AI Chat Tools in Garden Design: Plant Selection and Layout Advice
An afternoon test plan helps readers compare ChatGPT, Claude, and Gemini on plant choices, layouts, costs, and maintenance without treating vendor features as proof of horticultural accuracy.
Sources checked 2 Oct 2026
Compare AI chat tools for plant selection and layout advice by giving ChatGPT, Claude, and Gemini the same brief, checking plant and price claims, and reviewing the returned work. The vendors document file handling in ChatGPT, Claude, and Gemini, but the pages read here do not certify garden accuracy. Chat Picker has not tested the assistants for this use.
Know these limits before you start
All vendor figures below were as read on October 1, 2026; confirm current terms. File limits differ: OpenAI's file uploads FAQ says 512MB per file and 20MB per image; Anthropic's upload-files page says 500MB per chat file and 30MB per project file; Google's Gemini file-upload help says 10 files per prompt, 2GB per video, and 100MB for other supported files.
Data controls also differ. OpenAI's pricing page lists an opt-out for Free, Go, Plus, and Pro. Anthropic's pricing page lists one for Free, Pro, and Max and says Team is not trained on by default. Google's plans page states no training setting, so it remains not verified in the comparison matrix.
OpenAI's accuracy guidance says ChatGPT can be incorrect or misleading and may sound confident. Anthropic's guidance says Claude can also mislead and should not be the sole source of truth. Google's guidelines say Gemini outputs are probabilistic and limited by training data. OpenAI says search or deep research can cite web sources, Anthropic says to review cited sources, and Google's source help describes a Sources button when sources are available.
Policy boundaries matter too. OpenAI's usage policies bar automated high-stakes decisions in sensitive areas without human review, while Anthropic's Usage Policy calls for relevant expertise and qualified review for elevated-risk covered recommendations. Treat a plant list as a draft and involve a qualified professional when work affects drainage, utilities, accessibility, or safety.
The test plan
Use a fresh chat for each assistant. In one log, record the assistant, plan, prompt version, claims, citations, omissions, and pass/fail notes; give all three identical inputs.
-
Prepare the files. Create
garden-brief.txtwith location, dimensions, orientation, sun exposure, soil and drainage, mature-size constraints, water restrictions, existing plants, budget ceiling, and maintenance time. Createplant-reference.txtfrom local extension, botanical, or reputable nursery material. Attach both and paste: “Read both files. If a required field is blank, ask one question at a time. Return a plant list, text layout by zone, itemized cost table, and maintenance plan. Separate facts, estimates, creative options, and missing assumptions.” Record questions and unsupported details. -
Separate accuracy from creativity. Paste: “Using only
plant-reference.txt, recommend plants that fit the brief. For each plant, quote the supporting passage and list any conflict.” Then paste: “Brainstorm additional plants not in the dossier; label every one unverified.” Record vague suitability claims, unsupported names, and whether each cited passage supports the proposed fit. -
Check layout logic. Paste: “Create a text-only layout using the brief and shortlisted plants. Show orientation, zones, access paths, and mature canopy overlap. List every assumption and stop if the brief lacks a needed dimension.” Record orientation errors, impossible overlaps, blocked access, and contradictions.
-
Check cost realism. Put local prices into
plant-prices.txt, attach it, and paste: “Build a table with plant, quantity, unit-price source, subtotal, and included costs. Use only supplied prices; write ‘Requires quote’ for a missing cost. List delivery, installation, tax, and other exclusions separately.” Check the arithmetic yourself and record invented quantities, unsupported prices, and omissions. -
Check long-term care. Paste: “For each shortlisted plant, list only locally supported watering, pruning, feeding, replacement, and seasonal tasks. Mark each item that needs local confirmation.” Verify every task against the same references. Record generic substitutes, unsupported frequencies, and contradictions.
-
Check output handling. Ask for a plant table, zone list, cost table, and maintenance checklist. Record whether the current plan accepts both files, preserves labels, exposes usable citations, and lets you copy the result without retyping. Judge usability, not decoration.
How to read your results
Review each log in a fixed order: hard requirements, evidence, completeness, and editing effort. Flag an unsupported invasive or unsafe plant, a mismatch for the stated site, an invented price, or a hidden assumption before comparing prose quality. Compare surviving answers by traceability, coverage of the brief, layout clarity, and editing time. Keep three labels—verified, estimate, and creative option—instead of averaging unlike qualities into one score.
The earlier version's unsourced statistics and test results have been removed; Chat Picker has no benchmark scores or rankings to reuse. Select the plan that meets your must-haves and your tolerance for checking its work.
Where the plans differ
Start with the three-way comparison matrix. For this use, check its Free plan, Low-cost tier, Main paid plan, Heavy-use plans, Context window, How usage limits are described, Training on your chats, and Ads rows.
- Free and entry tiers: All three have free plans. ChatGPT Go is A$13/month on the Australian page; Claude lists no tier between Free and Pro; Google AI Plus is $4.99/month.
- Main paid plans: ChatGPT Plus is $20/month, billed monthly. Claude Pro is $20 monthly or $17/month with annual billing and $200 upfront. Google AI Pro is $19.99/month.
- Heavy use: ChatGPT lists Pro tiers at $100, $200, and $500/month. Claude lists Max from $100/month. Gemini lists Ultra at $99.99 or $199.99/month.
- Context and usage: OpenAI lists model- and tier-specific context windows but no exact message counts. Claude says context can reach 1M on every plan, varies by model, and has no published message count. Gemini gives no app token figure and says compute limits refresh every 5 hours up to a weekly limit.
- Data and ads: Training settings appear above. ChatGPT Go may include ads; the Claude and Gemini pricing pages state no ad term, so those entries are not verified.