26 AI Prompts for Landscape Architecture and Garden Design

TL;DR: Landscape architecture prompts fail the moment they describe "a nice border" instead of a border you could actually source and plant: a real species, a hardiness or climate zone, a mature size, and a season of interest. Below are 26 copy-paste prompts split between text-model plant schedules, specs and narrative, and image-model concept work, plus two honest limits on what neither can verify for you. If you are drawing the building as well as the garden, start with our AI prompts for architects and our architectural rendering prompts.

Four sibling posts on this site already cover building design: a step-by-step ChatGPT workflow for architects, a 30-prompt library for architectural rendering, one for interior design concepts, and one for floor plans and space planning. Landscape work isn't a fifth flavour of the same problem, so this page earns its own space rather than reskinning any of them. One thing first, same as every page in this cluster: Prompt Architects writes the prompt, not the render or the drawing. Everything below is text to paste into whichever text or image model you already pay for.

Jump to: planting (1-3) · safety and legal checks (4) · native and habitat briefs (5-6) · time and maintenance (7-9) · ground plane and hard landscape (10-12) · planning and client narrative (13-14) · image prompts (15-26).

Which Prompts Work for a Home Garden vs a Professional Scheme?

Both. If you are a homeowner asking AI for garden design ideas, use the beginner version below and skip the zone jargon. If you are a practitioner, go straight to the numbered prompts and treat the output as a draft to check.

Garden design for beginners: a plain-language prompt

I'm a beginner designing a small garden border, about [size, e.g. 4 m by
1.5 m], at the front of my house in [town or region]. It gets [sun
hours or "full sun / part shade / shade"] and the soil is [clay / sandy /
"I don't know"]. I want it to look good in most months and not need
much work. Suggest 6 to 8 plants, say roughly how big each one gets,
when it flowers or looks good, and what to plant it next to. Tell me
which of your suggestions I should double-check with a local nursery.

Which Tasks Need a Text Model and Which Need an Image Model?

What Makes a Landscape Architecture Prompt Different From a Building One?

A building is mostly finished the day the scaffolding comes down. A planting scheme is not finished for a decade, and most of what makes it good or bad only shows up over that time. Four things carry the weight here that a building prompt doesn't have to think about.

Planting is species-specific and seasonal. A model will happily generate a beautiful border full of plants that would never actually grow together in your climate, your light, or your soil. Naming a real species, a hardiness or climate zone, a mature size and a season of interest is the whole craft, and checking that the combination actually works on your site is your job, not the model's.

Time is a dimension buildings don't really have. A scheme has a year-one look, thin and gappy, and a year-fifteen look, closed and shaded, and a prompt that only asks for one of them shows a client half the truth. Ask for both, on purpose.

Seasonality runs underneath that same clock, at a faster tempo. The same bed reads as a different place in spring growth, high summer bloom, autumn seed head and colour, and bare winter structure, and a scheme that only performs in one of those seasons is a scheme that looks abandoned for the other three.

The ground plane and hard landscape carry decisions a building's facade doesn't have to make: grading and drainage falls, permeability, retaining walls, and steps or ramps resolving a level change, plus materials named by specification, setts, resin-bound gravel, corten, permeable paving, rather than the adjective a generic prompt would reach for instead.

What Goes Into a Real Planting Plan or Planting Schedule Prompt?

"Lush, mature planting" is not a prompt any more than "modern building" is one. It's a placeholder for six decisions you haven't made yet, and a model will make all six for you, usually badly and identically every time.

A working planting prompt names, at minimum: the real species or genus, the site's light and soil, a hardiness or climate zone (USDA, RHS, or your local equivalent), the plant's mature size, and its season of interest. Leave any of those out and the model substitutes something generic, most often a photogenic combination that would fail on a real site within two seasons.

1. Species-specific plant schedule generator · text model

I'm planting a [border type, e.g. mixed shrub and perennial border] in
[climate/hardiness zone] with [light: full sun / part shade / shade] and
[soil type: e.g. free-draining sandy loam]. Suggest a plant schedule of
[number] species covering [structure/mid-layer/groundcover], each with:
common and botanical name, mature height and spread, season of interest,
and one companion it's commonly paired with. Flag any species you're
not confident is genuinely suited to this zone rather than guessing.

Illustration only, not a measured result: this is the shape of answer to look for, with approximate sizes. Check every row against a nursery or extension-service source for your site.

A good answer also tells you which of these it is unsure about for your zone. If it never does, ask again.

2. Right-plant-right-place suitability check · text model

Here's a plant list I'm considering for this site: [list species]. Site
conditions: [zone, light, soil, drainage, exposure]. For each plant,
reason through whether the stated conditions are a realistic fit, note
anything you're uncertain about, and flag any pair that will compete
for the same light or root space as they mature.

3. Four-season interest audit · text model

Here's my current planting list: [list species with season of interest
for each]. Build a month-by-month table showing which months have no
species carrying visual interest, and suggest one or two additions per
gap, matched to the same [zone, light, soil] as the existing list.

The difference shows up immediately in the request itself. "Plant a shady border" produces a border. "Woodland shade border, USDA zone 6b, free-draining loam, three-season interest from Hakonechloa macra, Astilbe and Helleborus" produces something you could actually take to a nursery, a genuine line-up you'd still confirm against your specific site, and every clause in it is a decision you'd have had to make anyway.

Can AI Tell You Whether a Plant Is Safe, Legal, or Invasive?

That checklist is still genuinely useful, framed correctly.

4. Invasive-species and legal-status due-diligence checklist · text model

I'm considering these species for a scheme in [country/region]: [list].
List the categories I should check with my local biosecurity or
planning authority for each one (examples: listed as invasive/noxious,
sale or planting restrictions, toxicity to pets or livestock, protected
native status) without stating any specific answer, since those vary by
jurisdiction and I'll verify each one separately.

5. Native and pollinator planting brief · text model

I want a planting brief favouring native and pollinator-friendly species
for [region/climate zone], [site conditions]. List candidate species by
layer (tree/shrub/perennial/groundcover), noting bloom season for each,
so I can build a succession of nectar sources across the growing season.
Flag anything I should verify as genuinely native to my specific region
rather than native to the wider country.

6. Habitat and biodiversity intent narrative · text model

Write a short narrative describing the ecological intent of this
planting scheme: [describe layers, native proportion, structural
diversity]. Frame it as intent and reasoning, not a measured outcome,
since nothing about this text or any image of the scheme demonstrates
actual ecological performance, nectar yield or habitat use once built.

Why Ask for a Year-1 View and a Year-15 View?

A building doesn't really have this problem. A planting scheme has two honest photographs, and they look like different projects. Year one is thin: small nursery stock, visible mulch, gaps between plants sized for growth nobody can see yet. Year fifteen is a closed canopy, a different scale of shade, and a structure that has probably outgrown at least one original decision. Ask for one and you get half the truth; a client who only sees the year-fifteen render is signing off on a maturity that won't exist for a decade.

7. Growth-projection narrative, year 1 vs year 15 · text model

Here's a planting plan: [list species with mature size and growth
rate]. Write two short descriptions of how this scheme will read: at
planting (year 1) and at estimated maturity (year 15, or note if a
species takes longer). Cover canopy closure, shade cast, and anything
that will need thinning or removal as it matures.

8. Maintenance regime by year · text model

For this planting scheme: [list species and layers], draft a
maintenance regime as a table with columns for year 1-2, year 3-5, and
year 10+, covering pruning, thinning, mulching, and any species you'd
expect to need replacing or dividing on a specific cycle.

9. Phased planting plan memo · text model

I want to phase this scheme rather than plant everything at once:
[list full species list and site]. Write a memo proposing what goes in
now versus what gets added in a later phase (e.g. once structure plants
establish), and the reasoning for each phase split.

How Do You Prompt for Levels, Drainage, and Hard Landscape?

This is the section our floor plan and space planning post already made the case for on the building side: image models have no geometry engine, so they can't verify that a floor plan's walls close or its dimensions add up. The same limit applies to a garden's levels drawing, for the same reason. What a text model can do is hold your fall, drainage points and paving areas as structured facts and flag where the logic doesn't work, which is a genuinely different and useful job. None of the five prompts below draw anything. They reason in words about a problem usually solved with a tape measure and a spirit level, which is exactly the kind of check a text model is built for.

10. Grading and drainage reasoning check · text model

Here's my site: [describe existing levels, falls, and where water
currently drains]. Here's my proposed layout: [describe hard and soft
areas]. Check this for obvious drainage problems: water with nowhere to
go, a fall directed toward a building, or a paved area with no outfall,
and flag each one rather than assuming it's fine.

11. Permeability and sustainable drainage narrative · text model

I'm specifying [list hard surface areas and materials, e.g. permeable
resin-bound gravel, standard block paving] for [total area]. Write a
short narrative for a client or planning submission explaining the
permeability strategy, and flag any area where an impermeable surface
choice increases runoff I should reconsider.

12. Hard landscape materials specification draft · text model

Draft a specification section for these hard landscape materials:
[list, e.g. granite setts, resin-bound gravel, corten steel edging,
permeable block paving]. For each, cover: typical application in this
scheme, a base/sub-base note to flag for the engineer, and one
maintenance consideration. This is a draft for a professional to
review, not a finished spec.

13. Planning-statement narrative · text model

Write a planning-statement paragraph describing this landscape scheme's
approach to [levels/drainage/planting/access], for a [project type] in
[jurisdiction]. Ground it in the actual decisions I've described:
[list], and flag anywhere the statement makes a claim I haven't
actually confirmed, like a specific permeability percentage.

14. Client brief translator · text model

Here's my client's rough wish list for their garden: [paste notes,
however messy]. Turn this into a structured landscape brief: zones,
approximate hard vs soft split, level changes to resolve, and any
adjacency the client cares about. Flag anything vague enough that it
needs a follow-up question before it belongs in the brief.

What Can Image Models (Midjourney and Others) Do for a Landscape Scheme?

Concept and mood, not documentation. The prompts below use Midjourney-style --ar parameters, so delete or swap those for another image model. The same lighting and camera vocabulary that works for buildings works here, so we're reusing it rather than re-teaching it: our lighting vocabulary guide covers golden hour, blue hour and 38 more terms with copy-paste fragments, and our camera and lens terms guide covers focal length and framing the same way.

Scale works a little differently in a garden than on a facade. A distant figure does the same job here that it does for a building, but a garden carries its own scale references that read more naturally: a timber bench, a wheelbarrow, a retaining wall of a known height. Reach for one of those before adding a person, and keep named species to two or three per prompt, the same restraint that stops a facade turning into visual mush applies here to a border.

15. Border concept, spring · image model

Photoreal garden border concept, [planting: e.g. mixed shrub and
perennial border] in early spring, fresh green growth and spring
bulbs, soft overcast daylight, eye-level view from the path, 35mm
lens, --ar 3:2

16. Border concept, high summer · image model

Photoreal garden border concept, same [planting] at peak summer bloom,
full foliage, bright midday light with dappled shade, eye-level view
from the path, 35mm lens, --ar 3:2

17. Border concept, autumn colour · image model

Photoreal garden border concept, same [planting] showing autumn seed
heads and foliage colour, low golden-hour light raking across the
bed, eye-level view from the path, 35mm lens, --ar 3:2

18. Border concept, winter structure · image model

Photoreal garden border concept, same [planting] in winter, structural
stems and seed heads left standing, frost on the ground, flat overcast
light, eye-level view from the path, 35mm lens, --ar 3:2

19. Year-1 planting concept · image model

Photoreal garden concept of [scheme description] at year one, small
young nursery stock, visible gaps between plants, mulched bare soil
showing between them, soft overcast light, three-quarter view, 35mm
lens, --ar 3:2

20. Year-15 planting concept · image model

Photoreal garden concept of the same [scheme description] at estimated
year fifteen maturity, closed canopy, established scale, deeper shade
beneath larger specimens, same camera position and light as the year-one
version, three-quarter view, 35mm lens, --ar 3:2

21. Corten steel retaining wall and steps · image model

Photoreal landscape render of a [corten steel] retaining wall with
integrated steps rising through a [planting: e.g. grass and perennial]
bank, low afternoon light catching the rust patina, eye-level view
from the base of the steps, 35mm lens, --ar 3:2

22. Permeable resin-bound gravel courtyard · image model

Photoreal landscape render of a courtyard in [permeable resin-bound
gravel] bordered by [planting], flat overcast light, true-to-material
colour, straight-on view from the entrance, 50mm lens, --ar 4:3

23. Native meadow and naturalistic mood · image model

Photoreal naturalistic meadow planting concept, [native grass and
wildflower species for your region], informal drifts rather than
rows, soft morning light with dew, low-angle view at grass height,
50mm lens, --ar 3:2

24. Level-change concept illustration (not a levels drawing) · image model

A loose concept illustration, explicitly not a technical levels
drawing, showing a stepped garden descending through [number] level
changes with [planting/retaining materials], sketch-style linework, no
dimensions or spot heights, soft watercolour palette, --ar 4:3

25. Illustrative site-section-style concept · image model

An illustrative section-style concept sketch, explicitly not a
technical construction section, showing how [garden zones] stack
against a sloping site, flat colour blocks, no scale reference or
dimension text, clean line-weight, --ar 3:2

26. Dusk garden lighting mood · image model

Photoreal garden render at dusk, [hard landscape and planting],
warm low-level path lighting, blue hour sky still holding colour,
eye-level view from the entrance, 35mm lens, --ar 3:2

What Can't These Prompts Do?

Two honest limits are worth repeating on their own. First, image models cannot produce an accurate planting plan or a levels drawing, for the same reason they can't produce an accurate floor plan: nothing underneath represents a plant as an object with a real mature spread, or a slope as a surveyed number. These are concept and communication images, not documentation. Second, never let a model's confident-sounding answer stand in for a hardiness, toxicity, invasiveness or legal-status fact. Ask it for the category of check, then verify with your local authority or a qualified horticulturist, every time.

How Do You Keep a Working Prompt From Getting Lost?

Once a plant schedule or a materials spec actually works, the failure mode is administrative, not creative: it lives in an email, then a notes app, then nowhere, and six months later someone retypes it from memory for the next project and it's slightly worse. That's not hypothetical for us. In our own analysis of 2,170 customers on July 15, 2026, the average customer used 1.16 of our 7 features, and library adoption dropped from 69.7% on the core enhancer to 23.8%. Most people never get to the storage step.

Freeze the prompts that work for your practice: your usual zone, your default aspect ratio, the three materials you specify most often. Save them once as reusable prompt templates, and a schedule request becomes a structured output you can scan rather than a paragraph you have to re-read for the one species that doesn't fit. That's what Prompt Architects is for: we generate, refine and store the prompts, not the render or the drawing. The free plan includes 5 prompt enhancements per day, forever, and the paid plans add a bigger saved-prompt library. Check current pricing before you decide.

For more image-prompt patterns, see our interior and architecture image prompts, the Midjourney v8 architecture parameters guide, and prompts for design competitions and client presentations. If the project also includes the building itself, our architecture prompt generator covers massing, facade and site the same way this page covers planting, ground plane and hard landscape.