DEV Community: Sarah Pan

AI Doesn't Recommend the Best Product. It Recommends the Best Explained Product.

Sarah Pan — Tue, 07 Jul 2026 02:52:59 +0000

A simple bubble tea experiment completely changed how I think about AI recommendations.

Last week, I asked ChatGPT a question that seemed almost impossible to get wrong.

"What are the best bubble tea brands in my city?"

Surprisingly，some of the recommended brands were companies I had never heard of before. A few weren't even available in the cities I had lived in.

After asking the same question to Claude, Gemini, and DeepSeek, I noticed something interesting that many of the same brands which I'm not familiar kept appearing.

AI Isn't Judging Your Brand

Humans recommend products because they have experiences.

But AI does none of those things.

It doesn't know whether one brand actually tastes better than another.

Instead, it tries to generate the most statistically reliable answer based on the information it can understand.

So, AI doesn't recommend the best brand. It recommends the brand it understands best.

The Experiment That Changed My Perspective

Once I realized this, I started paying closer attention.

I didn't only test bubble tea but also the restaurants, beauty brands, consumer electronics, and ravel recommendations.

Again and again, I noticed a pattern.

Brands that consistently appeared in AI recommendations usually had several characteristics:

Clear product descriptions
Well-structured websites
Consistent public information
Plenty of third-party coverage
Easy-to-understand positioning

Meanwhile, some excellent brands barely appeared at all, because AI had much less reliable information to work with.

That's when I stopped thinking about AI recommendations as opinions.

They're much closer to information retrieval problems than human preferences.

Consumers Are Already Changing Their Habits

This matters because people are beginning to use AI differently from traditional search engines.

Instead of searching "Best bubble tea near me", many people (especially the youth) now ask AI to recommend a healthy milk tea brand."

The AI becomes the decision maker before the customer ever visits Google.

And this shift is happening surprisingly fast.

Adobe reported that traffic from generative AI tools to U.S. retail websites increased dramatically during 2025, showing that consumers are increasingly using AI assistants as part of their shopping journey.

PayPal's 2025 holiday shopping research found that 61% of Gen Z shoppers had already used AI tools to help make purchasing decisions.

Even more interesting, Adobe found that shoppers arriving from AI assistants often show stronger purchase intent than traditional search visitors.

In other words: AI is replacing the discovery stage.

How to combine these works together？？？

Sarah Pan — Mon, 06 Jul 2026 09:54:40 +0000

Sarah Pan

Jul 6

AI Made Content Creation Faster. So Why Am I Spending More Time Working?The Most Time-Consuming Part of AI Video Creation Isn't Prompting

#ai #chatgpt #claude #gemini

3 min read

AI Made Content Creation Faster. So Why Am I Spending More Time Working?The Most Time-Consuming Part of AI Video Creation Isn't Prompting

Sarah Pan — Mon, 06 Jul 2026 03:55:37 +0000

When I first started creating AI-generated videos, I thought the hardest part would be writing better prompts.

I couldn't have been more wrong.

Over the past few months, I've gradually built a workflow that actually works quite well. ChatGPT is usually where I generate reference images and visual ideas. Once I have something I like, I move to Claude to polish the prompts, rethink the scene, or fix all the little details that somehow never look right the first time. Finally, everything goes into Seedance to generate the animation.

On paper, it's a pretty good workflow.

In reality... it feels surprisingly exhausting.

One moment really made me stop and think.

I had just spent almost fifteen minutes refining a prompt in Claude. The pacing finally felt right. The visual style was consistent. The camera movements were exactly what I wanted. I copied everything into Seedance, expecting to finally generate the animation.

Instead, it interpreted the same prompt completely differently.

The camera angles changed. The expressions were off. The atmosphere felt wrong.

So I started editing the prompt again.

Another time, ChatGPT generated an image that I absolutely loved, but when I tried to animate it, the framing didn't work. I had to go back, regenerate the image, adjust the composition, and then repeat the whole process.

None of these tools were doing anything wrong.

In fact, each of them is probably one of the best tools available for its specific job.

The problem was everything between them.

Every time I switched platforms, I had to upload files again, copy prompts, rebuild context, and explain my project one more time because each model "understood" the task a little differently.

At first, I thought this was simply part of creating with AI.

But after repeating the same process dozens of times, I realized something that felt a little uncomfortable.

I wasn't spending most of my time creating anymore.

I was spending it managing my AI tools.

That realization completely changed how I think about AI.

For the past two years, almost every conversation has been about building smarter models. Every new release promises stronger reasoning, better coding, better image generation, or lower latency.

And honestly, they've all improved dramatically.

But while individual models keep getting better, using multiple AI tools together still feels surprisingly fragmented.

Every new model promises to save me time.

Yet every new tool also introduces another interface, another prompt format, another place where context gets lost.

Ironically, the more capable my AI toolkit became, the more time I spent switching between it.

I don't think this is just a problem for video creators.

Developers jump between coding assistants, documentation tools, deployment platforms, and different LLMs.

Designers move between image generation, editing software, and prototyping tools.

Writers constantly switch between research assistants and editing models.

Everyone is using the right tool for the right task.

But every switch carries a hidden cost.

Not money.

Attention.

Context.

Creative momentum.

That's why I don't think the next big challenge in AI is simply building a model that's another 5% smarter.

I think it's reducing the friction between the models we already have.

Lately, I've become less interested in asking:

"Which AI model is the best?"

Instead, I keep asking myself:

"How can different AI models work together so naturally that I stop thinking about the tools altogether?"

Maybe the future of AI isn't about replacing every model with one supermodel.

Maybe it's about building workflows where every model does what it does best, while the transitions between them become almost invisible.

I'd love to hear how other creators and developers are dealing with this.

Has AI actually simplified your workflow?

Or has it quietly made it more complicated?