Send Images to ChatGPT

ChatGPT isn’t limited to text. Drop a photo, screenshot, diagram, or scanned document into the conversation, then ask your question about it. The shift runs deeper than it looks: instead of describing what you see with words, you show it. According to OpenAI’s Help Center, this feature is available on all plans, including the free tier, and every ChatGPT model accepts images.

Supported formats and input methods

ChatGPT accepts PNG, JPEG, and non-animated GIF images up to 20 MB each. To add one, select + in the input box, then choose Add photos & files; drag the file into the text area; or paste an image copied to your clipboard. On mobile, you can also take a photo or pick one from your gallery.

There’s no fixed limit on the number of images per conversation: the cap depends on image size and the amount of accompanying text. If you run into trouble, the Help Center recommends reducing the number or size of images. ChatGPT also accepts video attachments, but analysis may be incomplete or inaccurate, especially for audio.

An image alone says nothing

The classic beginner mistake is always the same: sending a photo without instructions. Without intent, the model falls back on its default behavior—describing—and you get a correct but low-value paragraph.

Take a photo of a whiteboard after a meeting. With “What is this?”, you get a general description. With “Here’s the whiteboard from our kickoff meeting. Transcribe the decisions and action items, along with their owners, into a table,” you get a deliverable. The image is identical in both cases; only the intent changed. To draw attention to a specific area, the Help Center suggests annotating the image before sending it, for example by circling the relevant passage.

Multiple images in a single message

You can send several images together and request a cross-analysis. Two photos of the same product taken at different times, paired with “Compare these two photos and list the visible changes,” yield a structured comparative response, not two side-by-side descriptions you’d have to reconcile yourself. Keep the count reasonable: as the batch grows, fine details are more likely to get lost.

Make the photo reflex your habit

The highest-return uses are often the simplest: a screenshot of an error message to explain, a paper form to transcribe, a chart to comment on, a room to lay out, a label to translate. In every case, the same rule applies: a sharp image, a clear objective, and a precise output format.

What derails an analysis

A blurry or too-dark image remains the top obstacle: no prompt can rescue a bad shot, so retake the photo. OpenAI’s Help Center also lists known limitations to keep in mind:

• Medical imaging: The model isn’t suited to interpreting specialized medical images like CT scans and must not be used for medical advice.
• Text: Non-Latin scripts are read less accurately; small text benefits from enlargement; and rotated or upside-down text may be misinterpreted.
• Charts: Curves distinguished only by line style—solid, dashed, or dotted—can cause confusion.
• Space and counting: Precise localization, such as piece positions on a chessboard, is difficult, and object counts remain approximate.
• Shapes and files: Panoramic or fisheye images pose problems; the model doesn’t read file names or metadata, and the image is resized before analysis.

Overconfidence adds another risk: on ambiguous elements, ChatGPT may answer with unwarranted certainty and get it wrong. Therefore, verify any critical information—amounts, dosages, references—against the source. Finally, consider privacy before sending an invoice, ID document, or screenshot containing personal data: images follow the same data usage rules as the rest of your content.

Your turn

Show, don’t describe. Photograph a work document: a whiteboard, a form, or a printed chart. Send it once with “What is this?”, then again with a clear objective and a specific output format. Then verify three elements from the second response by comparing them to the original.

Official sources: ChatGPT Image Inputs FAQ · How your data is used to improve model performance (OpenAI Help Center) · Image inputs (ChatGPT documentation)

Key takeaways

• All ChatGPT plans and models accept images: PNG, JPEG, and non-animated GIF, up to 20 MB
• Always pair the image with an objective and an output format; annotate the area that matters
• Multiple images can be analyzed together, but oversized batches lose fine details
• Know the documented limits: medical imaging, non-Latin scripts, counting, and precise localization
• Verify all critical information and weigh privacy before sending an image