Understand your bills with an iPhone, beginners can follow along

The VoiceOver upgrade in iOS 27 is worth trying, but don't treat the recognition results as final answers. It can describe photos, screenshots, and bills, rather than just reading out text on the screen line by line. I actually tested it; over the past few days, recognizing a food delivery screenshot on the beta version took about ten-plus seconds, and the entry point is smoother than third-party tools. My judgment is that it's recommended to get started with, but for critical scenarios like finance, medical, or contracts, you need to assess the situation.

Let me clarify a few terms first. VoiceOver is Apple's built-in screen reader (in English, VoiceOver), which reads interface content aloud to you. Image Description is when the phone sees an image and uses text to explain what's in it. Action Button is the physical button on the left side of some iPhones, which can be set to quickly launch specific functions. Real-time Recognition means pointing at something with the camera and asking what it sees.

You can follow these specific steps. Open Settings, go to Accessibility, then tap VoiceOver and turn the switch on. The phone will enter screen-reading mode: swipe with one finger to select items, double-tap to open them. Expect to hear the phone start reading the current page. Open Photos, select a bill or screenshot. Swipe with one finger to the image, double-tap to open it. Then continue swiping down with one finger to find Image Description or Image Explorer, and double-tap. Expect to hear something like "A paper bill, date at the top, amount at the bottom right." Once inside Image Explorer, follow the on-screen prompts to input questions, such as "What is the amount?" or "Who said what in this screenshot?" If voice support is available, hold the Action Button or just speak the question. Expect a summary sentence back, not just the title being read out. If your model has an Action Button, go to Settings > Accessibility > Action Button, and select VoiceOver Image Description or Ask About Camera Content. Expect long-pressing the left button to quickly launch it. Open Camera, lay the bill flat, long-press the Action Button, and ask "What does the first line say?" Expect the phone to read out visible text or describe the image content.

In the update notes, Apple states that VoiceOver can now understand photos, charts, and screenshots in more detail and contextually, and can also retrieve answers faster via the Action Button.

My feeling is that the system-level entry point is smoother than third-party apps, no need to switch back and forth. Previously, when I first tried image input with GLM-4V, I succeeded once, but the process was still "open tool, upload image, wait for result." Apple's setup puts it directly in Photos and Camera, saving two steps. However, it's not omnipotent. I tested it with a crumpled utility bill; the first time it got the position of the amount wrong. Only after flattening it and retaking the photo did it stabilize.

Common pitfalls for beginners are mainly a few things. No response from the Action Button usually means it wasn't bound in Accessibility, or the entry point changed in the beta version. Inaccurate recognition is affected by blurry images, small text, and glare. Accidental touches in VoiceOver mode: single-finger swipe selects, double-tap opens. It takes getting used to initially, so I suggest practicing with an unimportant screenshot first.

The advantages are clear: no app switching needed, descriptions are more context-aware, saves eye strain for low-vision users, and makes handling screenshots easier for regular people. The disadvantages are direct too: menu names might not be consistent across versions, recognition is affected by image quality, and the beta version may drain battery or lag. For things like bills, contracts, and diagnostic reports, it's best to have a human double-check.

After learning this, the next step is to try Scan Documents in Notes. Scan a courier slip or contract page, then use VoiceOver to ask "Who is the recipient?" or "Is there a date?" Further ahead, keep an eye on natural language operations for Voice Control.

Useful, but key information still needs human verification.