tomorrow.garden fills in the same fixed record for every plant, and treats a fact as confirmed only when at least two independent sources agree on it. Where sources disagree or there is only one, we mark the value as uncertain and do not filter on it.
We work from horticultural sources, not from forums.
A fact is confirmed when two different sources agree on it. Numbers rarely match exactly: when sources give plausible but different figures we record one value, the larger for sizes and the longer period for flowering.
We mark a value as uncertain when we found a single source, when sources contradict each other, or when the numbers are far apart (more than double for sizes, more than 10 °C for hardiness).
What we could not establish at all is left empty. We never fill a gap with a guess; the page then says "no data".
An uncertain value is not necessarily wrong; it just lacks two agreeing sources. So we handle it with care:
We accept toxicity to dogs and cats only when one of the sources is the ASPCA. The ASPCA does not cover people, so the value for children needs two sources on human toxicity.
We rate a plant "non-toxic" only when a source says so explicitly. A plant that nobody mentions as toxic is not thereby non-toxic. That is why our lists of plants non-toxic to dogs, cats and children are short.
Toxicity data is for guidance only. If your pet or child has eaten a plant, contact a vet or a doctor.
Heat and drought tolerance, native status, invasive status, known problems and planting time are judged for the Carpathian Basin and the neighbouring countries: a continental climate, hot dry summers, winter lows between −15 and −25 °C, and often alkaline soil. So we do not take over a British or North American "drought tolerant" rating unchanged.
Hardiness is the lowest winter temperature the plant reliably survives. We derive it from USDA hardiness zones (zone 5 is −29 °C, zone 6 is −23 °C, zone 7 is −18 °C). When sources differ by one zone we take the warmer one, which is the more cautious value.
We keep two sizes. The ultimate size is that of a mature, unpruned plant. The ten-year size is roughly what the plant reaches in a garden ten years after planting: garden plans are built on it, so it is what the lists and cards show. For hedge plants we also record the height they are normally kept at when clipped.
No source publishes a safe distance from buildings and paving plant by plant, so we work it out with our own rule of thumb for woody plants: half the mature spread plus half a metre from a wall, and one and a half to two times that for plants with more vigorous roots. We give no such distance for herbaceous plants and climbers.
The design picture shows the first, third and fifth summer after planting, which needs a separate estimate. A woody plant starts at the size you buy it in and adds its own yearly growth, so a shrub is still small after three summers. A plant that dies back and regrows every year reaches about half its height in the summer it is planted and all of it by the third. These are estimates for the picture, not promises.
We do not recommend a plant that is on the European Union list of invasive alien species (Regulation 1143/2014) or on the official Hungarian invasive list. If a plant is listed only in a neighbouring Central European country, the plant page says so.
We also do not recommend a plant when a pest or disease common today makes it a poor choice for most gardens. These plants are on the site, but the plant finder and the lists do not offer them.
Some fields are horticultural judgement, not measurable fact: what the plant is used for (hedging, screening, ground cover), which garden style it suits, how far apart to plant it and how many plants a metre of hedge takes, and how much maintenance it needs. We do not require two sources for these.
We write the plant descriptions in our own words and do not copy sentences from sources.
Prices are rounded, indicative values by size, in the local currency, based on the public ranges of web shops in the country of the garden (Hungary, Austria or Slovakia). They are not offers: they show roughly what a plant costs as a small pot and as a larger specimen, and let us estimate the cost of a garden plan.
Only offers you can actually buy count: a sold-out product is left out, unless the plant is out of stock everywhere in the country.
A garden plan is made from the same database, and hard rules filter the plants first. A plant is only considered if its light and soil needs suit the spot, it survives the winter minimum of the garden's region, it copes without watering if you do not water, and it is sold in the garden's country. Next to a dog, cat or child only plants confirmed as non-toxic are used, and next to a child nothing thorny. No hard rule rests on uncertain data: a plant with uncertain hardiness, for example, never goes into a garden plan.
The layout follows professional design principles: German mixed perennial plantings (Staudenmischpflanzung), the methods of Piet Oudolf and of Thomas Rainer and Claudia West, and the advice of the Royal Horticultural Society. A plan usually has a framework (a tree or a structural shrub), larger repeated groups, a continuous low ground cover filling the rest of the space with a few airy plants scattered through it, and spring bulbs underneath. The number of plants follows from their planting distance, the estimated cost from the prices above.
A program checks the finished plan: whether every plant suits the light, the soil and the winter, whether the bed still shows something in winter, and whether something flowers in every month from April to September. If something is missing, the plan page says so.
Artificial intelligence helps find the sources and fill in the records, following the rules above. A program checks the rules on every value: it counts the independent sources, drops numbers outside a plausible range, and marks as uncertain whatever lacks enough sources.
In the plant finder, the language model does not make plants up. Rules pick the matching plants from the database for the conditions given; the model only orders the matches and words the explanation. It cannot recommend a plant that is not in the database.
The plants and the layout of a garden plan also come from rules; a language model only writes the short description of the plan. Where none of our proven combinations suits the spot, a model may choose from plants the rules have already filtered, but its plan is measured by the same checks and is not used if it is worse than the plan made from rules alone.
If you upload a photo, a model reads where there is plantable ground, where a wall or fence stands, and what must be left free (in front of a heat pump, for example). This gives the suggested bed, which you can redraw on the photo.
An image model paints the design picture onto your photo from the scale drawing of the plan, with its plants and numbers. Another model checks that the picture shows the plan and that nothing permanent has disappeared; if it finds a fault, the picture is made again. The picture is an illustration: it shows the garden at its best (lush, mown lawn, without movable objects) and the plants at their estimated size. What is exact is the plant list, the number of plants and the planting layout.
Besides the cookies the site needs to work, we would like to use analytics and advertising cookies (Google, Facebook), if you allow it. Details