OpenAI can curate a timeline based on your prior conversations and prompts. Why have people post and share content when ChatGPT can curate a hyper-personalized feed in whatever style, voice, or slant I want?

What if the content I want isn’t available? The model will create it for you. If I take a day off from posting on social media, ChatGPT, Gemini, and Claude can fill in for me. If I’m not discussing something you want covered, they can generate articles or posts in my voice.

From a user perspective, this is awesome. If you operate a website, app, blog, podcast, or social media community, this breaks contracts that make doing those things worthwhile. AI isn’t amazing for most builders and creators. As the incentives break down, so does most of the internet’s current foundations, but what will the next evolution look like?

A History Of Broken Contracts

Internet platforms have been breaking these contracts for years, long before the emergence of AI. Those early breaches set the stage for what’s happening now and give us a sneak peek. AI amplifies whatever trend or signal it finds. In this case, it’s accelerating the breakdown of online incentive structures.

On social media, people build large followings so everything they post is seen by that community. Something changed about 6 years ago. Social media companies stopped showing content to all of a creator’s followers. AI-curated feeds became the norm to boost ad revenue and engagement.

But that had a cost that most social media platforms don’t feel. Content with certain subjects or links can be shown to as little as 2% of the creator’s total community. That was enough to push many onto new platforms like TikTok and Substack, where monetization is built in and content restrictions are fewer.

What’s worse, LinkedIn now recommends that creators pay to boost their content so the social media company will show it to more of the creator’s followers. That’s a gut punch for many people who make their living by creating content and spent years putting in the work to build a community. If most users only scroll down to see 4-5 posts, and 1 or 2 of those are ads, the only way some creators can be seen is by boosting their content.

One of my posts from three weeks ago got almost a million impressions. I know creators with three times as many followers as I have who can’t get that number in a week. One creator I talked to for this article said,

“I spent two years grinding to get 300,000 followers. I posted daily for years and feel like I earned the right to talk to my community, and paid for that with sweat. They {LinkedIn} should have paid me. LinkedIn turns around and tells me I got to pay them!?”

What’s the point of building a massive following if it doesn’t matter? I haven’t reached the 200,000 follower mark yet, but my content is frequently seen by two to five times as many people. In the last week, I have reached almost 4 times as many people as I have followers. Some creators with fewer than 50,000 followers achieve similar numbers.

Building a large community is no longer worthwhile, and prominent creators are leaving social media altogether for alternative mediums, such as podcasts.

What’s the point of following someone if you’re going to be shown content from the highest bidder? The opposite side of the equation is breaking social media faster than a creator exodus. Creators leaving sounds serious, but it isn’t that big of a deal. When users leave, platforms collapse.

An AI Accelerated Restructuring

I left X because AI had taken over. Most of what was on my timeline was AI-generated slop. It may not seem like it, but X is fighting for its survival as users leave for a variety of reasons.

LinkedIn does a great job of keeping AI slop to a minimum, but promoted and suggested content is taking over most feeds. High-quality content will migrate elsewhere because it cannot stand out from the ads, even for creators with large communities. Platform design decisions that optimized revenue and AI-driven content curation started the cycle.

When quality content leaves, something must fill the empty space. AI-generated content will take over most social media platforms in the next two years, creating opportunities for new platforms to emerge. Human is premium, and a few platforms will use that tenet to create a competitive advantage, while most will drown in AI sludge.

Platforms like ChatGPT and Perplexity will find an opening to deliver the highest quality AI-generated content. Their generated social media timelines will be hyper-personalized. The business model requires fewer ads to be profitable, and slop isn’t incentivized. As the cost of reasoning models drops, AI social media feeds will prove to be exceptionally profitable.

Content products are branching in two directions. Short-form social media content favors a new, AI-dominated workflow catering to short attention spans. Platforms like X and Meta are investing heavily in AI to prepare for the shift. AI-curated timelines and low-quality AI slop have hurt X, but high-quality, hyper-personalized Grok-generated timelines might end up saving the platform.

Longer-form articles and podcasts favor human-dominated workflows. This is where the human is premium trend will take hold, and platforms will be able to charge more for subscriptions. High-quality long-form content will hold attention spans longer. As streaming services look like they’re reinventing cable, and news outlets lose their credibility, new platforms have an opportunity to dominate.

Workflow Reengineering

I discussed my Disruption framework in the last post. Defining the disruption reveals opportunities, but we need another step to reveal products that take advantage of the opportunity. Before we can design an AI product, we must evaluate the changes to the workflow.

This is where the concept of workflow reengineering steps forward. It’s most commonly implemented internally as business process reengineering. However, the concepts are more valuable when we reengineer customer workflows to take advantage of technical disruptions.

For social media, reengineering the workflow means taking creators out of the equation altogether. That gives the social media platform enough control to implement hyper-personalization. Platforms don’t have to hope someone is posting the perfect content for each user. They can ensure it happens at a very low incremental cost.

Workflow reengineering is as radical as it sounds. AI product designers implement disruptions in the customer workflow to improve it in a way that wasn’t possible before. In the process, we will break many institutions and contracts that the legacy platform is built on.

For generative search, workflow engineering removes the need to click through to a different website. Users don’t need to hunt through the pages to find what they’re looking for. If I want a new pair of sunglasses, the generative AI interface can recommend several options. It has access to pictures of me from iCloud or my local drive and can create customized advertisements with me as the model. I don’t need to leave the generative AI app or website to complete my purchase, either.

LLMs are slowly becoming a single pane of glass to the internet. How much of the internet will end up falling behind it is still an open question. However, there’s no going back to the old internet paradigm. Those institutions and workflows are fading because the contracts that created them have been broken. Disruption is often a destructive process, and the cycle is already picking up speed.