AI Shopping Assistant for Fashion Stores: What Chat Can Do

An AI chat assistant for a fashion store works best when it answers the same practical questions your staff already answer by hand: “Will this fit me?”, “Do you have size M in navy?”, “When will it arrive?”, and “What can I wear this with?”. It works worst when it is treated as a general-purpose chatbot that has no product data and no boundaries.

This guide is written for owners, marketers and e-commerce staff running online clothing stores on Shopify, WooCommerce, Haravan, Sapo or a custom site. You will learn what chat can do, what it should hand to a person, how to feed it accurate data, how to place it, and how to measure whether it is helping or hurting.

What an AI shopping assistant can reliably answer in a fashion store

The assistant should not try to do everything. In a clothing shop, it is most useful when it answers four types of questions precisely and from your own data.

Size and fit questions

This is the highest-value use case for fashion chat. A shopper usually wants to know whether a dress, pair of trousers or shirt will fit. The assistant should not guess. It should:

• ask for the shopper’s height, weight, bust/waist/hip or usual size where relevant;
• look up the product’s size chart and detailed measurements;
• check product notes such as “runs small”, “loose fit”, “stretchy fabric” or “model wears size S”;
• recommend a size based only on those inputs.

A safe answer looks like this:

“Based on your waist 72 cm and hip 96 cm, the size M in this skirt should fit closer to the waist and relaxed through the hip. The size L would give more room. The size chart shows M = waist 70–74 cm, hip 94–98 cm. If you are between sizes and prefer a looser fit, choose L.”

That answer is specific, uses the chart, and gives the shopper a decision rule.

Stock, variant and availability questions

Chat can answer “Is size 38 in black available?” if it is connected to your inventory or if your product feed is synced. On Shopify and WooCommerce, many chat tools can read live stock by variant if the integration supports it. On Haravan and Sapo, you may need to sync stock to a sheet or let the assistant use your product export data.

If live stock is not connected, the assistant should still answer from a recent stock file and add: “Stock can change quickly. I can confirm in the warehouse if you want.”

Shipping, payment, returns and order status

Policy questions are repetitive and easy to automate. The assistant should answer from a single approved policy source: delivery time by region, cash on delivery availability, return window, return conditions, exchange rules, and how long refunds take.

Do not let the assistant improvise legal or payment wording. For Vietnamese stores, consumer-protection and platform rules can change; the assistant’s policy text should be reviewed against the current official source before you publish it.

Style and pairing questions

A fashion assistant can suggest items from your catalog if your products have structured attributes: category, colour, occasion, fabric, sleeve length, neckline, fit, season, and style tags. It can say:

“This linen midi dress pairs well with the flat leather sandals in the same product collection and our straw tote. Both are available in neutral colours.”

The assistant should only recommend items that exist in your store and should not invent product names or links.

What the assistant must hand to a person

Some conversations should never be fully automated. Decide these handoff triggers before you turn the assistant on.

Hand the chat to a person when:

• the shopper wants a refund, exchange or compensation for a delayed or damaged order;
• the order has already been placed and needs to be changed;
• the shopper is angry, threatening a chargeback, or reporting suspected fraud;
• the size advice is unclear because the product measurements conflict or the body measurements are outside the chart;
• the question involves medical, skin-sensitivity or safety claims;
• the shopper asks for a discount, bulk price or custom order;
• the assistant has failed twice to answer the same question.

A good handoff message is short and does not make the shopper repeat themselves:

“I can pass you to Mai from our support team. She can check the order and help with the exchange. Please wait a moment.”

Then the assistant should attach a short summary: product, customer question, size advice given, order number if any, and the point where the chat stopped.

Photo: Yaroslav Shuraev / Pexels

How to feed the assistant accurate product and policy data

The most common reason a fashion chat assistant gives bad answers is not a weak AI model. It is that the store gave it thin, outdated or contradictory data.

Product data that matters

At minimum, each product should have:

• full product title with style or product code;
• fabric composition and care instructions;
• fit note: true to size, runs small, oversized, stretchy, stiff, lined, sheer;
• size chart or flat garment measurements;
• model height and the size the model wears, if you have it;
• stock status by variant;
• product-specific notes such as “elastic waist”, “no stretch”, “buttons at back”;
• links to the size guide, shipping page and returns page.

Good product data:

“The shirt is 100% cotton poplin, non-stretch, true to size. Model is 170 cm and wears size S. Size S chest 96 cm, shoulder 39 cm, length 67 cm.”

Bad product data:

“Nice shirt. Fits well. Imported.”

Policy and operational data

Create one source of truth for:

• delivery times by region and carrier;
• free shipping threshold if you have one;
• cash on delivery rules if you use COD;
• return window and condition;
• exchange process;
• refund timeline;
• who pays return shipping;
• contact channels and support hours;
• holiday or campaign exceptions.

If you sell in Vietnam on Haravan or Sapo and also sell internationally, keep separate policy blocks per market. The assistant should ask which country the shopper is in before quoting shipping.

Where to store the data on common platforms

Shopify

• Put product-specific fit and fabric notes in the product description or product metafields. Many chat apps can read metafields.
• Create a size guide page and link it from the product page.
• Keep shipping, returns and contact policy pages up to date in Settings > Policies or the equivalent area.
• If the chat tool supports live inventory, connect it through the app’s integration.

WooCommerce

• Use product attributes or custom fields for fabric, fit, measurements and model size.
• Add the size chart to the product description or a custom tab.
• Keep shipping zones and return policy pages current.
• If the chat plugin cannot read your inventory directly, export a daily product stock file and upload it to the assistant.

Haravan and Sapo

• Put product data in the product description fields so the assistant can read the same page content the shopper sees.
• Add policy and size-guide pages in the store admin.
• Many Haravan and Sapo themes allow you to paste HTML or script tags into the theme header, footer or a custom block, so a chat widget can be installed without a native app.

Maintenance checklist

Review every week or after any product or policy change:

• Remove discontinued products from the assistant’s knowledge base.
• Update stock if the assistant uses a file.
• Update delivery times before sales campaigns or holidays.
• Check the return policy wording.
• Update size charts when you change supplier or product batch.
• Test five common questions after every update.

Good and bad chat answers for a clothing store

The difference between a useful assistant and a trust killer is often one sentence.

Question: “I’m 160 cm and 58 kg. What size should I take in this high-waist trousers?”

Good answer:

“Based on the size chart and your measurements, size M should fit: waist 70–74 cm, hip 96–100 cm. The trousers have a high waist and no stretch. If your waist is 74 cm or more, size L may be more comfortable. You can check the full chart here.”

Bad answer:

“I think M or L. Most customers choose L.”

Question: “Does this dress run small?”

Good answer:

“Yes, this style runs small because the bodice has no stretch and the waist is fitted. If you are between sizes, choose one size up. Size S bust 84–88 cm, waist 66–70 cm.”

Bad answer:

“It fits true to size probably.”

Question: “Can I return this if it doesn’t fit?”

Good answer:

“Yes, you can return unworn items with tags within 7 days of delivery. Return shipping is paid by the customer unless the item is faulty. I can send the return form if you want.”

Bad answer:

“We accept returns, but it depends.”

When you configure the assistant, instruct it to:

• ask one question at a time;
• never invent a size or policy;
• always link to the relevant size chart or policy page;
• say “I don’t know, I’ll ask a human” when the answer is unclear;
• never make medical or safety claims;
• use the shopper’s language and keep answers under four sentences where possible.

Photo: Liza Summer / Pexels

Where to place the assistant and how to keep it useful

A chat assistant that appears too early can interrupt browsing. A chat assistant that is hidden on product pages misses the moment shoppers need sizing help.

Placement by page

• Product page: show the chat after a short delay or when the shopper clicks a size question. Prioritise size, fit, stock and delivery.
• Cart page: focus on stock, delivery time, payment and simple objections such as “Will it arrive before Friday?”.
• Checkout: only if you have order-status or payment support. Do not let chat slow down purchase.
• Order confirmation and contact page: helpful for order changes and delivery follow-up.
• Homepage: keep it quiet. Most shoppers do not need to chat immediately.

Basic installation approach

Most chat tools are installed with a script tag or plugin.

Shopify

• Use the chat provider’s Shopify app, or paste the script into the theme’s app embed or theme code. Put the widget on all pages, then use display rules to hide it until the product page if the tool supports that.

WooCommerce

• Install the chat plugin, or add the script via a child theme’s functions.php or header template. Use conditional logic to show different prompts on product, cart and thank-you pages.

Haravan and Sapo

• If the chat provider gives a JavaScript snippet, paste it into the theme header or footer area. If your theme has a custom HTML block, you can add the script there. Use the page-level editor to hide it from pages where it should not appear.

Measuring an AI chat assistant: metrics that matter

Do not only track “number of chats”. Measure whether chat improves the shopping experience and reduces employee workload without creating wrong-size sales.

Worked example: if a shop has 120 chats in a week, 75 are fully handled by the assistant, 30 are handed to a person, and 15 are abandoned before answering. Containment is about 63%. That number is only useful if answer accuracy is checked. Spot-check 20 transcripts; if 4 contain a wrong or unsafe size answer, fix the product data or handoff rules before scaling.

On Shopify and WooCommerce, you can often tag orders with a chat session ID or source if the chat tool supports events or custom fields. For Haravan and Sapo, you may track manually with a Google Sheets log or a link the assistant sends as a coupon code per channel.

Common mistakes that make fashion chat assistants useless

1. Turning it on without size data. The assistant guesses or gives generic advice, and the store gets returns.
2. No handoff path. A shopper with a failed delivery gets stuck in a loop.
3. Letting the assistant answer return and refund complaints. These need human judgement.
4. Using unrelated product descriptions. The assistant reads “beautiful, elegant, trendy” but no measurements.
5. Measuring chat activity instead of outcomes. More chats can mean more confusion, not more sales.
6. Hiding the assistant on mobile. Most fashion shoppers are on mobile; a widget that covers the Add to cart button is a conversion killer.
7. One policy for all markets. A store selling from Vietnam to Europe must not quote domestic COD terms to an international shopper.

A chat assistant can answer size and policy questions, but it cannot show a shopper how the garment looks on their own body. On the product page, the strongest setup is usually both: chat for questions, and a visual try-on button next to “Add to cart” for the moment a shopper wants to judge the look before buying. If you want to add that visual layer, Macgi offers an AI virtual try-on widget for clothing stores that works from your existing product photos. You can try Macgi’s free beta on Shopify, WooCommerce, Haravan, Sapo or a custom site.

If you are still improving your sizing help before adding chat, start with How to Help Customers Choose the Right Size Online for Boutiques and Customer Service for Online Clothing Store: Small Shop Standards. For turning chat conversations into actual orders, see How to close sales when selling clothes online via chat: 5 steps.

FAQ

How much product data do I need before turning on an AI chat assistant?

Start with at least: product title, fabric, fit note, size chart, stock status, shipping policy, return policy and contact hours. You can improve the assistant weekly. Do not turn it on with only marketing descriptions.

Should the assistant handle refund requests?

No. Refund and exchange requests, especially after delivery or a complaint, should go to a person. The assistant can collect the order number and explain the return window, but it should not promise refunds or compensation.

Can I use a general AI chatbot for my fashion store?

Yes, if you can connect it to your own product and policy data and set strict instructions. A general chatbot without your size charts and policies will either answer vaguely or make things up, which is riskier for a clothing store than for many other product types.

How do I prevent wrong size advice?

Give the assistant your actual size charts, flat garment measurements and product fit notes. Instruct it to ask for body measurements, never guess, and say “I don’t know” when the data is unclear. Check a sample of chat transcripts weekly.

Does the assistant need to work in Vietnamese and English?

If you sell to both local and international shoppers, yes. Configure the assistant to detect the shopper’s language and serve the correct policy data. Vietnamese shoppers on Haravan and Sapo often prefer short, direct answers; international shoppers may expect more detail.