O OpenChatGPT Start chatting

Ask AI about an image

Some questions are easier to show than to type. Upload a photo, a screenshot, a chart, or a receipt, and ask the AI what it shows — what the error says, what the graph means, how much the bill adds up to, or what is written on the sign. This guide shows you how to ask AI about an image on OpenChatGPT, what to ask for each kind of image, and how to get accurate answers.

Advertisement

How to ask about an image

There is no account to create and nothing to install. Open the chat and tap the attach button (📎) next to the microphone at the bottom of the screen. Choose an image from your phone or computer — it appears as a small thumbnail above the message box. Then type your question and press Send. If you send the image with no question, the chat describes what it sees. Large images are automatically shrunk in your browser before they are sent, so the upload is fast even on a slow connection. Keep asking follow-ups in the same chat — "What about the small text at the bottom?", "Is there anything I missed?" — and the AI looks at the same image again with your new question in mind. One image per message: if you want to ask about a second image, send the first question, then attach the next one.

Example questions by image type

Different images need different questions. A vague "what is this?" works, but a question that says what you need gets a far better answer. Here are the most useful patterns, one for each kind of image people actually upload.

Screenshot of an error or a problem. "What does this error message mean, and how do I fix it?" The chat reads the error text and explains it in plain words, then suggests the fix. Add context in your message: "This happened when I opened the app on my phone" — the fix for a phone error is often different from the fix on a computer. If the screenshot shows code, ask "Explain this error in simple words, then show the fixed code." For more on this, see the guide on AI chat for coding.

Chart, graph, or table. "Summarise what this chart shows in three sentences." Or, more usefully: "What is the trend in this graph, and what is the highest value?" Charts are where the chat earns its keep — it reads the axes, the labels, and the shape of the data, and tells you the story in words. Always check the numbers it quotes against the chart itself: AI can misread a small label or a crowded axis, and a wrong number in a chart summary looks just as confident as a right one.

Receipt, bill, or invoice. "Add up the items on this receipt and tell me the total. List each item with its price." Or: "What is the due date and the amount on this bill?" The chat reads the printed text and does the arithmetic. Check the total against the printed total before you pay anything — if a line is blurry or folded, the chat may miss it or read a 3 as an 8. For a bill in another language, ask "Translate this receipt into English and tell me the total."

Homework photo. "Explain how to solve this maths problem step by step, but do not give me the final answer yet." Or take a photo of a diagram: "Explain this diagram in simple words — what is each part doing?" Photos of handwritten work are fine as long as the writing is clear and the light is good. Blurry handwriting is the most common reason for a wrong answer, so retake the photo before you blame the AI. For more study uses, see the guide on AI chat for students.

Sign, menu, or label in another language. "Translate what this sign says into English, and explain what it means." This works for street signs, menus, medicine labels, and instruction sheets. Take the photo straight on, with the text filling the frame, and the translation will be far better than a photo taken at an angle in bad light. If the translation matters — a medicine label, a legal notice — check it with a person who knows the language.

A photo of something you want to identify. "What plant is this, and how do I look after it?" or "What kind of bird is in this photo?" The chat will give you its best identification and the key features it used. Treat the answer as a starting point, not a certainty: similar-looking plants and animals are easy to confuse from one photo, and a wrong identification for something you plan to eat or touch could be dangerous. Confirm with a reliable source before you act on it.

Accuracy tips

Three things decide whether the answer is right: the photo, the question, and the check. Take the photo well: good light, the subject filling the frame, the camera straight and steady, no glare on a screen. Crop out the clutter before you upload — a receipt photographed on a messy table is harder to read than the receipt alone. Ask a specific question: "What is the total on this receipt?" beats "What is this?" every time, because the chat knows what to look for. And check the answer: read the numbers, dates, and names against the image yourself. AI image reading is genuinely good now, but it still misreads small text, confuses similar characters, and occasionally describes something that is not there. For anything important — money, medicine, an exam, a legal document — your own eyes are the final check.

Privacy: what happens to your image

Your image is not stored on this site's server. It leaves your browser only as part of the chat request, sent to the AI provider so the model can look at it and answer — the same way your typed message is sent. The image is not kept in any account or gallery, and the only copy is the one in your own browser's chat history, which you can delete at any time. Even so, think before you upload: do not send photos of ID cards, passports, bank cards, passwords, or other people's private documents. Crop out anything you would not want a stranger to see, and never upload a photo of someone else without their permission. The privacy policy and the AI disclosure explain this in full.

What it cannot do

Be clear about the edges. The chat cannot identify a person in a photo — do not ask who someone is, and do not upload photos of people to find out. It cannot see anything outside the image: no location, no date, no context beyond what is in the frame and what you tell it. It cannot read text that is too small, too blurry, or partly hidden — if you cannot read it yourself, the AI probably cannot either, so retake the photo. It cannot verify that a photo is real: a screenshot can be faked, and the chat has no way to know. And it can be confidently wrong about what it sees, especially with unusual objects, similar-looking items, and text in a language it handles weakly. Ask it to quote the text it read, or to describe exactly what it sees, and compare that with the image before you trust the conclusion.

Common mistakes

Ready to try it? Start chatting — tap the attach button, choose an image, and ask your question. It is free, needs no account, and there is no daily limit on messages. If your question is about a document rather than a photo, see the guide on chatting with a PDF.

More guides