Edition 04 · Apr 2026 Filed under — vs ChatGPT Feasibly · The Brief

vs ChatGPT

Keep using ChatGPT. Add Feasibly.

ChatGPT is great at structure, explanation, and drafting, and bad at Australian planning data. It hallucinates zone codes, invents schedules, and can't cite sources. Feasibly plugs in via MCP, so your LLM stops guessing and starts citing.

FIG. 01 — SIGNAL VS SMUDGE

Head to head

Head-to-head on a real task

“Check the zoning and overlays for 12 Flinders Lane, Melbourne VIC 3000 and tell me if I can build a second dwelling.”

ChatGPT / Claude alone With Feasibly
Zone code Often right for famous addresses, often wrong for everyday ones. Plausible but not cited. Returned from live Vicmap WFS with source link. Point-in-polygon verified.
Schedule values (minLotSize, height, garden area) Returns defaults from training data. May quote a value that was changed in VC267, VC270, VC282, VC304. 620+ council schedule interpretations, each with last-verified date + schedule source URL. Daily amendment monitor.
Overlays May miss. May invent. No spatial check against the parcel. Spatial intersect against the parcel polygon. Only overlays that actually touch the lot.
Heritage Based on what's in training data. Misses HO additions since 2024. Live VHR + VHI + overlay registers. 1,700+ QLD State Heritage places mapped.
Source citations None, or fabricated URLs that 404. Every field links to the exact planning scheme page or Vicmap layer.
Confidence tag Always confident. Including when wrong. Per-field verified / default / unverified, with freshness date.
If wrong Shrug. Wrong-zone guarantee: free lookup + 10 credits, dataset gets corrected.

Decision

When to use each

Use ChatGPT / Claude alone for...

  • Explaining a planning concept in plain English
  • Drafting an email to council
  • Structuring a client brief
  • Anything that doesn't require a specific live fact about a specific property

Add Feasibly for...

  • Any zone / overlay / schedule question about a specific address
  • Yield and envelope estimates
  • Anything that needs to be source-linked to defend later
  • Batch screening of a pipeline

Closing

How to add it (2 min)

  1. Sign up free · 20 lookups/month, no credit card
  2. Copy your MCP URL from the dashboard
  3. Paste into Claude Desktop, ChatGPT Developer Mode, or Cursor
  4. Your LLM now cites instead of guesses