How to Upload Images to ChatGPT (Step-by-Step Guide for 2026)

ChatGPT stopped being just a text tool years ago. You can now drop in a photo of a handwritten note, a screenshot of a spreadsheet, or a picture of the inside of your fridge and have it read, analyze, or make sense of what it sees.

If you’ve never uploaded an image to ChatGPT before — or if it used to work and now it doesn’t — this guide walks you through every method that actually works in 2026, across desktop, mobile, and the API. It also covers the limits, the errors you’ll hit, and what to do when the upload button disappears.

Can you upload images to ChatGPT for free?

Yes. Image uploads are available to both free and paid ChatGPT users, on web and mobile. The catch: free users share a usage pool across GPT-4o and GPT-5 models, and once you hit the cap, you get downgraded to a lighter model that may or may not accept images depending on the day.

If you plan to upload images regularly — more than a few times per week — ChatGPT Plus or ChatGPT Go pays for itself by removing the cap and giving you faster processing for image-heavy prompts.

How to upload images to ChatGPT on desktop (web browser)

This is the method most people use. It works in Chrome, Safari, Firefox, Edge, and Brave.

Step 1. Go to chatgpt.com and log in with your account.

Step 2. Start a new chat or open an existing one.

Step 3. At the bottom of the chat window, look for the + icon (older interfaces show a paperclip) on the left side of the message box.

Step 4. Click it. You’ll see a menu with options like:

Upload from computer
Connect to Google Drive
Connect to Microsoft OneDrive

Step 5. Pick “Upload from computer,” select your image, and click Open. The image will appear as a thumbnail above the text box.

Step 6. Type your prompt in the message box — for example, “What’s in this picture?” or “Extract all the text from this receipt” — and hit send.

That’s it. ChatGPT will process the image and respond with its analysis.

Two faster shortcuts most people miss

Instead of clicking through menus, you can:

Drag and drop the image file directly from your desktop into the chat window.
Paste an image you’ve copied to your clipboard with Ctrl + V (Windows) or Cmd + V (Mac). This is the fastest way to send a screenshot — just press Print Screen or Cmd + Shift + 4, then paste straight into ChatGPT.

How to upload images to ChatGPT on mobile (iOS and Android)

The mobile app works slightly differently because you can also take photos on the fly.

Step 1. Open the ChatGPT app on your iPhone or Android.

Step 2. Tap the + icon to the left of the message bar.

Step 3. You’ll see three options:

Camera — take a new photo right now
Photos — pick from your camera roll
Files — upload from your device storage or iCloud/Google Drive

Step 4. Select or capture your image. You can upload more than one at a time by tapping multiple photos.

Step 5. Add your text prompt (optional, but highly recommended) and tap send.

One thing worth knowing: on the mobile app, you can also annotate the image before sending. After you attach it, tap the image thumbnail, then use the markup tools to circle, highlight, or draw arrows. This is a huge help when you want ChatGPT to focus on a specific area — a single cell in a spreadsheet, one ingredient on a food label, a line of code in a screenshot.

Supported image formats and upload limits

ChatGPT accepts these formats:

JPG / JPEG
PNG
GIF (static only — it won’t analyze animation)
WEBP
HEIC (from iPhones)

Per-image size limit: up to 20 MB.

Per-message limit: you can attach up to 10 images in a single message, though just because you can doesn’t mean you should — stuffing 10 images into one prompt tends to dilute the response quality.

Per-day limit: this one is fuzzy and changes based on your plan:

Free users: roughly 3–5 image analyses per 3-hour window on GPT-4o, then the feature becomes limited.
Plus users: effectively unlimited for normal use, though very heavy users may hit soft caps.
Pro and Enterprise users: no practical limit.

OpenAI doesn’t publish exact numbers because they adjust them based on server load.

What can ChatGPT actually do with an uploaded image?

This is where it gets fun. A few things that work reliably:

Describe the image — “What do you see here?”
Extract text (OCR) from receipts, screenshots, handwritten notes, or photos of documents.
Translate text inside an image — a restaurant menu in another language, for instance.
Identify objects, plants, animals, or landmarks — though for medical, legal, or plant ID cases, always verify with an expert source.
Explain charts, graphs, and diagrams — paste a screenshot of a dashboard and ask “what’s the trend here?”
Debug code from a screenshot — especially useful when error messages are inside a terminal or IDE you can’t easily copy from.
Critique design work — upload a mockup and ask for feedback on layout, typography, or color.
Solve math problems from a photo of a worksheet.

It’s worth giving ChatGPT context every time. “Here’s a screenshot of my landing page — tell me what’s wrong with the conversion flow” gets a far better response than just uploading the image with no prompt.

Common errors when uploading images (and how to fix them)

“Error in message stream” after uploading This usually means the image was too large, or the server is overloaded. Try compressing the image to under 5 MB, refreshing the page, and re-uploading. If that fails, switch to a different model from the dropdown.

“Upload failed” or the image never attaches Check your internet connection first, then clear your browser cache. If you’re on a corporate network, a firewall may be blocking the upload — try a different network or a mobile hotspot as a test.

The + icon is missing Most often this means you’ve been switched to a model that doesn’t support image inputs. Click the model name at the top of the chat and choose GPT-4o, GPT-5, or another vision-capable model.

ChatGPT says it can’t see the image Refresh the page and re-upload. This is usually a temporary glitch, not a permanent limitation.

Upload images to ChatGPT via the API

If you’re a developer, image inputs are supported through the Chat Completions API using vision-capable models.

You can pass an image either as a URL or as a base64-encoded string inside the image_url field of a user message. The basic flow:

Encode your image to base64 or host it at an accessible URL.
Send a request to the /v1/chat/completions endpoint with a model like gpt-4o or gpt-5.
Include the image alongside your text prompt inside the content array.

The OpenAI vision documentation has current code samples in Python, Node.js, and cURL — always check it before building, because parameter names occasionally change.

Is it safe to upload images to ChatGPT?

A few things worth knowing before you upload sensitive images:

Free and Plus accounts: by default, OpenAI may use your conversations (including uploaded images) to improve their models. You can turn this off in Settings → Data Controls → “Improve the model for everyone.”
Enterprise and Team accounts: content is not used for training by default.
Faces and private documents: ChatGPT will often refuse to identify specific people in photos for privacy reasons, which is a feature, not a bug.
Before uploading anything sensitive — medical records, financial documents, ID cards, passwords visible in screenshots — consider whether you’d be comfortable with that data sitting on OpenAI’s servers. For truly confidential material, redact first.

Quick FAQ

Can ChatGPT upload images in a PDF? Yes — but you upload the PDF itself, not the images inside it. ChatGPT will read the PDF and analyze both the text and any embedded images.

Can I upload multiple images at once? Yes, up to 10 images per message. They’re processed together, so you can ask comparative questions like “which of these two designs is cleaner?”

Can ChatGPT generate images, not just read them? Yes, but that’s a separate feature. Ask it to “create an image of…” and it’ll use its built-in image generator. Uploading and generating are two different flows.

Does ChatGPT remember images from earlier in the conversation? Within the same chat, yes. Across separate chats, no — unless you have memory turned on and ChatGPT has saved relevant details.

Why does ChatGPT refuse to analyze my image? Most common reasons: the image contains a real person’s face, graphic content, copyrighted material, or something the safety filter flagged. Try cropping or using a different image.

Uploading images turns ChatGPT from a text tool into something closer to a general-purpose assistant — one that can read your mail, debug your code, and explain what’s on your screen. The process takes about five seconds once you know where the button is, and the real skill is writing a good prompt to go with the image.

If you’re looking to go deeper, check out our guide to the top 10 best free AI tools or our roundup of free AI tools for content creation — many of them pair well with ChatGPT’s image capabilities.