You are looking at something you want explained: a chart in a PDF, an error dialog, a line of code in a video, a practice question. The slow way is well known. The faster ways depend on your operating system and on how much you want to send.
1. Screenshot, then upload
Take a screenshot, open a chat tab, attach the image, type the question.
- Works everywhere, with any chatbot that accepts images.
- Slow when repeated. Four or five steps per question, and your screenshots folder fills up.
- Sends the whole image, including anything around the part you care about.
Fine for one question a day, tiring for twenty.
2. The ChatGPT desktop app
In the ChatGPT app for macOS, click the plus icon in the prompt window or the launcher and hover over Take Screenshot. It lists your open windows; pick one and it is attached to your message. macOS asks for the screen recording permission the first time, and the app may need a restart afterwards.
- Good if ChatGPT is already your main assistant and you mostly ask one model.
- You attach a whole window, then describe which part you mean.
- One provider. There is no second opinion unless you repeat the question elsewhere.
3. Click to Do on Windows 11
On a Copilot+ PC, Windows key + Q (or Windows key + click) turns the screen into something you can act on: copy text out of an image, search for it, or ask Copilot about it. The screen analysis itself runs on the device, and content leaves the computer only when you choose an online action.
- Built in, no extra app.
- Copilot+ PCs only. It needs a neural processor (40 TOPS), 16 GB of memory and a recent Windows 11 build, so most existing PCs cannot use it.
- Features vary by region and language.
4. Google Lens in Chrome
In Chrome on a computer, open the three-dot menu (or right-click the page) and choose Search with Google Lens, then click, highlight text or drag over part of the page. Results open in a side panel, where you can type a follow-up question about what you selected.
- Quick for anything in a browser tab.
- Only inside Chrome. It cannot see a PDF reader, an IDE or a desktop app.
- A screenshot of the page goes to Google. Chrome asks you to agree to this the first time.
- Built on search. It is strongest at identifying things and finding related pages.
5. A select-and-ask app
A small group of desktop tools is built for exactly this: press a shortcut or pick a menu item, drag over a region anywhere on the screen, and the answer appears next to it.
LuciQA works this way on macOS, Windows and Linux. What sets it apart from the options above:
- You choose what is sent. Text is read on your device with OCR and only the text goes out; for a chart or diagram you send that region as an image.
- Your own API key, your choice of provider. Gemini, Groq, OpenRouter, Cloudflare and Mistral have free tiers; OpenAI and Claude work too. Requests go straight from your computer to the provider.
- A second opinion when it matters. The Pro version asks several models at once and shows where they disagree.
The trade-off is setup: you need to create an API key once. Getting a free Gemini key takes about two minutes.
Which one should you use?
- One question now and then: screenshot and upload is enough.
- You live in ChatGPT on a Mac: its desktop app.
- You have a Copilot+ PC: try Click to Do first, it is already there.
- Everything you ask about is in a browser tab: Google Lens.
- Many questions a day, across apps, or on Linux: a select-and-ask app.
Whatever you use, remember that the answer can be wrong. For anything that matters, check it against a source you trust.