Keep one spreadsheet row per garment holding its measurements, condition grade and every flaw, then have an AI assistant turn those rows into descriptions and a five-post countdown. Schedule the posts in Instagram's own scheduler, shoot all pieces from one shot list, and approve every claim yourself before the drop goes live.
The usual failure is asking an assistant for "vintage-sounding" copy. Given a vague brief, it will supply a decade, a maker and a backstory that nobody gave it, and none of those survives the first buyer who asks where the claim came from. The cure is to hand it facts rather than a mood, so that every claim in a listing traces back to a cell the seller filled in with the garment on the table.
Where this seller's eleven hours went before the sheet
Take a one-person shop that lists 24 pieces every Thursday at 7pm, on its own shop page and on Instagram. Every number here is an illustration for planning, not a measurement, so time your own week with a stopwatch before you trust any of them.
Before the sheet, the week looked like this. Measuring, grading and noting each piece took about two hours, scribbled on paper. Photographing and editing took three and a half. Then came Wednesday night: 24 descriptions at roughly ten minutes each, four hours in all, plus an hour and a half of countdown posts. That is eleven hours, and the worst of them are the final five and a half, written when the seller is tired. By description fifteen every piece was "a lovely example in great condition", and flaws were recalled from memory rather than read from a note.
Writing and posting are the two jobs an assistant can take most of. Measuring, grading and photographing stay with the seller, because that is where the truth about each garment enters the system. Everything below rests on that split: the seller records what is true, the assistant turns it into words, and the seller reads the words against the garment before anyone else does.
One row per garment: the fields that stop invented history
The spreadsheet is the single source of truth. The assistant reads rows; it is never asked to look at a garment and guess. An empty field means "say nothing", not "make something up", and that rule is what keeps invented eras and makers out of the copy. Resellers who list second-hand stock at volume use the same one-row-per-item habit, and the tutorial on how resellers list second-hand items faster covers the general case. The vintage version needs these fields.
| Field | What you enter | What the assistant may do with it |
|---|---|---|
| ID | Drop number plus item number, such as D0612 | Appears in photo file names and the listing reference |
| Category and colour | The garment type, and the colour as you see it in daylight | Use as given; add no extra style detail |
| Label text | Exactly what is printed or woven, or "name unreadable" or "no label" | Quote it; never name a maker from the cut or style |
| Fibre and care label | From the care label only, such as "55% wool, 45% polyester; dry clean only", or "no care label" | Repeat it; never guess a fabric or a care instruction |
| Tagged size and flat measurements | The size on the tag, then measurements taken flat in the same order every time | Copy them exactly; no sums, no conversions |
| Era estimate and evidence | "Probably 1980s", plus what points to it: shoulder pads, fastener type, care label style | Write an era only if the evidence cell is filled |
| Grade and flaws | A, B, C or R from the ladder below; each flaw on its own line with size and position | Quote the grade wording; list every flaw |
| Cleaning | "Dry cleaned", "hand washed and steamed" or "not cleaned" | Repeat it; say nothing about smell unless you wrote it |
| Documented provenance | "None known" by default; add only what you could show a buyer | Mention origin only when this cell holds something |
| Price | Yours, set before the assistant sees the row | Never suggested, never changed |
Here is one filled row, exactly as it gets pasted into the chat. The era cell holds an estimate with its evidence, and the provenance cell says none known, which tells the assistant to write nothing about where the blazer came from.
ID: D0612
Category: blazer, double-breasted, lined
Colour: navy
Label text: woven maker's label, name unreadable
Fibre and care label: 55% wool, 45% polyester; dry clean only
Tagged size: 14
Flat measurements (in): pit to pit 20.5; shoulder 17; sleeve 24.5; length 29
Era estimate: probably 1980s. Evidence: shoulder pads, wide lapels, style of the care label
Grade: B
Flaws: 1) faint mark below left pocket, about 8 mm; 2) one replaced button, not an exact match
Cleaning: steamed and aired
Documented provenance: none known
Price: [set by you]A condition ladder the assistant quotes instead of improvising
"Great condition" means one thing to the seller and another to the person opening the parcel. Write a four-step ladder once, keep it in the standing instructions and make the assistant quote it word for word.
| Grade | What it means in this shop | Wording the assistant quotes |
|---|---|---|
| A | No flaw found at arm's length or in close-up, inside or out | Excellent vintage condition. No flaws found. |
| B | Light wear only; every flaw under 1 cm, or a minor repair; all listed | Very good vintage condition. Light wear, every flaw listed below. |
| C | Flaws a buyer would notice; each listed, photographed and reflected in the price | Good vintage condition with visible wear, listed and photographed below. |
| R | Needs repair, or sold as found | Sold as found and needs repair; see flaws. |
Two word lists travel with the ladder. "Mint", "perfect", "flawless" and "pristine" are banned in the prompt, because "Excellent vintage condition. No flaws found." says the same thing more credibly. "Character", "patina" and "lived-in" are banned as well: they are what a flaw sounds like when somebody would prefer not to name it. Where a flaw is minor, the sheet gives its size. A 2 mm hole is minor; "small hole" is not information. For fabric, fit and care errors that assistants make in clothing descriptions generally, the checklist of fabric, fit and care errors is a good companion.
The standing prompt, and what it returned for one blazer
Keep the standing instructions in a ChatGPT Project or a Claude Project, so each week you paste only the new rows. Custom GPTs are the wrong home, because OpenAI stops them running on 11 December 2026. On Claude, the Free plan allows up to five projects, instructions included. ChatGPT Plus is $20 a month and Claude Pro is $20 monthly or $17 a month billed annually; both services also have free tiers you can test on a few pieces first.
You write listings for a one-person vintage clothing shop. Use only the facts in the row I paste. If a fact is not in the row, it does not go in the listing.
Rules
1. Never state a decade, maker, country, fabric or backstory that is not in the row. If "Era estimate" has no evidence beside it, write nothing about age. If "Documented provenance" says "none known", write nothing about where the piece came from.
2. Copy measurements exactly as given, in this order: pit to pit, shoulder, sleeve, length. Do no sums and no conversions.
3. Use the wording for the piece's grade from the ladder below. List every flaw in the "Flaws" cell, one per line, numbered to match the photos. Never soften a flaw with words such as character, patina or lived-in.
4. Never use: mint, perfect, flawless, pristine, rare, authentic, investment.
5. Format: a description of 40 to 60 words, then Size, Measurements, Condition, Flaws and Care lines.
6. If you are unsure whether a fact is in the row, ask me instead of guessing.
Ladder
A: Excellent vintage condition. No flaws found.
B: Very good vintage condition. Light wear, every flaw listed below.
C: Good vintage condition with visible wear, listed and photographed below.
R: Sold as found and needs repair; see flaws.Fed the row for the blazer, the assistant returned the draft below. It is tidy, usable and wrong in two places.
Navy wool-blend double-breasted blazer with padded shoulders and wide lapels, lined, with a nipped waist. A woven maker's label is inside but the name can't be read. Probably 1980s, judging by the padded shoulders, the lapels and the style of the care label.
Size: tagged 14
Measurements (flat, inches): pit to pit 20.5, shoulder 17, sleeve 24.5, length 29
Condition: Very good vintage condition. Light wear, every flaw listed below.
Flaws: 1. Faint mark below the left pocket, about 8 mm (photo 5). 2. One replaced button, not an exact match (photo 6).
Care: Dry clean only, per the care label. Steamed and aired; no smell.The description says "nipped waist", and nothing in the row mentions the waist; the assistant borrowed it from the way blazers like this are usually described. The care line says "no smell", and the row only says the piece was steamed and aired. Expect slips like these in any first draft, which is why the read-through before posting is not optional. After a catch, add a line to the standing instructions: "Describe only features named in the row, including cut and fit."
When a draft softens a flaw: before and after
The slip that costs most is a flaw wrapped in flattering language. Three entries from the same drop show it, and the right-hand column shows the line after the euphemism rule went into the prompt.
| What the sheet says | First draft said | After the rule |
|---|---|---|
| Moth hole, 2 mm, edge of left cuff lining | "A touch of character at the cuff" | Flaw 2: moth hole about 2 mm at the edge of the left cuff lining (photo 7) |
| Pilling across both sleeves, light | "A soft, well-loved knit" | Flaw 1: light pilling across both sleeves (photo 4) |
| Underarm discolouration, both sides, faint | "Some underarm wear typical of its age" | Flaw 3: faint discolouration under both arms (photo 8) |
A second pass catches whatever the rule misses. Before posting, paste the finished description back into the chat with the sheet row and ask the assistant to check itself:
List every flaw in the row. Next to each one, quote the words in the description that cover it. Write MISSING for any flaw the description does not state plainly.This takes seconds per piece and turns the read-through into a check against a list instead of a feeling. A MISSING means the description is wrong. A clean pass means only that the words match the notes; the notes still have to match the garment, which is why the seller's own eyes come last.
Flat measurements: one place does the doubling
Here is a mistake to expect, because two helpful instructions can collide. The sheet had a formula that turned each flat pit-to-pit figure into a chest measurement, and the prompt also said "give chest as well". The assistant doubled a figure that was already doubled, and the listing for the blazer read "chest 82 in". Nobody noticed until a buyer asked whether it was a typo. The flat figure was 20.5 in, so chest is 41.
The fix is one rule: arithmetic happens in the sheet and nowhere else. The prompt says "copy measurements exactly as given, do no sums and no conversions". The sheet holds the flat figure (20.5), a doubled figure (=B2*2, giving 41 in around) and a metric figure (=ROUND(B2*2.54,0), giving 52 cm flat), and the description quotes them together as "pit to pit 20.5 in flat (41 in around, 52 cm flat)".
A sense-check column catches the rest. Set limits from your own size chart and flag anything outside them, for example =IF(AND(C2>=34,C2<=48),"ok","check") where C2 holds the doubled figure for a jacket. Vintage pieces rarely match their tagged size, so both numbers go in the listing: the tag tells the buyer what the maker called it, and the tape measure tells them whether it fits.
Words that need evidence before they appear
Some words carry promises, and an assistant will use them freely unless the sheet stops it. Each word below needs evidence in a cell before it may appear in a listing or a post.
| Word | What you need first | Safer wording otherwise |
|---|---|---|
| Vintage | An age that qualifies; Etsy's rules, for one, draw the line at 20 years | "Second-hand" or "pre-owned"; "vintage-style" for newer pieces |
| Deadstock | Original tags, or unworn old stock you can show in a photo | Say "unworn" only if it is true; otherwise say nothing |
| A maker or designer's name | A label you can read in a photo | "Label name unreadable" |
| Rare | Proof you could show a buyer | Leave it out |
| Authentic | A way to authenticate the piece | Leave it out |
| Handmade | Knowledge of who made it | "No label; maker unknown" |
| Sustainable, eco-friendly | A claim you could substantiate | "Second-hand, so reused rather than made new" |
Provenance is the claim that tempts most. If the seller was told a story, the sheet records it as the seller's account: "bought from a seller who said it came from a theatre wardrobe" is accurate, while "from a theatre wardrobe" is not. On Etsy the vintage rules go further. Etsy's house rules for vintage items say an item must be at least 20 years old, and Etsy can ask where you sourced it, how you worked out its age and for photographs of the details that show it. A sheet with an era-and-evidence column is already the answer to those questions.
One more category deserves a line. If you have customers in the EU, generic environmental claims such as "eco-friendly" fall under EU Directive 2024/825, which has applied since 27 September 2026 and bans such claims unless you can show recognised excellent environmental performance. "Second-hand, so reused rather than made new" says something true without a claim you would have to defend. Ask a solicitor if you are unsure where a particular phrase falls.
One shoot, one shot list, a checklist per garment
Photography is the step where AI helps least and consistency helps most. Shoot all 24 pieces in one block, in daylight from one window with no flash, against the same backdrop, with the camera at the same height and distance each time. Lock the white balance if your phone allows it; a colour that drifts from piece to piece is the quickest way to make a buyer doubt every photo.
The shot list changes with the garment, because the buyer's questions change with it.
| Garment | Photos every time | Add when the sheet lists a flaw |
|---|---|---|
| Tops, shirts and knitwear | Front, back, label, care label, collar and cuff detail, flat lay with the tape across pit to pit | A close-up of each flaw with a ruler for scale, plus a wider shot showing where it sits |
| Dresses and skirts | Front, back, waist and fastening detail, hem, label, care label, lining | The same pair of shots for each flaw |
| Trousers and denim | Front, back, waistband and label, inside leg seam, hem, fly detail | The same pair of shots for each flaw |
| Outerwear | Front open and closed, back, lining, label, cuffs, pockets | The same pair of shots for each flaw |
Let the assistant turn the sheet into the checklist for the day. Paste the category and flaw columns, give it the table above and ask for one numbered list per piece. For the blazer it returns eight shots: front, back, label, care label, lapel and shoulder detail, lining, then two close-ups with a ruler, one for the mark below the left pocket and one for the replaced button. Printed, or open on a second screen, that list ends the classic discovery on Wednesday night that nobody photographed the label.
Let the assistant look at the photos before a buyer does
ChatGPT and Claude both accept image uploads (check your plan's limits), which gives the seller a second pair of eyes before a buyer's. Upload one piece's photos with the finished description and ask for contradictions only:
Compare these photos with the description below. List (1) any colour that looks different from the colour named, (2) any mark, crease or damage visible in a photo but not covered by a listed flaw, (3) any listed flaw you cannot find in the photos. Do not guess about anything you cannot see.On the blazer, the answer came back with three points. The first was useful: photo 1 looked closer to dark teal than navy under that light. Checking the garment by the window showed it was navy, so the fault was the camera's white balance, and photo 1 was shot again. The second pointed at a crease across the front, which the seller added to the flaws as "storage crease, steamed, may reappear". The third could not find the mark below the left pocket in photo 5, which is a reason to reshoot that close-up nearer and sharper. Treat every flag as a prompt to look at the garment, never as a verdict: assistants can miss small flaws and occasionally report ones that are not there.
Editing needs its own boundary, and one question settles most cases: does the edit change what the garment is?
| Fine to do | Changes the garment, so don't |
|---|---|
| Crop and straighten | Smooth creases that are in the cloth |
| Match exposure and white balance to the real colour | Remove pilling, marks, holes or fading |
| Tidy the background, such as a cable or a wall mark | Push saturation until the colour is wrong |
| Add the ID or drop date as text | Add a generated model or background, or run an "enhance" filter |
Generated try-on images suit repeatable stock, and the tutorial on AI model shots for boutiques covers when they earn their place. They do not suit one-off vintage, because the picture has to show this garment's drape, colour and wear: the buyer is paying for that exact piece. Compare every edited photo with the garment in daylight before it goes into the queue.
A five-post countdown queued on Monday morning
The countdown is five posts, all written on Monday morning from the sheet and queued so the seller does not need to open Instagram again until Thursday, unless the teaser's event reminder has to be added by hand.
| When | Post | Job it does |
|---|---|---|
| Monday, 6pm | Teaser: one hero piece, size and flat measurements only, with a reminder set for Thursday 7pm | Tells followers when to be there |
| Tuesday, 12pm | "How I measure": the tape-across-pit-to-pit photo and what each number means | Answers the most common message before it is sent |
| Wednesday, 6pm | Preview of six pieces with sizes and measurements, no prices | Lets buyers decide before the clock starts |
| Thursday, 6pm | One hour to go: the shop link and the returns line | A last nudge for people who set no reminder |
| Thursday, 7pm | Live now: link, the first three pieces, how to ask about a flaw | The moment itself |
Make the teaser's image from a template, so the announcement looks like the shop and not like an advert: the hero piece on the usual backdrop, with the day and time added as text. Build it once, and each week only the photo and the time change. Instagram's own scheduler handles most of the rest. Its help page on scheduled posts and Reels says, as of October 2026, that the account must be public, that you can schedule up to 25 posts a day and up to 30 days ahead, and it names reels, photos and carousels as the formats. A week's countdown sits well inside those limits. The page also says product tagging, collaborative posts, sponsored posts and fundraisers are not compatible with scheduled content at this time, and it does not mention Stories, so check whether your own tool can queue them or post them by hand. Instagram caps hashtags at five per post, and the prompt below says so.
The Monday teaser can also carry an event reminder, which suits a drop with a fixed start time. Instagram's help page on event reminder posts says they are for professional accounts, can be created from one hour to three months before the start, and send followers who tap the bell a notification 24 hours before, 15 minutes before and at the start. The scheduling page does not say whether a post with a reminder can be queued, so try it once; if the option is missing, post the teaser by hand, which takes a minute.
Ask for all five posts in one go, naming the six pieces so the assistant has nothing else to draw on:
Write the five countdown posts for Thursday's 7pm drop using only the rows I paste after this message. Schedule: Monday 6pm teaser; Tuesday 12pm "how I measure"; Wednesday 6pm six-piece preview; Thursday 6pm one hour to go; Thursday 7pm live now. Rules: under 90 words each; no prices before Thursday 7pm; nothing that is not in the rows or in this message; at most five hashtags; exact sizes and flat measurements; end both Thursday posts with "Flaws are listed and photographed on every piece." Return five separate blocks I can paste.Two of the five came back like this, and both needed only a trim:
Monday 6pm: Thursday, 7pm: 24 new pieces, one at a time. First look: a navy wool-blend double-breasted blazer, tagged 14, 20.5 in pit to pit flat. Tap the bell on this post and we will remind you. #vintageclothing #vintageblazer
Thursday 7pm: Live now. This week's drop is on the shop page, link in bio. First three: the navy blazer (tagged 14), a cream cable-knit cardigan (tagged M, 22 in pit to pit flat) and straight-leg jeans (tagged 30, waist 15.5 in flat, 31 in around). Flaws are listed and photographed on every piece.The jeans line shows why tagged and measured sizes travel together: a tag of 30 and a flat waist of 15.5 in, which is 31 in around, is exactly the gap a buyer would otherwise discover with the parcel open.
Four drops in: reading a sell-through table
After a month the sheet doubles as a record of what sells. Add two columns, sold within 48 hours (yes or no) and days to sell, then paste the totals by category into the chat. The numbers below are invented to show the method.
| Category | Listed over 4 drops | Sold within 48 hours | Sell-through |
|---|---|---|---|
| Knitwear | 20 | 17 | 85% |
| Denim | 20 | 16 | 80% |
| Dresses | 24 | 15 | 63% |
| Outerwear | 16 | 9 | 56% |
| Shirts and blouses | 12 | 5 | 42% |
| Accessories | 4 | 2 | 50% |
| All pieces | 96 | 64 | 67% |
Asked which categories to list more or less of, the assistant gave a fluent and slightly over-sure answer: shirts and blouses were "the weak category" and worth dropping. Twelve listings across four drops is a hint, not a verdict, and four accessories is too few to read at all. The better question is why: were the shirts priced above the rest, or photographed worse than the knitwear? Asking "what would I need to know before dropping a category, and treat anything under 15 listings as too small to judge" gets a more careful answer than asking for conclusions. The mix for next Thursday is the seller's call.
The hours and the bill for one drop
With the sheet in place, the same drop looks like the table below. These are illustrations again; replace each figure with your own stopwatch numbers after the first week.
| Job | Before (hours) | After (hours) | What changed |
|---|---|---|---|
| Measure, grade and log each piece | 2.0 | 2.5 | Typed into the sheet as you go, which is slower but now the only place facts live |
| Photograph and edit | 3.5 | 3.0 | A checklist per garment means no forgotten label shots and fewer reshoots |
| Write 24 descriptions | 4.0 | 1.2 | Drafts arrive together; about three minutes each to read and correct |
| Countdown posts and scheduling | 1.5 | 0.6 | Five posts drafted at once and queued on Monday morning |
| Total | 11.0 | 7.3 | 3.7 hours back each drop |
The bill is small. A $20 a month plan spread across 96 pieces (four drops of 24) is about 21 cents a piece, and Instagram's scheduler needs no extra tool. The saved 3.7 hours a drop come to 14.8 hours a month, and what an hour of the seller's time is worth is a number only the seller knows. At $15 an hour it would be $222 against a $20 plan, but the honest version of the sum starts with your own stopwatch. The first drop will run longer than the table shows, because the ladder, the prompt and the sheet all have to be built; set aside a spare evening for that.
Thursday at 6pm: what only the seller decides
An hour before the queue fires, the seller runs one short check. It takes five minutes, and it is the part nobody else can do.
- Read the top six descriptions aloud with each garment in hand: measurements, flaws and colour.
- Open the live shop page on a phone and tap through from a queued post.
- Confirm every price in the sheet matches the shop.
- Check that every era word has an evidence cell and every provenance cell either says "none known" or holds something you could show.
- Read the countdown posts once more for anything the sheet does not say, such as a promise of free postage you do not offer.
What never goes to the assistant: the grade, the price, the era estimate, what goes into the drop and how to answer the buyer who says a flaw was not clear. Everything else is words, and words are cheap to redo. If you also email subscribers on drop day, the new-arrivals email routine reuses the same rows, and the quieter days of the week can be planned with the method in planning a month of marketing content with AI in one hour.
Questions sellers ask before they try this
How many pieces should a weekly drop have?
Let your own clock decide. Time one drop from washing to logging, divide the hours you can spare by the minutes each garment took, and round down. If that gives 18, list 18. A smaller drop you can describe accurately is better than a bigger one you rush, and you can add two pieces once a drop runs to time.
Do I have to say that AI helped write my listings?
Rules differ by marketplace and by where your buyers live, so read the current policy of each platform you sell on, and ask a solicitor if you are unsure about disclosure law. What protects buyers and you is accuracy: true measurements, every flaw named and photographs of the real garment. Platform rules on AI-generated or edited images deserve a separate read.
How do I keep a title inside a marketplace's character limit?
Put the limit in the prompt, then count in the sheet with =LEN(), because assistants miscount characters. eBay titles allow up to 80 characters (checked in September 2026; confirm before relying on it), and other platforms differ. Ask for three title options, paste your favourite into a cell and let the formula confirm it fits.
Can I follow this with a private Instagram account?
The sheet, descriptions and photo routine work on any account. Instagram's help page says scheduled posts need a public account, and its event-reminder feature is for professional accounts, so a private personal account would post the countdown by hand. Collab posts and product-tagged posts can't be scheduled whatever the account type.
Further reads
- AI for Thrift Stores: List Donated Items From a Phone Photo — A phone-photo listing routine for second-hand stock that arrives in bulk.
- Clothing Return Policy Template: Write One With AI and Check It — Write the returns wording that sits behind every drop.
- How to Create Outfit and Styling Posts for a Boutique With AI — Styling posts for the quieter days between drops.
- Charity Shop Pricing: Price Donated Items Faster With AI — Price donated items from two phone photos, a band card and a sold-price check, with a 90-second target per researched item and the brand mistakes AI makes.
- Yarn Shop Marketing: Fill Knitting Classes and Promote New Stock With AI — A class calendar, listing and caption prompts, a monthly newsletter template and a three-hour routine for a small yarn shop's marketing.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Instagram's help pages on scheduled posts and event reminder posts; Etsy's house rules for vintage items; ChatGPT and Claude pricing pages; EU Directive 2024/825 on environmental claims.