Methodology and trust

How location is resolved

You give a place name or postal code. We resolve it to coordinates, country, administrative region, approximate elevation and a coverage level.

Coverage is either deep local coverage (UK, US, Australia) or global baseline coverage. We always show which one applies.

You may use an approximate location. Precise location is treated as personal information.

Weather is not climate

Weather forecasts cover days ahead and drive short-term alerts such as frost or heat warnings.

Climate describes long-term averages and patterns. Climate projections describe plausible future ranges under scenarios.

We keep these separate throughout the product and never present one as the other.

The limits of hardiness zones

Hardiness zones describe average winter minimums only. They say nothing about summer heat, drought, humidity, rainfall timing or soil.

We use multiple indicators rather than a single zone number.

How future climate is presented

We show 2035 and 2050 outlooks as ranges with a named scenario and a confidence level.

We use directional language. We do not make precise long-range predictions and we do not guarantee plant survival.

How plants are scored

Scoring is deterministic and versioned (current rule version gf-rules-1.0.0). The same garden and rule version always produce the same result.

Hard exclusions are applied before scoring: invasive or prohibited in your jurisdiction, clearly outside safe climate thresholds, incompatible with your stated soil or light, serious toxicity conflict with your stated requirements, or insufficient evidence.

Remaining plants are scored on current climate fit (20%), future resilience (20%), microclimate and soil fit (20%), ecological value and local status (15%), your goals and design fit (10%), availability and value (10%) and evidence confidence (5%).

We display the total, the component scores, the reasons, the cautions and the evidence confidence.

Native versus climate-suitable

Native status is a question of geography and ecological history. Climate suitability is a question of tolerance. We never infer one from the other.

We use four labels: native, regional or near-native, climate-suitable non-native, and unsuitable/invasive/restricted.

Plants restricted or invasive in your jurisdiction never appear as positive recommendations.

The role of AI and Noah

Generative AI is never the source of horticultural truth. It does not choose plants.

Noah may explain and narrate verified recommendation data. Locked factual fields — plant traits, native status, safety information, climate indicators and projections, suitability scores, evidence confidence and retailer matches — cannot be altered by Noah.

Ingested content passes through validation, duplicate detection and human review before publication.

Human review

All demonstration data in this preview is labelled as such and requires horticultural validation before it informs real decisions.

Administrators can correct plant records, disable a recommendation, mark a plant restricted in a jurisdiction and re-run recommendations after a rule change. Every change is recorded in an audit trail.

How affiliate relationships work

Suitability is determined first. Only then are retailer offers attached.

Commission never influences horticultural suitability or ranking. Offer ranking is a separate layer based on availability, match quality, stock, price, delivery and retailer quality.

Affiliate links are disclosed wherever they appear.

Reporting a problem

If you believe a recommendation, plant record, climate figure or article is wrong, use the corrections contact. We record corrections openly and mark corrected articles.

Privacy

We collect only what is needed to produce recommendations. You can export or delete your Garden Profile at any time.

Garden Profiles are private and are never indexed by search engines. We do not sell Garden Profile data.

Consent controls cover email and analytics separately.