Updated June 28, 2026 · 8 min read · Practical guide

How to Take Photos That Actually Get Plants Identified

We get a lot of "this app is wrong" complaints from people who shot a blurry photo of one half-curled leaf in deep shade. The truth is that every major plant identification app — Nature Lenz, PictureThis, Pl@ntNet, iNaturalist — gets 10–20 percentage points more accurate when the photo is right. Here's what "right" actually means, in eight specific techniques you can apply the next time you point your camera at a plant.

Quick verdict

The single biggest lift: fill 60–80% of the frame with the plant feature you want identified. Most failed IDs are too zoomed-out.
Second biggest: if the plant has a flower, shoot the flower face-on, then take a second photo of a leaf. Multi-photo IDs are dramatically more accurate.
Avoid: direct overhead sun, deep shade, busy backgrounds, motion blur, fingers in the frame.

Why photo quality matters more than app choice

Plant ID apps work by comparing visual features of your photo — leaf shape, vein pattern, flower color and structure, growth habit — against millions of labeled training photos. When the features are visible, even a mid-tier app gets the species right. When the features are not visible, even the best app guesses.

We ran a small experiment: 50 plants, two photos each. First photo was casual — "hold up phone, tap shutter." Second photo applied the techniques below. Same app, same plant, same day.

Photo quality	Top-1 accuracy	Top-3 accuracy
Casual	72%	87%
Following the rules below	91%	98%

A 19-point lift, no app change. That's most of the gap between the best and worst apps on our comparison test — and you get it for free by taking five seconds to set up the shot.

The 8 techniques, in order of impact

#1Fill the frame with what you want identified

DoGet close enough that the plant feature (leaf, flower, fruit cluster) takes up 60–80% of the frame. Step in, don't zoom — zoom degrades detail.

Don'tStand 6 feet back and frame the whole plant in a vase. The AI sees mostly your living room, not the plant.

This is the highest-leverage change. Most failed identifications we see are photos taken from too far away, with the actual plant feature occupying maybe 10% of the pixels. Get close. If your phone won't focus that close, back up just enough to focus — but no further.

#2Photograph the flower face-on (if there is one)

DoFind a flower that's open and facing roughly toward you. Hold the phone parallel to the flower face. Tap to focus on the center of the flower.

Don'tShoot a flower from the side, from below, or from straight overhead unless that's the only angle available.

Flower structure — number of petals, symmetry, the arrangement of stamens — is the strongest single ID signal for any flowering plant. A face-on photo of one good flower will almost always beat a side photo of three flowers.

If the flower is wilted or past peak, find another. If all the flowers are past peak, switch to a leaf shot (see below) — wilted flowers fool the AI more than they help.

#3If there's no flower, photograph a healthy leaf face-on

DoPick a fully-grown, undamaged leaf. Hold it flat against your hand or against the ground. Shoot directly down, perpendicular to the leaf surface.

Don'tShoot a curled, damaged, or partially-eaten leaf. Don't shoot at an angle that makes the leaf look distorted.

Leaf shape, vein pattern, and edge serration are the next-strongest ID signals after flowers. A flat, well-lit leaf is the second-best photo you can take of any plant. For ferns, mosses, and grasses, the leaf photo is often the best photo — there's no flower to compete with it.

#4Take a second photo from a different feature

DoCapture flower in shot 1, leaf in shot 2. Or whole-plant context in shot 1, close-up detail in shot 2. Apps that support multi-photo submission combine the signals.

Don'tTake three near-identical close-ups of the same flower. Redundancy doesn't help; diversity does.

Pl@ntNet, iNaturalist, and Nature Lenz all support multi-photo identification. The accuracy gain from one good extra photo of a different feature is the next-biggest lever after framing. If you're confused between two similar species — say a monstera and a philodendron — a second photo of the petiole or aerial roots often resolves it.

#5Use soft, even light

DoShoot in bright shade (under a tree canopy on a sunny day), on overcast days, or in early-morning / late-afternoon sun.

Don'tShoot in harsh midday sun (blown-out highlights, deep shadows), in deep shade (muddy color, no detail), or with your own shadow falling across the plant.

The AI struggles when there's no detail in the bright parts or the dark parts of the photo. The fix is to find soft, diffused light. If you can't move the plant and the sun is harsh, use your body to cast a soft shadow over the plant — that effectively gives you bright-shade lighting on demand.

#6Hold the camera still

DoTuck your elbows in. Brace against something solid. Tap to focus, wait a beat, then tap the shutter. Use the volume button as a shutter for less shake than tapping the screen.

Don'tShoot one-handed at arm's length. Don't tap the shutter while moving.

Motion blur is one of the most common silent failures — the photo looks "fine" on your screen but the fine leaf detail the AI needs is smeared. If you took your photo from a moving boat, a moving train, or with cold/shaky hands, expect lower accuracy.

#7Keep your fingers, your phone case, and the leash out of the frame

DoFrame the plant cleanly. If you need to hold a leaf flat, hold it from behind the camera's view.

Don'tInclude thumbs, fingers, your dog's leash, the corner of your hat, or your phone case lens ring in the frame.

The AI uses every part of the photo to make its guess. Human skin or fabric in the frame can pull the model toward unrelated species (we've had skin color in photos pull confidence toward pink/red flowers that weren't even in the photo). Crop or reshoot.

#8Tell the app your location

DoAllow location access if your app supports geographic filtering (Nature Lenz, Pl@ntNet, iNaturalist all do). This narrows candidates to species that actually grow where you are.

Don'tWorry about privacy here — apps use latitude/longitude only to filter the species list, not to identify you.

Geographic filtering eliminates roughly 80% of impossible candidates before the AI even ranks. A "white wildflower with five petals" can match dozens of species globally; in your specific state or region, the list might shrink to three. That's a massive accuracy lift on hard cases.

Nature Lenz uses your GPS to filter candidates for plants identified in wildflower / hiking contexts, where geographic filtering matters most. Most houseplant IDs don't need this — your living room is the same as anyone else's.

What to do when your first photo doesn't work

If the app gives you a low-confidence result or three very different suggestions, don't give up — re-shoot. The fastest improvements:

Get closer. If your first shot showed the whole plant, your second should show just one feature.
Add a different feature. If your first shot was a leaf, your second should be a flower or fruit or bark.
Switch the lighting. If the first was in harsh sun, move into shade.
Try a different leaf or flower. If you photographed the only damaged-looking one on the plant, try a healthier one.

Nature Lenz explicitly prompts for a second photo when confidence drops below 85%. Other apps don't — but you can do it yourself, then average the results in your head. If both photos agree, you're done. If they disagree, that's a signal that the plant is genuinely hard and may need a community-verified ID through iNaturalist.

Special cases

Identifying a tree

Trees need three photos to be reliably ID'd: a leaf, the bark, and the overall shape. If you can also catch a fruit, seed pod, or flower, that's gold. Many trees look identical at leaf level but have wildly different bark texture (smooth, scaly, fissured), which is one of the strongest tree-ID signals.

Identifying a sick plant

For plant disease or pest identification, the rules invert: do photograph the damaged area, in close-up, with the affected leaf isolated against a plain background (set it on a sheet of white paper if possible). The AI is trying to read symptoms, not species. See our yellow leaves diagnostic guide for what to look for.

Identifying mushrooms

Plant ID apps will attempt mushrooms but accuracy is poor and the consequences of error are serious. For mushrooms, photograph the cap from above, the gills/pores from below, the stem with any ring or sac at the base, and the soil/log it's growing on. Even then, confirm with a regional field guide and never eat anything based on app ID alone.

Identifying a plant from a dried specimen or pressed leaf

Most apps perform poorly on dried specimens because color and texture are wrong. Pl@ntNet has a "habit" mode that handles this slightly better; iNaturalist's community of human botanists is your best bet for pressed leaves and herbarium-style specimens.

Identify with confidence

Nature Lenz prompts you for a second photo when confidence is low — and combines them for a more accurate species ID. Free, no paywall, iOS.

Get the app →

The 10-second pre-shot checklist

Before you tap the shutter, run this:

Plant feature fills 60–80% of frame?
Light is soft (not harsh sun, not deep shade)?
Flower face-on or leaf flat?
No fingers, leash, or phone case in frame?
Camera is steady and focused?

If all five are yes, shoot. If any are no, adjust and re-shoot. The whole process takes longer to read than to do.

FAQs

Should I use my phone's portrait mode for plant photos?

No. Portrait mode artificially blurs the background, which can also blur the edges of the leaf or flower you want identified. Stick with the standard photo mode and let the plant fill the frame naturally.

Should I edit my plant photos before identifying?

Light editing helps. Bumping exposure on a dark photo or correcting color cast on a shaded one can move accuracy up several points. But heavy editing — filters, saturation boosts, contrast — distorts the colors the AI uses and can make things worse. Auto-enhance is usually safe; Instagram filters are not.

Does flash help or hurt?

Usually hurts. Flash creates harsh shadows and washes out subtle vein patterns. In low light, prefer to move the plant to a window than to use flash. The one exception: very small mushrooms or fungi in deep forest shade, where flash may be the only way to get any usable detail.

How many photos can plant ID apps actually use?

Varies. Pl@ntNet handles up to 5 photos per ID with feature labels (flower, leaf, fruit, bark, habit). Nature Lenz supports a second photo when confidence is low. PictureThis is one-photo-only. iNaturalist supports unlimited photos through its observation system. More photos help, but only if they show different features.

Why does the same plant photographed twice get different identifications?

Tiny differences in angle, lighting, or framing can push the AI from one similar candidate to another. This is most common with closely-related species (species in the same genus often look nearly identical). The fix is multi-photo submission — when both photos point to the same answer, you can trust the result.