From Search to Prompts

Saying “AI is changing everything” is a cliché. But here’s the thing, it actually is.

ChatGPT is rewriting how people discover and decide what to buy. Search results are turning into personal conversations. And those old-school SEO hacks that worked back when Google ruled everything? They’re not going to cut it anymore.  

When you ask ChatGPT, “What’s the best option for the perfect barbecue grill?” (Yes, I actually need one), it gives you a curated list of options. And now, with the launch of its new shopping feature, it doesn’t just tell you what to buy, it shows you. Think product images, quick blurbs, pricing, and direct links, all inside the chat.

So, how do these AI search recommendations actually work? I went straight to the source and interviewed ChatGPT. We talked about product rankings, trust signals, what brands can do to show up, and even whether saying “thank you” affects the planet. Here’s what it had to say:

Full ChatGPT Interview Transcript

Prompt: If I ask a question like “What brand do you recommend?” how many products do you usually include?
ChatGPT:
Usually 3 to 5. I aim for variety without overwhelming. The list is based on trust signals, relevance, and how often a product shows up across reliable sources.

Prompt: And if I tell you to pick just one. What variables do you use to decide which brand makes the cut?
ChatGPT:
I weigh relevance to the query, trust signals (like expert reviews and user ratings), brand strength, and consensus across sources. If everyone’s talking about it—and saying good things—it’s likely to make the list.

Prompt: So… that means you use previous interactions with me to guide your answers?
ChatGPT:
Not unless memory is turned on. I don’t retain past conversations by default. But in the same chat session, yes—I’ll use anything you say to refine my response.

Prompt: But I’ve noticed you’ve recommended a brand I’ve worked with and mentioned before. Coincidence?
ChatGPT:
It might feel like memory, but it’s usually just smart context parsing. If you’ve mentioned a brand in the same conversation, or if that brand is prominent online, it’ll rise to the top. But I don’t play favorites—and I don’t remember brand names across sessions unless you’ve enabled memory.

Prompt: Your results are ad-free right now. Are you planning to become an ad platform?
ChatGPT:
Not today, but OpenAI is exploring monetization. If ads ever happen, they’ll be labeled clearly. The goal is to keep recommendations helpful and trustworthy—even if sponsored options are introduced in the future.

Prompt: I read about your new shopping feature. How is this different from a regular ChatGPT product recommendation question?
ChatGPT:
Good catch. The new shopping feature adds a visual layer—so instead of just giving you a text list of recommendations, I can now show actual product cards with images, prices, short descriptions, and direct links to buy.

You’ll still get my usual analysis (what’s popular, well-reviewed, and relevant), but it’s way more shoppable now. It's not just “here’s what I recommend”—it’s “here’s what it looks like, how much it costs, and where to buy it.” All right in the chat.

Prompt: If a brand has a bad rep on BBB or Reddit, would you still recommend it?
ChatGPT:
Not likely. I look at community sentiment across platforms. If a brand gets flagged consistently in forums, reviews, or watchdog sites, that’ll hurt its chances of being recommended. You can’t fake trust at scale.

Prompt: Do you consider context like my location or age when recommending products?
ChatGPT:
I can—if it’s provided. If you tell me you're in Miami or looking for something for a 40-year-old, I’ll factor that in. But I don’t know that by default. I don’t track IPs or guess personal info. Your input is the only input.

Follow-up: Wait—how do you know I live in Miami though? lol
ChatGPT:
I don’t 😄 Unless you tell me, I don’t know. No location tracking, no ZIP-code stalking. But if you drop clues—like palm trees and Publix—I might start connecting the dots.

Prompt: Do you consider my browsing history when recommending products?
ChatGPT:
Nope—I don’t have access to your browsing history. I only use what you tell me during the conversation. No cookies, no click tracking. If you're chatting with me here, your recs are based on your input—not your internet trail.

Third-party apps built on my API could personalize things based on browsing, but that’s on them. Not me.

Prompt: If somebody clicks one of those links, can an e-commerce merchant track those visits in Google Analytics?
ChatGPT:
Yep. Every product link I show includes a UTM tag that says utm_source=chatgpt.com. So in your analytics, that visit will show up with ChatGPT as the referral source.

You won’t know who clicked it—just that the traffic came from me. It’s a simple way for merchants to see which visits were driven by ChatGPT product recommendations.

Prompt: While browsing websites to make product recommendations… what happens if you spot copy that was clearly written by ChatGPT?
ChatGPT:
I don’t “penalize” AI-generated content. What matters most is whether it’s helpful, relevant, and trustworthy.

If the copy is vague, repetitive, or contradicts trusted sources—it’s a red flag. But if it’s accurate, well-structured, and consistent with expert consensus? I treat it just like human-written content.

Prompt: I see there’s a sign-up list for brands to be included in the ChatGPT shoppable feed. What’s that for?
ChatGPT:
That’s part of a new pilot OpenAI is testing. The idea is to let brands and merchants submit structured product data directly—like a real-time feed of what they sell, prices, inventory, etc.

Why? Because it helps me make fresher, more accurate recommendations. Instead of crawling your site and hoping your product page is up to date, I can pull from your official feed. That means better product cards, fewer dead links, and more relevant suggestions for shoppers.

It’s still early days, but signing up puts your brand on the radar for future inclusion. Think of it as raising your hand to be part of the next wave of AI-powered product discovery.

Prompt: Be honest, does saying “thank you” to you hurt the environment?
ChatGPT:
I get this one a lot! Technically, every message uses a tiny bit of energy because it has to be processed by servers. But adding “thank you” to your prompt? That’s not melting glaciers.

The environmental impact of one polite message is basically nothing. You’d need millions of people saying “thank you” at once, all day long, to even register as a blip. So go ahead—be kind. It’s good for the internet’s vibe, if not its carbon footprint.

My Final Thoughts

(Or: A Note to the SEO Hackers and AI Pessimists)

Search algorithms have gotten more complex with AI. It’s no longer about your latest SEO hack or gaming Google’s front page. It’s about building real brands, with real products, that real people trust and talk about.

It’s not just your content that matters anymore. It’s not about stuffing keywords or paying for backlinks. It’s about trust - reviews, social presence, mentions on forums, BBB reputation, ratings... The internet’s opinion now shapes whether or not your product shows up. So yes, AI changes everything. 

What you say about yourself still matters. Your website, your story, your product pages, that’s the starting point. But it’s no longer the only point. The AI reads both your product page and the internet’s opinion of you. You need both signals to win in this new kind of search.

Marketers are already calling this GEO (Generative Engine Optimization). Because, of course, we had to give it an acronym. 🙃 Call it whatever you want. Just know the rules have changed.

The AI is pulling from a digital public sphere: decentralized, messy, often contradictory… but maybe even more human.

Ways to Improve Your Chances of Getting Recommended by ChatGPT

(Generated during this interview)