Google is creating a bigger and wider chasm between users and the source of data. Currently, my blog's traffic from Google searches has dropped significantly since AI Overviews launched. Where users once clicked through to read my articles, they now get their answers directly from Google's AI summary and never visit the source.
Before, it was straightforward: you searched for something, Google showed the website that had the information, and you clicked on it. Now, when you search for information, you're presented with an AI Overview that tries to answer your search query. This is fine from a user's standpoint. You had a question, now you have an answer. But who answered your question?
Google crawls the web, finds websites that have the information you need, then summarizes them neatly for end-users. The problem is, with AI summaries, you never get to see the source information. Sure, there's a small link behind a collapsible menu, but it now means you rarely click on links anymore. Links, the very thing that made the web hyperconnected, take a back seat. Long term, since users aren't clicking on links, there are fewer incentives for anyone to create content. And Google will eventually have to find a way to source content from somewhere.
But before we get there, I want to put my cards on the table. The next frontier for Google Search is video. And the technology is already here.
How Video Search Works Today
For videos, Google often presents a YouTube video in the search results and highlights the part that's relevant to your search query. You still watch the video, and there are still incentives for the person who created the instructional video to continue doing so. The creator gets views, ad revenue, subscribers, etc. The ecosystem still works.
When you search for "how to fix a leaky faucet," Google shows you a YouTube video and jumps to the 2:30 mark where the actual fix is demonstrated. You watch that person's content, maybe subscribe, maybe watch their other videos. They directly benefit. But this is just the stepping stone to something much bigger.
What happens when Google starts showing AI Video Overviews?
My Previous Predictions About AI-Generated Video
A few years back, I wrote about how YouTube uses machine learning to predict the most likely video you will want to watch. Their goal is to keep you on the platform for as long as possible. Based on your history, and that of millions of people sharing the same watch pattern, they keep you watching by recommending the most appealing next videos.
Earlier this year, I wrote that Google (through YouTube) has all the ingredients to create the perfect video for you. In my article "The Perfect YouTube Video", I explored how YouTube tracks every aspect of how you watch their video and react to it. Using the different data points you generate, they could prompt Veo (Google's video generator) to create the perfect video for you. A video so enticing that you'd have a hard time skipping it.
This might not have been possible when I wrote that article, but at the rate AI is progressing, I wouldn't be surprised if in a couple of years Veo creates video in real time.
The Technology Is Already Here
Now, Google has Genie 3, an impressive world-building model that creates a world you can navigate in real time. It operates at 720p resolution and 24 frames per second. Combine this with Veo's video generation capabilities, and you have all the ingredients needed to create real-time AI Overview videos.
Here is what Google's AI can extract from videos right now:
- Visual elements: Objects, tools, ingredients, environments, lighting setups
- Actions and sequences: Step-by-step processes, timing, hand movements
- Audio content: Narration, background music, sound effects
- Text overlays: Ingredient lists, measurements, temperature settings
- Style and presentation: Camera angles, pacing, editing choices
And then here is what they can generate:
- Realistic environments: Through Genie 3's world modeling
- Human avatars: With perfect lip-sync and natural movements
- Coherent narratives: Combining information from multiple sources
- Optimal pacing: Based on user engagement data
Putting It All Together
Let's walk through a scenario. You have some free time today, and you finally want to try your hand at baking cookies. You search for a recipe online, and Google gives you the ingredients from an Old Family Recipe, the link buried somewhere below. Now, you go to the store, buy the ingredients, and you're in your kitchen wearing your apron and chef's hat, ready to bake some cookies.
This time you Google "how to bake cookies." You're presented with a wall of text from the AI Overview listing those same ingredients you bought before. But you're not much of a chef or a reader. Instead, you want to see how the cookies will look because you're a visual learner.
What's that in the top right corner? A new Google feature? It says "AI Video Overview."
You click the button and a new window appears. It loads for just 15 seconds, and you're presented with a hyper-realistic kitchen, with an AI-generated avatar narrating the steps with perfect lip-sync and text overlays listing ingredients. The video is just 30 seconds, cutting all the fluff usually found on cooking channels. In a 30-second video that you can scrub through, you can see all the steps for baking your cookies.
Of course, at the end of the video there's a card that appears where you can click and see the source videos Google used to generate this 30-second clip. But who clicks on that? There is a rise on zero click searches.
This will be extremely convenient for users. Why waste time hearing all the fluff and a Nord VPN sponsorship when all you need is the steps to bake? But here is what will remain unseen:
- The cooking channels that spent hours creating detailed tutorials become invisible training data.
- Their personality, expertise, and hard work get synthesized into a "perfect" but soulless AI version.
- Users get their answer without ever engaging with the original source.
- No views, no ad revenue, no subscribers for the people who actually created the knowledge.
The Invisible Attribution Problem
This isn't science fiction. Yes, it doesn't exist just yet. But it's the logical next step in Google's evolution from search engine to answer engine. Just as my blog now gets fewer clicks because people read the AI Overview instead of visiting my site, video creators will soon face the same reality.
The old value exchange model of the internet is breaking down. We were used to Google sending traffic our way when we created high-quality information that helped users. As a reward, we got views, revenue, and built a following.
With the new model:
Google uses our content as training data → AI generates competing content → Users get information → We get nothing.
Sure, there will be attribution buried in a menu somewhere, just like there is for text overviews now. But when was the last time you clicked on those source links after reading an AI summary?
The chasm between users and creators isn't just widening. It's becoming a canyon. And unlike text, where you might still want to read the original article for depth or personality, AI video overviews will be so polished and efficient that there will be even less reason to click through to the source.
For video creators, what's your value when an AI can synthesize your expertise, replicate your techniques, and present them more efficiently than you ever could? The future may lie in what AI cannot easily replicate. Like live interaction, community building, unique personality, and the kind of deep, original insight that goes beyond answering simple informational queries.
I understand that it might take another leap in efficiency before these videos can be generated in real time, but the work is being done. All the major AI players are heavily investing more data centers and research to improve their product.
But first, we need to acknowledge what's happening. Google is building a world where your content fuels their answers, but your audience never finds you.
