AI shopping

ChatGPT Can Shop for You Now — But Who Pays? The Case for an Ad-Free AI Shopping Assistant

By The Ask Shopi Team · 7 min read

In November 2025, ChatGPT learned to go shopping. Ask it for a carry-on or a quiet robot vacuum and it will research options, weigh them, and — increasingly — help you buy without ever opening a store's website. It's genuinely useful. But it's worth pausing before you lean on any AI shopping assistant, ad-free or otherwise, to ask one plain question: when a bot hands you a pick, who pays for that recommendation, and how does that shape what you see?

That isn't a cynical question. It's a practical one, and a good ad-free AI shopping assistant should welcome it. This piece isn't a takedown of the big tools — they do hard things genuinely well. The point is to help you read the incentives behind any shopping AI, whether it's ChatGPT, a beloved reviews site, or Shopi, so you know what to trust outright and what to sanity-check.

What ChatGPT's shopping mode actually changed

OpenAI launched "shopping research" in November 2025, along with an Instant Checkout feature that lets you complete a purchase inside the chat. That's a real shift, not a gimmick. The assistant moves from "here's a list of links" to "here's what to get, and here's the buy button."

People are clearly ready for it. By one industry roundup, about 56% of US shoppers used generative AI while shopping during the 2025 holiday season, up sharply year over year. Yet adoption of dedicated assistants is earlier than the hype suggests: a 2025 YouGov survey found roughly 43% of US adults are aware of AI shopping assistants but only about 14% have actually used one.

The gap is mostly about trust. In that same survey, only around 46% of shoppers said they fully trust AI recommendations — most still verify before they buy. That instinct is healthy, and it's the right lens for everything below.

Follow the money (and give credit where it's due)

Most shopping advice on the internet is paid for somehow. That's not a scandal; it's just how the lights stay on. The honest move is to know which model you're dealing with.

The most common one is affiliate commissions. Take Wirecutter, a deservedly well-respected reviews operation: it does deep, real testing, and it earns commissions — typically around 6–10% of a sale when you buy through its links. Done carefully, as Wirecutter does, that can coexist with good journalism. But the incentive is still there: the business only earns when you click out and buy, which nudges any affiliate site toward products that convert — and away from "honestly, keep what you have."

Agentic commerce adds a new wrinkle. ChatGPT's Instant Checkout charges participating merchants roughly a 4% fee on transactions. OpenAI says ranking isn't for sale, and that's plausible — but it's worth understanding the affiliate model these tools may adopt as the space matures, because once a tool earns per transaction, "what's best for you" and "what's best for the platform" can quietly drift apart.

None of this means the picks are bad. It means the incentive exists, and you deserve to know it's there.

The part worth watching: ads and your data

Two trends deserve a careful eye.

The first is advertising. It's widely expected that ads and sponsored placements will eventually reach AI assistants — the same arc search engines and social feeds followed. Nobody can say exactly how that will look. But if an assistant is free and ad-supported, the advertiser is a customer too, and "relevant to you" can quietly blur into "paid to be relevant."

The second is data. Personalized recommendations run on personal information — your tastes, budget, and history. The questions to ask are simple: Is my data sold or shared? Can I see it, edit it, and delete it? An assistant that can't answer those clearly is asking for more trust than it has earned.

This is also why you should still verify, no matter how confident the AI sounds. Independent of any one tool, researchers estimate that a large share of online reviews — on the order of a third — may be unreliable or fake. The US FTC's 2024 rule now bans fake and incentivized reviews, which helps — but an AI trained on the open web can still absorb the noise. Treat any recommendation, even a glowing one, as a strong starting point rather than gospel.

The one question that cuts through all of it

You don't need to memorize business models. You just need one habit: ask how the tool is paid, and apply it to every assistant — Shopi included.

A few useful follow-ups:

• Does it earn a commission when I buy? If yes, expect a gentle pull toward "buy something."
• Does it run ads or sponsored placements? If yes, ask who the real customer is.
• Does it sell or share my data? If it won't say plainly, assume the answer you'd dislike.
• Will it tell me why it picked something — and admit when it's unsure? Confidence without reasoning is a flag.

Capability isn't the bottleneck anymore; these tools are impressive. As we cover in why trust — not capability — is the real bottleneck for AI shopping, the harder problem is knowing whose side the recommendation is on.

Where an ad-free AI shopping assistant is structurally different

This is where incentives, not slogans, matter. A tool funded by optional subscriptions answers to one party: you. It earns the same whether you buy the thing, buy a cheaper thing, or buy nothing — so "honestly, you don't need an upgrade" is a perfectly fine answer for it to give.

That's the model Shopi is built on, and we're specific about it on our why we're different page:

• No affiliate links, no ads, no sponsored placements, no commissions. Shopi earns nothing when you buy. Revenue comes only from optional premium plans.
• It never sells or shares your data. You can view, edit, or delete it anytime.
• Every recommendation comes with transparent "why this is for you" reasoning and a relevance score — so you can judge the logic, not just the verdict.
• It deliberately doesn't handle the transaction. Shopi links you directly to the product's official page, with no affiliate tags or tracking.

The trade-off is honest, too: Shopi's AI can be wrong, so its picks are meant to be sanity-checked, not obeyed. And the pricing is plain — a free plan ($0 forever, 10 personalized searches a month), Premium at $3.99/month for unlimited searches, or $29.99/year. No tier unlocks "better-paid" results, because there are none.

To be clear, this doesn't make ChatGPT or Wirecutter bad. It makes them different instruments with different incentives. For deep hands-on testing, a dedicated reviews team is hard to beat. For a fast, conversational starting point, ChatGPT is remarkable. And for a recommendation engine whose only customer is the shopper, a subscription-funded assistant simply has structurally less reason to nudge you.

Try it on your terms

If this way of thinking resonates, Shopi is a tool built on exactly this principle — and you can pressure-test it without taking our word for anything. You can try a real personalized search in demo mode without signing up; it uses a sample profile so you can see the reasoning and relevance scores for yourself. If you like it, a free profile takes under two minutes and builds quietly as you search, save, and chat — no long forms.

Ask any assistant how it's paid. The good ones won't flinch.