What you’ll be able to do

Snap a photo of that thing growing in your garden, or the one your neighbour gave you a cutting of three years ago and you’ve never known what it is, and get an actual name back. Not just “it’s a plant” but the Latin name, the common name, and how to keep it alive. You don’t need to know a thing about botany first. You just need a phone and something green to point it at.

Before you start

Here’s the thing, you barely need anything. A phone with a camera does the job. That’s it, really.

What AI can’t do is guarantee it’s right, especially with tricky families where lots of species look almost identical, or with young plants that haven’t flowered yet. It’s working from what’s visible in your photo, same as you would if someone handed you a leaf and asked “what’s this then?” Sometimes it’ll nail it in a second. Sometimes it’ll give you its best guess dressed up as confident fact, which is worse than not knowing.

That’s why a second opinion matters, and I mean it, not just as a nice-to-have. If you’re thinking of eating something, treating a skin condition with a leaf poultice, or letting a toddler near berries you can’t identify, one AI tool saying “this looks like elderberry” is not enough to act on. Poisonous look-alikes are genuinely common. Get a second tool to agree before you do anything with it.

Get set up

For plant ID specifically, Google Lens is the easiest starting point if you’ve got an Android phone or the Google app, since it’s built right in and works from a photo you take or one already in your gallery.

PlantNet is worth having too. It’s free, it’s run by a non-profit, and it was built specifically for plant identification, which shows in how it handles the job. Similarly, iNaturalist’s Seek app is aimed at exactly this and leans into the nature side of things, so if you’re the sort who wants to log what you find, that’s a nice bonus.

PictureThis is the paid option that comes up a lot. It’s popular, does a solid job, and if you want ongoing plant care reminders and a slicker interface it’s worth a look, but you’re paying a subscription for something the free tools mostly cover.

And then there’s ChatGPT, Gemini, or Claude, which weren’t built as plant ID tools but can absolutely do the job if you upload a photo and ask. The advantage here is you get a conversation, not just a label. You can ask follow-up questions, get care instructions, ask about look-alikes, all in the same thread.

Honestly, I’d download PlantNet, keep Google Lens in your back pocket if you’ve got it, and use ChatGPT or Gemini for the “what do I do with it now” conversation afterwards. Belt and braces.

Try it yourself

The photo matters more than the app, honestly. AI can only work with what it sees, so give it something decent to look at.

Take three photos if you can: the whole plant so it can see the shape and size, a close-up of a single leaf showing its edges and vein pattern, and the flower if there is one, since flowers are often the single most useful clue for narrowing things down. Good natural light, no blur, plant filling most of the frame. If it’s raining sideways and you can only manage one shot, get the leaf.

Once you’ve got your photo, here’s how to ask ChatGPT, Gemini or Claude properly. Don’t just say “what’s this”, give it something to work with.

Here's a photo of a plant from my garden. Please identify it, giving me both the Latin (scientific) name and the common name. Also tell me two plants it could be commonly confused with, and how to tell them apart.

If you want the care side sorted at the same time, try this one straight after it’s replied:

Assuming that identification is correct, what does this plant need in terms of light, watering, soil and pruning? I'm in the UK, so please give advice suited to a UK climate.

And if the first answer feels vague or hedgy, push it:

How confident are you in that identification, and what specific features in the photo led you to that conclusion? What would you need to see to be more certain?

That last one’s useful because it forces the AI to show its working, and you’ll often see it admit uncertainty it glossed over the first time round.

Check the result

Right, so you’ve got a name back. Don’t just take it and run. A few checks first.

Search the Latin name it gave you, not just the common name, and look at images from a proper source like the RHS website. Does it actually match your photo? Leaf shape, colour, growth habit, all of it. Sometimes an AI will give you a plausible-sounding name that’s in the right family but not quite right.

Ask it directly for the two most likely look-alikes, if it hasn’t already given you them, and compare the differences it lists against your actual plant. If it says “the main difference is leaf serration” and your leaf’s smooth-edged, that’s a red flag worth chasing.

Then run the same photo through a second tool. If ChatGPT says one thing and PlantNet or Google Lens says another, you’ve learned something useful, namely that this one needs more digging before you trust either answer.

Watch for confidence that doesn’t match the evidence. If the AI says “this is definitely X” off one blurry photo of a leaf with no flower, no fruit, nothing distinctive, that’s overconfidence, not accuracy. A good answer hedges when it should hedge.

If it doesn’t work

Sometimes you’ll upload a photo and just get a vague answer, “this appears to be a flowering plant, possibly in the daisy family,” which is technically true and completely useless. Usually that’s a photo problem rather than a tool problem, so go back and get a clearer shot, ideally with a flower or seed pod in frame, since foliage alone is often not enough to narrow things down to species level.

Now, the privacy bit. When you upload a photo to any of these tools, whether it’s ChatGPT, Gemini, Claude or PictureThis, you’re sending that image to a company’s servers. If your photo happens to include your house number, your car, your face, or your garden with your address visible in the background, that’s now sat somewhere you don’t control. Crop tight to the plant if that’s a concern, or take the photo somewhere neutral.

And here’s the one that actually matters most: never eat, drink, or apply anything to your skin on the strength of an AI identification. Not PlantNet, not ChatGPT, not any of them. Foraging mistakes kill people, genuinely, every year, and poisonous look-alikes for edible plants are far more common than people expect. Deadly nightshade berries look enough like edible ones to fool someone who’s rushing. Wild garlic gets confused with lily of the valley and autumn crocus, both of which are seriously toxic, every single year, by people who thought they knew what they were looking at.

If you’re identifying something to eat or to use medicinally, treat the AI answer as a starting point for research, not a green light. Cross-check with the RHS, or better still, with an experienced forager or a botanist in person. This is the one place where “good enough” isn’t good enough.

Keep exploring

If you enjoyed this, have a look at “How to get AI to tell me what is in a photo”, which covers the same idea but for things other than plants, or “How to get AI to design my garden” if you’re thinking bigger than one identification and want to plan out a whole space. And if all this plant talk has you dreaming of somewhere warmer to see them growing wild, “How to get AI to plan a trip” is worth a read too.

Sources and review notes

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