How do I write a garden brief when I don't know what I want or what good design even looks like?
We had a scruffy garden, a rough plan, some photos and a vague sense that it should be nicer. What we did not have was any way to turn that into something anyone could design or build against.
What it was costing us
Weeks of circling. Every idea sounded reasonable, nothing could be ruled out, and there was no way to say no to a beautiful suggestion that had nothing to do with how we actually live.
What we were trying to do
Work out what we actually wanted, before spending anything.
We had a rough measured plan, a set of photographs of every part of the garden, and a shared sense that the space was scruffy and did not work. Beyond that, nothing. Neither of us has any design training. We could not have told you what good looked like, only that we would probably recognise it.
What we tried
We looked at a lot of pictures. Instagram, Pinterest, a couple of gardens we had visited. It made things worse rather than better, because everything looked lovely and nothing told us what to do with our own space.
Then we asked AI to design the garden, which is what everyone does first. It produced something confident and generic — the average of every garden it has ever seen. Pleasant, plausible, and completely disconnected from us. We could not say why we did not want it, which was itself the problem: with no stated intent, there was nothing to judge against.
What went wrong
We went round in circles for weeks, and the circling was invisible because every individual conversation felt productive.
Without a brief there is no basis for saying no. A beautiful suggestion arrives and the only available response is "yes, nice" or an unexplainable "no". So ideas accumulated instead of resolving, and we kept revisiting settled questions because nothing had actually been settled — we had only stopped talking about them.
The real cost was not the weeks. It was that we were about to start spending money with no way to tell whether anything we bought served us, or was simply the last attractive thing we had been shown.
Why it happens
- 1Why can't I just write down what I want?Because taste is stored as recognition, not as specification. You know it when you see it, which is a completely different faculty from being able to describe it in advance to somebody who cannot see inside your head.
- 2Why doesn't collecting inspiration pictures fix that?Because a photograph of someone else's garden encodes their plot, their light, their budget and their life along with the bit you actually liked. Without separating those, you end up chasing an aesthetic that cannot exist on your site.
- 3Why is a designer's first meeting almost entirely questions?Because the deliverable of that stage is the brief, not the drawing. Extracting what a client wants, including the things they have not realised they want, is the professional skill. The pretty plan comes later and is comparatively mechanical.
- 4Why does AI help with that particular job?Because it is an inexhaustible interviewer. It does not get bored at question forty, it does not mind that you lack the vocabulary, and you are not embarrassed telling it you have no idea what a pleached hornbeam is. Volume and patience are exactly what this stage needs and exactly what a paid hour discourages.
- 5So why does asking AI to design your garden fail?Because you are asking for the answer before the question exists. With nothing to go on it produces an average of every garden it has ever seen, which is why those first renders look simultaneously impressive and like nothing you would ever build.
What actually works
Stop asking AI for the answer. Ask it for the questions.
The move that broke it open was two-part. First we gave it everything we had — the measured plan, the photographs of every zone, and a plain description of what we thought we were planning. Then, instead of asking for a design, we asked what a world-class garden designer would do with this, and what they would want to know before drawing anything.
That reframing changes the output completely. Asked to design, it designs. Asked what an expert would ask, it produces the interview.
Then we let it run that interview on us. It questioned us both, separately and together, about how we use the garden now, how we would like to, what we hate, who visits, what the evenings look like, what we would miss if it were gone. Question after question, for far longer than we would ever have sat with somebody we were paying by the hour. That is the part that would not have happened any other way: there was no clock running and no embarrassment about not knowing the words.
The output of that interview became the brief. Feelings, functions, fixed constraints, and the things we had ruled out and why.
And then the brief stopped being a document and became the spec. Every plan, render, quote and compromise since has had to answer to it. It is what lets us reject a genuinely attractive idea in one sentence, and it is the reason the project has direction rather than momentum. It is published in full, and it is still being revised, because a brief that never changes is one nobody is really using.
You can read the brief this produced. It is the single most useful thing we made, it cost nothing, and it happened before a spade went anywhere near the ground.
- A brief is not a design. It records what the garden must do and how it should feel, not what it looks like.
- It only surfaces what you can recognise when prompted. It will not invent a taste you do not have.
- It is worth nothing without the real site facts beside it. Preferences that ignore your aspect, levels and drainage produce a wish list, not a brief.
- Ours took several sessions and it is still being revised. Anyone promising a finished brief in one prompt is selling something.
Words used above
- Brief
- A written statement of what the garden must achieve, how it should feel, and what is fixed. The thing every later idea has to answer to.
- Expert frame
- Asking what a world-class practitioner would ask or do, rather than asking for the output directly. It shifts the model from producing an average answer to producing the questions behind a good one.
- Spec
- The agreed description that later work is checked against. Once the brief is written down, it becomes the spec for plans, renders, quotes and decisions.
The explanation is free. The instrument is paid.
You now know why it fails. Here is what fixes it.
- BriefingThe Brief Interview — The exact prompt sequence that turns AI into an interviewer, plus the forty questions it should be asking you and the four it must ask before anything else. Run it in an evening with a glass of wine.
- CourseWrite the Brief That Everything Else Answers To — Ninety minutes, ending with a written brief for your own garden covering feelings, functions, fixed constraints and rejections, in a form a designer or a landscaper could actually quote from.
- BlueprintBuild a Brief Builder for Domestic Projects — The full build spec for the tool: guided interview, structured brief output, revision history, and a checker that tests any later plan against the brief it is supposed to satisfy.
Free trial, then paid. Everything above is inside The AI Garden Project.
Published 25 July 2026