Jo — The Etsy Seller Who Stopped Answering the Same Question Twice

The 10 PM message pile

Jo makes ceramics in a studio behind her house. Mugs, bowls, small sculptural pieces, the occasional commission. She sells through Etsy and at two local markets a year. She has been doing this for six years and built it from a side project into most of her income, entirely on her own.

Being entirely on her own means she also does all the customer service, all the photography, all the listing copy, all the social media, all the packing and shipping coordination, and all the answers to the questions. The questions arrive at all hours, but they seem to peak around 10pm.

Is this food-safe?

Do you ship to Australia?

Can I get this in blue?

How long until it ships?

Is this dishwasher-safe?

She had answered every one of these questions, personally, individually, dozens of times. She answered them warmly. She answered them late at night on her phone in bed, because that's when they came in and she felt guilty leaving them unanswered until morning. She was spending roughly four hours a week on messages — and the number went up every time the shop had a good month.

The FAQ that took thirty minutes and saved three hours a week

Someone in a ceramics-sellers forum mentioned using AI to sort customer messages. Jo had heard of ChatGPT, hadn't tried it, assumed it was for tech people. She tried it on a Tuesday evening when she had forty-seven unread messages and very little patience.

She pasted in the last six months of messages she'd copied from Etsy — everything she could pull quickly — and ran this:

It came back with eight categories. Seven of them were questions she already knew she was answering. The eighth — about whether her glazes were non-toxic and safe for children — was a question she hadn't realized was as common as it was, because she'd been answering it the same way every time without noticing the pattern.

Then:

The draft it returned was fine but not her — a bit generic, slightly formal in places. She rewrote four of the eight answers and published the page that evening. The fifth question in the next week's messages was the same food-safety question, and she replied with a single link. That was the last time she answered it by hand.

Product descriptions in ten minutes instead of forty

The listing descriptions were the part of the job Jo hated most. Not because she didn't have things to say about her work — she had too much to say, and the wrong amount of space to say it in. Every Etsy listing wanted 200 words. She either overwrote (three paragraphs about the specific firing temperature she used, which no buyer cared about) or underwrote (four words: "handmade mug, food-safe glaze").

The prompt she developed over a few iterations:

The first draft always needed editing — the AI didn't know her work the way she did, and sometimes leaned into a warmth that felt generic rather than genuine. But it gave her a starting shape to work from instead of a blank page. The blank page had been costing her forty minutes per listing. The editing pass cost her ten.

The other thing she noticed: when she gave the AI accurate technical details about a piece and asked it to write for a buyer who didn't know ceramics, the output was more useful than when she tried to write it herself. She was too close to the work to see it the way a buyer would. The AI had no such problem.

Social media: five options, ten minutes, her choice

Jo had been treating Instagram like a homework assignment she was always behind on. She knew she was supposed to post. She knew photos of work-in-progress got better engagement than finished pieces. She knew she should write captions that were more than just the item name. She knew all of this and still found herself staring at a photo she'd taken three weeks ago with no caption written.

The shift was small. She stopped trying to think of the right thing to say and started generating options.

She doesn't use AI captions directly. She uses them to unblock herself.

The feeling of not being stuck was worth more than the time savings.

The traffic problem she didn't know she had

While she was going through six months of customer messages to build her FAQ, Jo noticed something she hadn't noticed in six years of selling.

The most common question, by a significant margin, was whether her mugs were dishwasher-safe.

She looked at her mug listings. Not one of them mentioned the dishwasher.

She had assumed buyers would ask if they wanted to know. She had assumed her FAQ was covering it. She had not, at any point, thought to put it in the listing itself — the place buyers were looking before they sent a message.

She updated every mug listing that week. The "is this dishwasher-safe" message volume dropped by roughly half over the following month. Her conversion rate on mugs went up in the same period, though she can't say for certain that one caused the other.

The insight had nothing to do with AI. It came from reading six months of her own customer messages in one sitting for the first time. The AI made that possible by sorting them in about ninety seconds instead of an afternoon.

What Jo will tell you

Start with your messages. Not your product descriptions, not your social media — the thing customers are actually asking. Paste in a few months of them and ask AI to find the patterns. Whatever shows up most is your first project.

Give it real details. The vaguer your input, the more generic the output. "Please write a description for a mug" produces nothing useful. "Please write a description for a 12oz mug with a speckled blue-grey glaze, thrown on a wheel, fired in an electric kiln to cone 6, food-safe and dishwasher-safe, with a generous handle that's comfortable for large hands" gives it something to work with.

Edit before you publish. This is not optional. AI doesn't know your actual shipping times, your actual materials, your actual policies. It will make plausible-sounding things up. You are the fact-checker, and you are the one whose customers will email when something is wrong.

Use it to get unstuck more than to get done. The most useful thing Jo found wasn't a faster way to produce content — it was a way to stop staring at a blank page. The first draft doesn't have to be good. It has to be something.

And pay attention to what you notice while you're editing. Jo's most useful business insight this year came from reading her own customer messages. The AI organized them. She had to look at them.