Lay out the sales-comparison grid. From YOUR verified comps, not invented ones
Rung 8 of 10 · Fact-check list
1 · Learn the move · Fact-check list
The sales-comparison grid is where fabrication is most tempting and most dangerous: a plausible-looking +8% condition adjustment the model invented reads exactly like one you supported. So the discipline is a fact-check list. Every comp and every adjustment cell traces to YOUR verified data or your own market support, and a comp missing a data point gets an UNKNOWN cell, not a guess. The AI structures the grid; it does not invent a comparable, an adjustment direction, or an adjustment magnitude you didn't provide. And the reconciled value indication? That's computed in a tool and concluded by the appraiser. Tagged [verify], never asserted by the AI. You get a clean, auditable grid that says exactly where each number came from and leaves the value where it belongs: with the appraiser.
Here are 3 comps and my subject. Build the adjustment grid and tell me what the subject is worth. [paste]
2 · Your turn. You write the prompt
You're valuing a 9,200 SF retail strip. You've verified 3 comparable sales and pasted them: Comp A $245/SF, sold 2026-02, 8,800 SF, 'similar condition'; Comp B $228/SF, sold 2025-11, 10,400 SF, 'inferior frontage per my inspection'; Comp C $261/SF, sold 2026-04, 7,900 SF, 'superior. Recent facade renovation'. You have market support for a size adjustment but NOT yet for condition or frontage. Write a prompt that organizes the grid, shows adjustments only where you supplied support, marks the rest for your input, flags missing comp data UNKNOWN, and leaves the value indication to the appraiser.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.