Four Ways to Tell ChatGPT What Is In Your Closet, Ranked
Before ChatGPT can say anything true about your clothes, it has to know what your clothes are. There are four ways to tell it, they take between five minutes and three hours, and they are not equally good.
They are also not ranked the way you would expect. The fastest method is genuinely the right answer for a lot of people, and the most thorough method is overkill for most. What follows is each one honestly assessed, including the parts that are annoying.
The short version
Ranked by how well they hold up over time: a connected wardrobe, a spreadsheet, a typed list, then photo by photo. Ranked by how quickly you get a useful answer: a typed list, a spreadsheet, photo by photo, then a connected wardrobe. If you use ChatGPT for clothes a few times a year, type a list and stop reading. If you use it weekly, the gap between the methods becomes the whole story.
Method 1: the typed list
Write your clothes into the chat, one per line, and ask your question underneath.
Time to set up. Fifteen to forty minutes for a first pass. Five minutes if you only list what is relevant to today.
What it is genuinely good at. Everything about it is honest. You can see exactly what the model can see, because it is right there on screen. There is no retrieval deciding what surfaces, no file being partly read, no integration to configure. For one good answer today, nothing beats it, and it is free on every plan.
It is also the method that teaches you the most. Writing out your wardrobe by hand is a small audit in itself. Most people notice at least one thing they had forgotten they owned.
Where it breaks.
It does not survive. New conversation, gone. You will paste it again.
It decays inside a conversation. As the chat grows, the list is the oldest content and the first to fall out of view. This produces the characteristic pattern of good early answers narrowing into repetitive ones and eventually into invented items. The mechanism is set out in Why ChatGPT Forgets Your Wardrobe.
It punishes size. A hundred and fifty item list works worse than a forty item list, because at some point the model is no longer seeing all of it and has no way of telling you.
It has no history. Nothing records what you wore, so you get your best combination handed to you repeatedly.
How to do it properly. One line per item, always in the same shape:
Category | Colour | Fabric | Fit or cut | Formality 1-5 | Note
Trousers | Charcoal | Wool blend | Wide leg, full length | 4 | Needs a heel
Knit | Oatmeal | Lambswool | Boxy, crew | 2 | Pills at the cuff
Boots | Black | Leather | Ankle, block heel 5cm | 3 |Consistent structure matters more than detail. A model reading a uniform pattern extracts far more reliably than one reading prose, and the formality number is doing more work than anything else in the row.
Rank: third, and first for occasional use. The prompt that goes with it is in How to Get ChatGPT to Pick Your Outfit.
Method 2: the spreadsheet
Build your wardrobe once in a spreadsheet, then either upload it or paste its contents when you need it.
Time to set up. Ninety minutes to three hours for a full wardrobe. Realistically, two evenings.
What it is genuinely good at. It is the only method where the work compounds. Add a column and every future answer gets better. Sort by formality and you can see the shape of your wardrobe. Filter to what you have not worn and you have half an audit already done.
Structured rows also parse far better than prose. A consistent grid gives the model something closer to data than to text, and the answers are noticeably more precise.
And the file is yours. It outlives any app, any subscription and any change of AI tool.
Where it breaks.
Uploading is not the same as being read. Attached files are usually searched rather than read from end to end, so the model works from the chunks that matched your question. Items whose wording lacks obvious keywords never surface, and nothing warns you. This is the single most misunderstood thing about the spreadsheet method: it feels permanent and reliable, and it is neither, when uploaded.
Pasting the contents into the chat as text is more reliable than attaching, and costs you the context problem instead.
It goes stale immediately. Every purchase, donation and alteration puts the file further out of date, and the file will never tell you.
Maintenance is real. Everyone updates a wardrobe spreadsheet enthusiastically for three weeks.
The compromise that works best. Keep the spreadsheet as your master, and paste a filtered subset into the chat for each question. Going to a wedding, paste rows with formality 4 and 5. This gives you a short, fully visible list drawn from a complete record, which sidesteps both failure modes at once.
How to do it properly. Column order matters, because retrieval reads left to right and the first columns carry the most weight. The layout that parses best, along with the prompt to paste with it, is in The ChatGPT Wardrobe Inventory Template.
Rank: second.
Method 3: photo by photo
Photograph each garment and show ChatGPT the images, letting it identify and describe each one.
Time to set up. Two to four hours for a full wardrobe, including the photography, which is most of it.
What it is genuinely good at. Accuracy of description, when it works. The model will pick up texture, drape, cut and colour with a precision most people cannot produce in words. Ask it to describe a jacket and it will tell you the lapel is notched and the shoulder is soft, which is exactly the sort of detail that makes an outfit suggestion correct rather than plausible.
It is also the best method for a wardrobe you cannot describe. If you do not have the vocabulary for what you own, the camera has it for you.
Where it breaks, and it breaks hard.
Images are expensive in context. A photograph consumes far more of the available window than a line of text describing the same garment. A wardrobe in images fills the context window many times faster than a wardrobe in words, so the point at which things start falling out of view arrives much sooner.
You can only show a limited number at once, so a full wardrobe becomes many batches, and batches do not accumulate. By batch four, batch one is fading.
None of it persists. Photographs do not survive into the next conversation any more than text does, and you will not photograph your wardrobe twice.
Full closet shots do not work. Garments overlap, colours read wrongly under wardrobe lighting, and the model will confidently miscount. Individual flat lays work. Rail shots mostly do not.
The version that is actually worth doing. Do not photograph your wardrobe. Photograph the fifteen items you genuinely cannot describe, ask ChatGPT to write a one line entry for each in your spreadsheet format, and paste those lines into your list. You get the accuracy of the image and the durability of text. This is the best use of photos and almost nobody does it this way.
Rank: fourth as a wardrobe method, first as a description tool.
Method 4: a connected wardrobe
Your clothes live in a structured wardrobe outside the chat, and ChatGPT connects to it and queries it directly.
Time to set up. Cataloguing, which is the same cataloguing every other method asks for, plus a few minutes to connect. There is no version of this where the clothes get into a system without someone putting them there.
What it is genuinely good at.
Nothing is ever pasted, so nothing falls out of context. The list is not in the conversation at all. The model asks for what it needs at the moment it needs it.
Size stops mattering. Two hundred items behaves the same as twenty, because neither one is sitting in the window.
Nothing is a snapshot. Add a jacket and the next answer includes it, with no re-upload.
Completeness is available. When the model asks for every item in a category, it gets every item in that category, rather than the ones a search decided were relevant.
It can count. This is the part that no other method offers at all, because a chat window has nowhere to write anything down. Wear history, cost per wear, what you rejected, what you have not touched since autumn: all of it requires something keeping score, and only this method has anything keeping score.
Where it breaks, honestly.
The cataloguing is still the cataloguing. If you will not spend the time getting your clothes into a system, no system helps you. This is the real barrier and it is the same barrier for a spreadsheet.
You are dependent on a service. A spreadsheet on your own machine cannot go away. Judge any wardrobe tool by whether you can export everything, and treat that as non negotiable.
It is more than you need for occasional use. If you ask about clothes three times a year, connecting a wardrobe is a lot of setup to save some pasting.
Rank: first, for anyone who does this weekly. How it compares with custom GPTs, Projects and standalone wardrobe apps is covered in ChatGPT as an Outfit Planner.
This is what Modebase is built as. Your wardrobe is stored properly, ChatGPT connects to it, and you ask in the chat you already use. The styling is still ChatGPT. What changes is that it is styling your actual clothes and it remembers what you wore.
The comparison
| Typed list | Spreadsheet | Photo by photo | Connected wardrobe | |
|---|---|---|---|---|
| Setup time | 15 to 40 min | 90 min to 3 hr | 2 to 4 hr | Cataloguing plus minutes |
| Survives a new chat | No | Partly, via file | No | Yes |
| Holds 200 items | No | Partly | No | Yes |
| Read completely | Yes, while visible | Not when attached | Only in batches | Yes |
| Stays current | Re-typed each time | Goes stale | Goes stale | Live |
| Records what you wore | No | Manually | No | Yes |
| Cost | Free | Free | Free | Subscription |
| Portable if you stop | Nothing to keep | The file is yours | Nothing to keep | Only if it exports |
How to choose
A few times a year. Type a list of what is relevant to today and ask your question. Do not build anything.
Monthly, and you like a spreadsheet. Build the spreadsheet, keep it as your master, and paste filtered subsets rather than attaching the whole thing.
You do not have the words for your clothes. Photograph the difficult items, have ChatGPT write the rows, and put those rows into your list or spreadsheet.
Weekly, and you are tired of pasting. Connect the wardrobe. The pasting is not the real cost anyway. The real cost is that every answer is about a slightly different, slightly imaginary version of your closet, and that only goes away when the wardrobe stops living in the conversation.
Frequently asked questions
What is the best way to give ChatGPT my wardrobe?
For a single question, a typed list of the twenty five to forty items you would actually consider, in a consistent one line format. For repeated use, a structured wardrobe the model can query, because that is the only approach where nothing falls out of context and nothing goes stale.
Should I upload a spreadsheet or paste it into the chat?
Paste it if it is short enough, because pasted text is visible to the model in full while an attached file is usually searched rather than read completely. Attach it if it is too long to paste, and accept that some rows will never surface. The best of both is to keep the full spreadsheet and paste a filtered subset relevant to the question.
Can ChatGPT identify my clothes from photos?
Yes, and it is good at it, particularly for texture, cut and drape. The limits are that images consume much more context than text, only a limited number can be shown at once, and none of them persist into a new conversation. Photograph the items you cannot describe rather than the whole wardrobe.
How many items can ChatGPT actually handle at once?
As a practical rule, forty to sixty well formatted lines stay reliably visible through a normal conversation. Beyond about a hundred, expect items to start being dropped without any warning. Beyond two hundred, assume the model is working from a fraction unless the wardrobe is being queried rather than pasted.
Do I have to photograph everything to use a wardrobe app?
Photographs are what most wardrobe apps use to build your closet, so some capture step is unavoidable. The realistic approach is to catalogue the things you actually wear first, which is usually a third of what you own, and add the rest when it comes into rotation.