Hot take: Your approach to synthetic media is likely flawed. And it's costing you more than you think.
I've seen it hundreds of times now. In pitches from founders, in board meetings, in the marketing reports of my portfolio companies. Everyone is scrambling to adopt AI-generated media, churning out images and videos with tools like Midjourney, Stable Diffusion, and DALL-E. They're mesmerized by the technology. They show me photorealistic faces and fantastical landscapes, proud of how they’ve prompted a machine to create something beautiful. My response is always the same: “Who cares? Is it making you money?”
The silence that follows is usually pretty telling.
As someone who has built and sold two tech companies and now invests in over 200 more—including some of the foundational players in the AI space like Anthropic, OpenAI, and Scale AI—I get a front-row seat to the hype and the reality. The reality is that 99% of companies are using synthetic media as a novelty, a toy. They’re stuck in a loop of creating pretty pictures that don’t move the needle on a single important metric. They’re burning time and resources on an approach that simply doesn’t convert.
It’s time for some uncomfortable truth.
The Siren Song of Photorealism
The biggest trap I see is the obsession with photorealism. Founders and marketing teams spend countless hours trying to generate an image that is indistinguishable from a real photograph. They tweak prompts for days, run hundreds of iterations, and celebrate when they finally create a synthetic human who doesn't have six fingers.
I get the impulse. It’s a technical challenge, and it feels like a victory. But in most business contexts, it’s a colossal waste of time. Back at MovieLaLa, which was acquired by Gfycat, we were obsessed with creating the perfect promotional assets. We spent weeks with designers to get the gloss and lighting on a movie poster just right. When we started experimenting with early-stage generative tools, we fell into the same trap, trying to replicate that manual perfection. We thought more realistic meant more engagement.
We were dead wrong. We ran an A/B test. On one side, our “perfect,” manually-crafted graphic. On the other, a slightly janky, clearly AI-generated image that we’d spent about 5 minutes making. The janky one got a 30% higher click-through rate. Why? Because it was different. It broke the pattern of glossy, overproduced marketing people are trained to ignore. It was intriguing.
Your customers don’t care if your image could fool a photography expert. They care if it grabs their attention and communicates value. Chasing perfect realism is a vanity metric. The top 1% of operators aren't asking, “Is it realistic?” They’re asking, “Is it effective?”
The Silent Killers of Creative Output
Beyond the realism trap, there are a few other silent killers I see torpedoing creative strategy.
1. You’re Ignoring the Funnel: You’re generating assets in a vacuum. You have a team making cool images and another team responsible for hitting a sales number, and they barely talk to each other. A stunningly creative video is worthless if it’s used as a top-of-funnel ad when it’s clearly designed to be a bottom-of-funnel product demo. You have to map your media to the customer journey. An image for a social media ad needs to stop the scroll. A hero image on your landing page needs to build trust. A visual in an email needs to drive a click. If you don't define the job of the image, you can't measure its success.
2. You’ve Forgotten Your Brand: Your AI output looks like everyone else’s. You’re using the default Midjourney style—that glossy, ethereal, slightly-too-perfect look. Your content is generic because your inputs are generic. You haven't developed a unique visual language. A strong brand has an ownable aesthetic. You should be able to look at an image, with no logo, and know it came from that company. Most companies are just renting their visual style from the latest trending model.
3. You Have No Feedback Loop: This is the most critical failure. You generate, you publish, and you hope. That’s not a strategy; it’s gambling. Are you tracking the performance of your synthetic assets? Are you analyzing which images lead to higher conversion rates, lower bounce rates, or more time on page? Are you feeding that data back into your creative process? Probably not. You’re treating it like an art project, not a growth engine.
What the Top 1% Do Differently
I have a few portfolio companies that are absolutely crushing it with synthetic media. They aren’t just making pretty pictures; they’re building systems. Here’s what they do.
First, they focus on speed and volume over perfection. Instead of spending a week on one perfect image, they spend a day generating 100 different versions. They test different characters, backgrounds, color palettes, and compositions. They embrace the “good enough” principle. This allows them to learn exponentially faster than their competitors. While one company is still polishing a single image, my guys have already run 50 A/B tests and know what actually works with their audience.
Second, they build a proprietary style. They don't just use off-the-shelf models. They use techniques like LoRA fine-tuning on Stable Diffusion to train the model on their own brand assets, products, and existing marketing creative. They create a custom model that outputs images in their style, with their products, reflecting their brand voice. This is their creative moat. It’s defensible and unique. No one else can copy it by just writing a better prompt.
Third, and most importantly, they build a conversion-centric workflow. It looks something like this:
- Hypothesis: They start with a clear hypothesis. “We believe an image showing a happy customer using our product in a real-world setting will increase sign-ups more than our current abstract graphic.”
- Generation: They generate dozens of variants based on this hypothesis. Different “customers,” different settings, different angles.
- Testing: They use a platform to systematically test these variants in their ads, on their landing pages, and in their emails. They measure everything—clicks, sign-ups, sales.
- Analysis: They analyze the data to see what won. But they don’t stop there. They analyze the attributes of the winning images. Was it the color? The composition? The emotion on the person’s face? They tag their images with this metadata.
- Iteration: They feed these insights back into the generation process. The system gets smarter with every cycle. “The data shows that images with a blue background and a person looking directly at the camera have the highest CTR. Let’s generate 50 more of those.”
One of my investments, a small e-commerce startup, used this exact system. They went from a 1.2% CTR on their social ads to over 5% in two months. They didn’t hire a massive creative agency. They built a small, scrappy system using open-source tools and a data-driven mindset. They’re not just making art; they’re manufacturing conversions.
Stop Playing, Start Winning
Synthetic media is not a magic wand. It’s an industrial tool. It’s a machine that can mass-produce creative options at a scale and speed we’ve never seen before. But if you operate it without a clear goal, without a system for measurement, and without a feedback loop, you’re just making expensive noise.
Stop chasing the perfect image. Stop treating this like a fun experiment. Start by defining the one metric you want to move. Build a simple system to generate, test, and measure. Find a way to create a unique style that is yours alone.
This is how you go from being a part of the 99% who are playing with AI to the 1% who are winning with it. The choice is yours.
Frequently Asked Questions
What's the most common pushback you get on this?
People often push back by citing exceptions or edge cases. And they're usually right that exceptions exist. But building a strategy around exceptions rather than patterns is a losing game for most founders.
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.