PlantCraft Guides · Meta Reading time ~10 min

How to Use the PlantCraft AI App — From Photo to Trust Score to Guide

The app is the field guide. These guides are the kitchen book. Here is how the two halves fit together.

PlantCraft AI does one thing the old field guides could not: it reads your specific photo of your specific plant, cross-checks it against four AI vision models and six botanical databases, and tells you exactly how confident it is and why. What it cannot do — what no software can do — is make the final call for you. That decision still happens in the kitchen, on your knees in the dirt, or over a cup of tea with an older forager who has been doing this since before you were born. This guide walks you through the mechanics of the app so you can use the confidence number correctly, not just trust it.

Before you open the app: the photo is everything

PlantNet, the visual ID engine behind PlantCraft, was trained on tens of millions of herbarium photographs and citizen-science observations. What kills a good ID is not a bad species — it is a bad photograph. Four shots is the minimum. Take them in this order, at this distance:

Whole plant from two feet back. You want the growth form — whether it is a basal rosette, an upright herb, a climbing vine, a shrub. This is the "habit" shot. Lay a hand or a boot next to it for scale if you can.
Leaf close-up, both surfaces. Flip the leaf. The underside carries hairs, glands, and vein pattern that the topside hides. Get within six inches if you can hold the camera steady. Natural light from the side is better than direct sun, which washes out color.
Flower, fruit, or seed head close-up. Get the whole inflorescence in frame — not just a single floret. If the plant is not in flower, photograph the seed head, the berry cluster, or the bare stem with buds. Something is usually there even out of season.
Habitat context shot from five feet back. What is this plant growing next to, and in what kind of soil or light? Woodland edge, roadside gravel, wet ditch, dry hillside — habitat narrows the field faster than most people realize. It also makes your library find more useful when you come back a year later.

The one thing the app cannot see: smell, texture, and taste. For families where a dangerous look-alike exists — the umbel family (wild carrot versus poison hemlock) is the canonical example — the nose and fingertips are still required. The app will flag the risk. You still have to do the check.

If you only have one shot and it is a full-plant image in decent light, upload it anyway. PlantCraft will do what it can and tell you how much coverage it detected. You can always add photos later — the app merges them into the same specimen and recomputes the score.

Uploading: the Add button and what happens next

Tap Add in the bottom navigation bar. The upload zone accepts one or more photos at once. If you have all four shots, drop them together — the app detects that they belong to one plant and stacks them into a single specimen, which earns better Photo Coverage than four separate uploads would.

After upload the pipeline runs automatically. PlantNet identifies the species. Four AI vision models independently examine the images and vote on whether they agree. The result arrives as a specimen card in your My Plants view, showing the common name, scientific name, Certainty Index score, and the V3 Forager Caution badge.

If you disagree with the identification, you have two options on the specimen detail page. Reconsider with notes lets you add context — "grows in standing water," "stem hollow, no spots" — and the AI re-runs the ID with that information. I'm telling you what this is accepts your declared species outright; the score pill switches to "USER" instead of a number, and the override is logged in the specimen's audit trail so the source is never hidden.

Reading the Certainty Index

Every specimen gets a Certainty Index from 0 to 100. The number is not a guess — it is a weighted sum of three independently measured components. Tap the score pill on any specimen card to expand the breakdown bar chart.

Component	Max pts	What is measured
AI Models	65	Four leading vision AI models examine your photos independently. Each model that agrees with the consensus species contributes points, calibrated against its own typical confidence threshold. A model only earns full credit when it hits its own "I'm really sure" zone. Full agreement across all four earns the maximum 65 points.
Photo Coverage	20	AI scans your photos and marks which of the four plant parts it can clearly see: roots, stems, leaves, and flowers/seed/fruit. A single sharp whole-plant shot can satisfy multiple parts at once — no manual tagging required. Three of four parts visible earns full credit. Mark a part N/A (amber pill flag) when it is genuinely unreachable — off-season flowers, buried roots on a mature tree — and the cap recalculates to not penalize the missing part.
Database Match	15	The identified species is checked against six botanical databases: iNaturalist, GBIF, a worldwide citizen-science plant-ID source, USDA, Kew POWO, and Wikipedia. Sources with reference photos get AI vision comparison — 85% or greater visual similarity counts as a match. Text-only sources match when they confirm the species name. Four of six matches earns full credit and the green VERIFIED badge.

What the score bands mean in practice

80 – 100 Strong agreement across AI models and databases. Still verify before eating. Strong agreement among machines is not the same as the certainty you need for a member of the umbel family.

50 – 79 Worth a second look. Add more photos, check Photo Coverage gaps, or tap Re-identify. Models partially agree but are not unanimous.

0 – 49 Treat as a guess. Consult a local expert before any consumption. The machine does not know what this is with meaningful confidence.

A 100/100 Certainty Index is not a guarantee. It means the machines agree strongly. PlantCraft is an information aggregator, not a botanist standing next to you in the field. The score tells you how well the software can see what you photographed — nothing more and nothing less.

The V3 Forager Caution badge

Separate from the Certainty Index — and more important for foraging decisions — is the V3 Forager Caution Check. This runs after identification and asks a different question: does this species have a dangerous look-alike, and do any of the four AI models think your photos could be that look-alike instead?

The result appears as a colored badge on the specimen card and at the top of the detail page:

GREEN — No dangerous look-alikes detected by the AI verifiers. The four models agree the specimen does not visually overlap with any toxic, lethal, or severe-dermatitis species in the database.
YELLOW — Look-alikes exist. Some may be distinguishable on close inspection of specific features. The badge opens a detail panel naming the candidate species and the features to check. Read it before you do anything else with the plant.
RED — Dangerous look-alikes flagged. One or more of the AI models cannot confidently rule out a toxic species based on your photos. Verify every visual feature carefully — stem spots, hollow versus solid, hair pattern, leaf margin, smell — before any consumption. A RED badge is not a verdict. It is a stop sign that requires human verification.

The V3 check runs automatically after identification for most specimens. If you see a "Run V3 Check" button instead of a badge, tap it — it takes about thirty seconds and cross-checks against the four models fresh. You can also tap the Re-check button on any existing badge to re-run with the latest AI models.

RED badge protocol: Do not eat, do not nibble, do not make tea. Add more photos — close-up of stem (look for spots, hairiness, hollow vs. solid), close-up of leaf underside, habitat context. Re-run the V3 check with better coverage. If the badge stays RED, take the specimen to an experienced local forager or your county extension office before any use.

From species page to the skill guides — this catalog

The app identifies the plant. The guides teach you what to do with it. The two halves are connected.

On a specimen detail page, scroll past the Certainty Index breakdown and the V3 badge to the Knowledge section. Any specimen with a proven or likely status will surface links to relevant guides from this catalog — tincturing, drying and storage, oxalate leaching for high-oxalate edibles, seasonal harvest windows, look-alike verification checklists. Tap Guides in the bottom nav bar to browse the full catalog directly, organized by topic: identification, safety, processing, preservation, herbal use, cultivation, legal.

Think of it this way: the app tells you what the plant is and whether it is safe to handle. The guides tell you what to do once you are back at the kitchen table. You would not fry chanterelles based on a field guide description alone, and you would not try to tincture a plant based on an app score alone. Both sources together are where the old forager knowledge and modern machine vision earn their keep.

Logging a find to your personal library

Every photo you upload creates a specimen in your private library, visible under My Plants. The library is yours — other users cannot see it. Each entry records the identification, the Certainty Index, the V3 caution rating, the photo set, and any notes you added.

To get the most out of your library:

Add location notes in the Notes field. "North-facing slope, moist soil, edge of cedar thicket" is more useful a year later than GPS coordinates alone. The Locations tab in the nav bar gives a map view once you start tagging finds.
Use the Seasonal tab to see what should be harvestable right now in your region, based on what your library already knows is growing near you.
Check the Library tab (the scroll icon in the nav bar) for the curated plant encyclopedia — browse species pages, edibility notes, historical uses from the public-domain book passages, and the look-alike entries for any species in your library. The Bibliography tab shows the source books behind every claim.
Run Re-identify if you add photos later in the season — a specimen you found as a basal rosette in March might hit a much higher Certainty Index once you photograph the flowering stalk in June.

Contributing to the community forum (Tier 2 and above)

If you are on a paid tier, the specimen detail page shows a Community section below your personal notes. This is where confirmed identifications and field observations get shared with other PlantCraft users — not raw AI output, but finds that have passed a minimum threshold of evidence.

To contribute a find to the community:

Make sure the Certainty Index is at least in the yellow band (50 or above). The community forum is not for guesses.
Add a field note describing what you observed beyond what the photos show — smell, texture, habitat, associated species. This is the part the camera cannot capture.
Tap the Share to Community button. Your find will be visible to other users browsing that species, alongside your Certainty Index and caution rating but not your personal location unless you choose to share it.

Community posts carry the same disclosure banner as AI-rendered content — readers can see whether the underlying data is from a confirmed field observation, an AI-generated knowledge card, or a passage grounded in the public-domain book corpus. The source is never hidden.

The forum is the apprentice tier above the books and below the local expert. Read other people's field notes on a species before you harvest it for the first time. Two dozen people who found the same plant in your county across four seasons know things a single trip to the field will not show you.

Cross-links

← All guides Open the app → Start here. Every other guide assumes you know how to read the Certainty Index and the V3 caution badge.