Can AI design your garden?
Yes… and no.
With the easy availability of AI products such as Gemini and ChatGPT, many homeowners have found the idea of creating fun, unique garden designs with nothing but a smartphone, understandably appealing, especially when it’s completely free to use. I definitely understand this thinking, because it’s not a lot different to when I get myself stuck down Pinterest rabbitholes for hours, thinking I’m going to build a water feature out of cork (I might still do that).
There are some garden designers who have used AI for generating ideas, but for me, generating ideas is my job, and why would I deny myself the most enjoyable bit? I certainly don’t like the idea of cheating my clients out of the expertise they pay for! The one thing I do like about AI is its relatively decent ability to project regional lighting schemes onto my own drawings. This enables me to check how well my garden design will respond in the real world, at different times of the day. Or to turn a 3D render I’ve drawn into a photorealistic version to help a client visualise, but I am not stealing other designer’s work to do that.
Unfortunately, I just see don’t lighting checks as a good enough trade-off for using AI instead, AI can’t do my job.
Recently a friend, who is an established independent property developer, showed me a planting presentation all generated by AI. At first glance it was a beautiful presentation, and his client would no doubt have been impressed. The trouble was, when my friend asked me to look it over, there were three problems obvious to me, as a professional garden designer:
1- the planting images had been optimised for unrealistic aesthetics. This made a very pretty “menu” for the client to enjoy, but didn’t reflect the true size, shape and colours the plants would have in reality. It also didn’t take into account which angles the human visitors to the garden would be viewing from, so the layouts were unappealing and would not be easy to enjoy from the windows of the home.
2- worse still, AI had used whatever information was available online (which could have been the RHS website, or an inexperienced layperson’s YouTube video; AI can’t tell the difference) to draw a “planting plan”, but had not calculated the optimal planting distances to allow for healthy establishment and a cohesive aesthetic. It also couldn’t account for seasonality, so the garden would be dull and less appealing for over half the year. That is a lot of money to spend on plants that end up looking tatty or sparse.
3- even worse still, AI did not have the “training” to plant according to conditions. I take into account soil type, light, PH, wind, etc; all the horticultural elements necessary in varying degrees for different plants! In this plan, the shiny, striped grass (sedge) Carex oshimensis looked dazzling next to billowing Lavender, but one comes from damp east asian woodland and one comes from scorching hot Mediterranean rock crevices. They can’t both have their needs met in the same planting space, so one plant is definitely going to be working harder to survive than the other. When I design a planting plan, I want to make maintenance easier, not harder. Again, this wastes time and money, but not just once if they have to plant again!
Being the good friend I am, I did give him some tweaks to make the plan work (and I didn’t charge him for the service!), but this is where homeowners who want to build a beautiful new garden need to be careful. None of us know what we don’t know. That’s why using AI to design your garden isn’t necessarily a good idea.
At GI Landscapes, the entire landscaping team understands from experience, that every garden makeover needs an expert human eye. Once the initial “wow” of a new garden has passed, badly planned details suddenly become noticeable, and over time they become intolerable.
An example of this was in a lovely garden I visited here in Kent. I saw the “before” pictures, so I could understand how much of an improvement the new garden was, especially framed by the owners’ pretty cottage planting, but I also noticed something before they pointed it out and it was a shame to hear how much it was annoying them a year on from the finished build. It was an uneven paver. Just one. It would be tempting to blame the landscaper here (it was in fact, a builder, not a landscaper; perhaps a topic for another day), but it was the result of on-the-spot, lesser-of-many-evils decisions, due to an unfavourable clash of budget, time and a miscalculation of rise and run on a slope. Of course, it can be fixed, but I’m sure you’ve heard the old adage, ‘Buy cheap, buy twice’? A human garden designer would have avoided that AND stayed in budget and on time.
So ignoring my obvious bias against a product that claims to do my job, this is a bit of cautionary advice to let AI do what it does best; performing simple tasks to make human jobs easier; but when it comes to the creativity, horticulture and construction knowledge of garden design, we can do much better than that.
Til next time!
Mel