The difference between a vague ChatGPT shopping answer and a useful one is almost always the prompt. These are the prompts I use with Phia in ChatGPT to find a specific item, check whether it is cheaper pre-owned, and style a piece with in-stock products, plus the habits that make each answer sharper.
ChatGPT Prompts for Fashion Shopping That Find Real Products
Talk to it like a personal shopper
Every prompt here works with the Phia plugin in ChatGPT. And the Phia app keeps your closet, so your prompts can start from what you already own.
Betty Halbreich started at Bergdorf Goodman in 1976, and the store built a personal shopping office around her. Her first client was Babe Paley, and for more than four decades she dressed socialites, actors and women who walked in off Fifth Avenue. What made her good was not a secret catalogue. It was that she asked the right questions, and told people the truth.
ChatGPT prompts for shopping work the same way. The model is only as useful as what you tell it. Ask it for a nice coat and you get a nice paragraph about coats. Tell it you need a camel wool coat under $400, single-breasted, long enough to cover a midi skirt, and you get something you could buy this afternoon.
With the Phia plugin in ChatGPT, a good prompt also gets you real products. Phia searches fashion, bags, shoes and accessories across hundreds of retailers and resale marketplaces, and its answers arrive as cards with a photo, price, store and Shop button. The prompt is what decides whether those cards are the right ones.
What I would skip: asking ChatGPT what is trending and buying from the answer. Trends are fine as a starting point. Your closet, your budget and your week are a better one.
Start with @Phia, then be specific
The single most useful habit is typing @Phia at the start. When you type the @, ChatGPT lists the plugins you have added, and choosing Phia tells it to answer with Phia's catalogue rather than its general shopping results. If you have not added it yet, search Plugins for Phia or open the Phia plugin page.
Then give it the five things a salesperson would ask. What is it, exactly? What color and material? What is your budget? What is it for? And do you want it new, pre-owned or either? A finding prompt I like: @Phia find a black leather tote that fits a laptop, under $500, new or pre-owned. A narrower one: @Phia cream cashmere crewneck under $150.
You can stack constraints and refine in the same chat. Make it cheaper. Show me only pre-owned. Only items on sale. Different color. Phia can filter by price, gender, condition and sale items, so these follow-ups change the results rather than just the wording.
What I would skip: brand-only prompts like "Loewe bag". Name the model, Puzzle or Flamenco or Hammock, and the size if you know it.
The prompts I actually use
Most of my shopping questions fall into three prompts, and I have stopped trying to be clever beyond them.
Find it. @Phia find black pointed kitten heels under $150. @Phia a navy wool blazer that is not double-breasted. The answer is a row of products with prices and stores, so you can see the range before you click anything.
Compare it. @Phia is the Loewe Puzzle bag cheaper new or pre-owned? When I asked that, Phia's results put a pre-owned small Puzzle at $1,505 at Rebag against a new Mini at $2,985.26 at Bluefly, and ChatGPT pointed out that the sizes differed, which is exactly the kind of catch you want. @Phia what does a Bottega Veneta Jodie cost secondhand? works the same way.
Style it. @Phia what should I wear with a black leather biker jacket for a dinner date? @Phia outfit ideas with brown ballet flats for fall. Phia's stylist returns the look grouped by pairing, two in-stock options for each part, and links to the Phia edits behind the advice, like Brown Ballet Flats Outfits for Fall.
What I would skip: combining all three into one message. Ask one, read the answer, then follow up.
Make the follow-up do the work
The first answer is rarely the one you buy from. It is the draft you react to. The follow-ups are where a good shopping chat earns its keep: that is too formal, what if it is cold, I already own black jeans so skip those, show me the same thing under $100.
I always tell it what I already have. Not my whole closet, just the relevant pieces. I own a camel coat and dark jeans, what shoes and bag would finish it? That one sentence stops the answer from suggesting things you have, and it makes the outfit feel like yours rather than a mannequin's.
If you want that context without typing it each time, that is what the Phia app's closet is for. It builds itself from your order emails and the photos on your phone, then its Style analysis tells you your top colors, vibes and brands, and you can add up to ten styling rules you swear by. Knowing your own top colors makes your prompts better too. I explained how that works in What Is My Style? Find It in Your Digital Closet.
What I would skip: starting a fresh chat for every follow-up. Stay in the same one so it remembers what you have already ruled out.
Before you tap Shop
Phia shows you where to buy, and Shop opens the product on the retailer's own site, where you check out. Phia does not take payment or place orders, and it does not need an account or personal information in ChatGPT. Prices and availability come from the stores and can change, so the price on the product page is the one that counts. Right now the plugin works in the United States with prices in US dollars.
My last check is always the same three things: is my size actually in stock, what is the return window, and for anything pre-owned, what is the condition grade. Ten seconds of checking saves a lot of return labels.
If you are new to all of this, add the Phia plugin in ChatGPT, try one prompt from each of the three kinds, and see which one you reach for again. Then download the Phia app so your closet can start building itself in the background.
