What you’ll be able to do

Point your phone at pretty much anything, a plant on your windowsill, a weird bit of machinery, a page of text you can’t quite read, and ask what it is. Most of the mainstream AI tools now take photos as input, not just words. So you snap it, you ask, and you get an answer back, usually in a few seconds.

It won’t always be right. But it’s often close enough to be genuinely useful, and that’s what this guide is about, getting you set up to try it properly.

Before you start

A few things worth knowing before you start pointing your camera at stuff.

These tools are pattern matchers. They’ve seen enormous numbers of images and learned to recognise shapes, textures, common objects, landmarks, logos, text. When you show them a photo, they’re comparing what they see against everything they’ve learned and giving you their best guess, dressed up as a confident answer.

That confidence is the tricky bit. An AI will often sound just as sure when it’s wrong as when it’s right. It won’t say “I’m not sure” unless you ask it to. So the skill here isn’t really about taking the photo, it’s about knowing what to ask and how to check the answer.

There are also things you should never use these tools for. Mushroom identification is the classic example, get it wrong and someone could get seriously ill or worse. Same goes for medication, don’t rely on a photo ID to tell you what a pill is or whether it’s safe to take. And don’t upload photos of other people to try and identify who they are. That’s a privacy line, and most mainstream tools will actually refuse to do it anyway.

Keep this in mind: good for curiosity, plants, landmarks, products, general “what is this thing” questions. Not good, ever, for anything safety critical.

Get set up

You don’t need anything fancy. Just your phone, or a computer with a camera or the ability to upload a photo, and one of these tools.

Google Lens is built into the Google app, Google Photos, and Chrome on most phones. It’s specifically designed for this: point it at an object, a landmark, a plant, a product, even a page of text, and it’ll try to identify it. Genuinely one of the easiest ways in, because there’s a dedicated camera icon waiting for you.

ChatGPT, Claude, and Gemini all accept photos now. You can upload an existing photo or, on mobile, take one directly in the app, then ask a question about it in plain English. This is where you get more of a conversation rather than just a label slapped on the thing.

Gemini Live can use your phone’s camera continuously, so instead of snap-then-ask, you’re pointing your camera and talking to it in real time, a bit like narrating what you’re looking at and getting responses as you go.

Apple’s Visual Intelligence, on recent iPhones, works a similar way, letting you point your camera at something and get information without leaving the camera view.

Bing Visual Search is Microsoft’s version, built into Bing and Copilot, and works well for products and shopping-type identification in particular.

Pick whichever you’ve already got installed. Honestly, for a first go, Google Lens is the lowest-friction option because there’s no typing involved at all.

Try it yourself

Right, let’s actually do this.

Start simple. Take a photo of something you’re mildly curious about, an object on your desk, a building you walk past, a bit of packaging. Crop the photo if you can so the thing you care about fills most of the frame; less clutter means a better guess.

Then ask. If you’re using Google Lens, tap the camera icon, point, and it’ll surface suggestions automatically. If you’re in ChatGPT, Claude, or Gemini, upload the photo and type a question underneath it.

Here are three prompts to copy and use, each doing a slightly different job.

For a straightforward identification:

What is this? Please tell me what it is and how confident you are in that answer.

For something with a specific detail you need, like a product or a model number:

What model or version is this? Look closely at any text, logos, or design details in the photo.

For anything involving safety, food, or “should I do something with this”:

Based on this photo, what do you think this is, and how confident are you? 
Also tell me what else it could realistically be, and remind me if this is 
something I should double-check with another source before acting on it.

Notice that last one builds in the check. That’s deliberate. Get into the habit of asking for confidence and alternatives every time, not just when you remember to.

Check the result

Once you’ve got an answer, don’t just take it. Test it a bit.

Ask the AI directly how sure it is. Something like “how confident are you out of ten” or “what’s your second guess if this isn’t right” works well and most tools will give you a genuinely useful answer rather than dodging the question.

Look at the language it used. “This is definitely a…” is different from “this looks like it could be a…” AI answers do vary in how hedged they are, and that hedging is often a real signal, not just politeness.

Cross-check with a second tool if it matters to you. Run the same photo through Google Lens and then ask ChatGPT the same question. If they agree, that’s reassuring. If they don’t, that’s useful information too, it tells you the identification is genuinely uncertain rather than you having asked badly.

And do a basic sense check yourself. Does the answer match what you already know? If it’s told you a common garden bird is a species that doesn’t live anywhere near you, something’s gone wrong.

For anything that matters, a plant you’re thinking of eating, a part you need to order, a document you’re relying on, treat the AI’s answer as a strong hint, not a final word. Verify with a proper source before you act on it.

If it doesn’t work

Sometimes you’ll get a vague answer, a wrong one, or no answer at all. A few common reasons and what to do about them.

The photo’s too cluttered or too far away. Crop it tighter, get closer, try again. AI tools do much better with one clear subject than a busy scene with ten possible things in it.

The lighting’s poor or the angle’s odd. Try a different photo if you can, from a clearer angle or in better light.

The thing genuinely is ambiguous, lots of similar-looking objects, plants, or products exist and even a person would struggle to tell them apart from one photo. In that case, ask for the AI’s best few guesses rather than expecting one definitive answer.

Now the important bit: some things you shouldn’t be asking about at all in this way. Never rely on an AI photo ID for mushrooms, wild plants you’re thinking of eating, or anything about medication, dosage, or whether something is safe to ingest. Get those checked by an actual expert or a proper reference source, every time, no exceptions.

And don’t upload photos of other people to try to work out who they are. It’s a privacy issue, it’s not something these tools are meant for, and most mainstream ones will refuse the request anyway if you try.

If the answer still looks wrong after a retry, just trust your own judgement. You know your kitchen, your garden, your neighbourhood better than any AI does from a single photo.

Keep exploring

If this was useful, there are a few related things worth trying next. “How to get AI to identify a plant” goes deeper into garden and houseplant identification specifically, including the bits that go wrong most often. “How to get AI to read my handwriting” covers getting text out of scrawled notes and old letters. And “How to get AI to edit a photo” moves from identifying what’s in an image to actually changing it.

Sources and review notes

Review date: 15 September 2026 — every source above was opened and checked against the text on that date.