I Asked AI to Design My Raised Vegetable Garden – Here’s What a Real Garden Designer Made of It

Like many enthusiastic gardeners, I’ve spent more than a few evenings staring at an empty raised bed wondering what exactly should go in it. Tomatoes? Herbs? Something ambitious like eggplant? There's too much to choose from, quite frankly.

At the behest of my editor, and in the spirit of experimentation, I decided to outsource the decision process to... well, to artificial intelligence. Which means, yes, I asked AI to design a planting plan for a standard 4-by-8-foot raised vegetable bed and waited to see what it came up with.

Now, an important caveat here: I'm deeply wary of AI (blame I, Robot). I don't even have my own account, so I had to borrow my partner's Gemini. The result looked impressively confident – a neatly organised grid of crops, a planting calendar, and practical advice on soil, watering and fertilizer.

Can AI Design a Raised Vegetable Garden?

At first glance, it seemed like exactly the kind of clear, beginner-friendly guidance many new gardeners are searching for. But how good was the advice, really?

In a bid to find out, I decided to embrace my suspicious side and ask a real (by which I mean a living, breathing, human) garden designer to take a look. And not just any old garden designer, either; I'm talking about the award-winning Zoe Claymore, who is an RHS Chelsea medallist, an NAS Ambassador, and a member of the editorial board of The Garden Design Journal.

Her verdict? Well, it wasn’t terrible... but it definitely wasn’t foolproof either. Let's dive on in...

What the AI Got Right

Before we dive into the bad, it's worth mentioning that the AI-generated plan for a raised vegetable garden does get a few fundamental principles right.

“All of the plants suggested are edible crops, which sounds obvious but is a good starting point,” she explains. “And the idea of placing taller plants toward the north side of the bed is broadly correct because it reduces shading on smaller plants.”

The timeline is also roughly in the right order for a growing season, and the overall concept (mixing vegetables, herbs and fruit in a compact bed) is a sensible approach for beginners.

“For a very simple layout with a small number of plants, it’s not a terrible starting point,” says Zoe. However, she warns that the plan relies on a lot of hidden assumptions that an inexperienced gardener might not realise.

“If someone followed it exactly without understanding gardening basics, there’s a reasonable chance they’d end up disappointed.”

Where the AI Planting Plan Falls Apart

Once Zoe looked more closely at the AI-generated plan for my raised vegetable garden, though, she started spotting problems. Lots of problems.

For starters? The layout itself isn’t especially balanced. “Who wants eight bean plants but only two tomatoes?” she says. “That’s not really how people tend to cook or eat.”

Some plants are also placed incorrectly for how they actually grow. Zucchini, for example, takes up far more space than the grid allows. “Courgettes [zucchini] have a spreading growth habit, so the numbers and placement here simply wouldn’t work.”

Spacing is another issue with the AI-generated design, with Zoe pointing out that the bed is extremely densely planted, and different varieties of the same vegetable can grow very differently.

“A Roma tomato behaves very differently from other varieties,” she explains. “Some beans grow bushy, some climb, some need more support. Without specifying varieties, the plan could be completely wrong depending on what the gardener buys.”

The Problem with AI's Garden Advice

Perhaps the biggest flaw is that the design doesn’t account for the garden itself. Instead, the AI assumes the raised bed is positioned in a particular direction so that taller plants sit on the north side – but that might not be true.

“Very few garden beds are perfectly aligned with the compass,” says Zoe. It also doesn’t consider important real-world factors such as sunlight levels, wind exposure, underlying soil conditions, pest pressure, or ongoing plant care.

“These are the details that determine whether a garden thrives or struggles,” she says, noting that without them, even a well-intentioned planting plan can miss the mark.

Zoe also flagged some environmental concerns in the advice, particularly with regards to the fact that AI recommended using peat-based compost, which many gardeners and retailers are moving away from because peat extraction damages fragile ecosystems and releases stored carbon.

“Gardening advice should be evolving alongside environmental knowledge,” she says. “Otherwise we risk encouraging practices that are harmful in the long term.”

If Zoe were designing the bed, she says the first step would be understanding the gardener themselves. “I’d start by asking what the client actually enjoys eating,” she says.

From there, she’d assess the site, taking care to look at sunlight, soil and orientation, before choosing specific plant varieties suited to those conditions. She’d also make a point of including perennial vegetables or fruit to reduce maintenance, edible or cut flowers to improve pollination and visual appeal, exact plant varieties suited to the local climate and planting zone, and proper spacing to ensure plants can thrive.

“I’d also recommend more sustainable gardening practices that reduce cost and resource use over time,” she says.

Raised Vegetable Garden Essentials:

So, can AI design your raised vegetable garden? Well, while Zoe doesn’t think AI is entirely useless, she does insist we start thinking of it as a starting point than a finished plan.

“I think AI-assisted design will probably enter the lower end of the garden design market,” she says. “It can provide quick, simple layouts for people on a tight budget.”

The problem comes, though, when people treat that output as expert advice. “AI is only as good as the data you give it and your ability to spot mistakes,” she explains. “If you don’t know what to check, you could end up making expensive errors.”

In that sense, she compares it to AI tools used in coding. “It works if you already have expertise. Without that knowledge, it can embed errors that become costly later.”

My AI-designed vegetable garden wasn’t a total disaster, I guess, but it also wasn’t the foolproof plan it first appeared to be (sometimes it pays to be suspicious!). And, as Zoe points out, when it comes to living systems like gardens, nuance matters.

“You still need human knowledge,” she says. “Otherwise you risk creating gardens that are inefficient, less biodiverse, and more expensive to maintain in the long run.”

Something well worth bearing in mind, then, the next time you plug a prompt into your ChatGPT account...