A Checklist for Every Appraisal

Opening Scene

Even a genuinely expert appraiser follows a deliberate, practiced checklist for every single examination — not because they might forget something obvious, but because a consistent, systematic process catches things ad hoc judgment alone sometimes misses. Reducing hallucination benefits from this same kind of practical, systematic checklist, applicable immediately at the prompting level, without needing to fine-tune a model or rebuild infrastructure.

In Plain English

Practical, immediately applicable techniques for reducing hallucination include explicitly instructing a model to say “I don’t know” when genuinely uncertain, asking it to cite specific sources for factual claims, breaking a complex question into smaller, more verifiable steps, and explicitly asking it to flag any part of its answer that isn’t well-supported. These connect directly to the prompt engineering techniques covered in this content library’s dedicated series, applied here specifically to hallucination reduction.

The Old Way

Before these practical, prompting-level techniques were widely known and systematically applied, most users didn’t have an accessible way to reduce hallucination risk themselves:

• Most users had no accessible, systematic way to reduce hallucination risk themselves, relying entirely on whatever mitigation the underlying system happened to provide.
• There wasn’t yet a well-established, shared checklist of practical prompting techniques specifically targeted at hallucination reduction.
• Explicit instructions to express uncertainty or cite sources weren’t yet a widely known, standard practice for everyday model use.

A practical, shared checklist of prompting-level techniques emerged specifically as prompt engineering matured into the rigorous discipline covered throughout this content library’s dedicated series.

What’s Changing (and Why AI Is the Reason)

1. Practical, prompting-level hallucination mitigation techniques are increasingly well documented and widely applied, connecting directly to this content library’s dedicated prompt engineering series.
2. This connects directly to the calibrated uncertainty concepts covered in Article 8, since explicitly prompting for “I don’t know” responses is a direct, practical application of that principle.
3. These techniques are increasingly taught as standard practice for everyday model use, not just specialist knowledge for AI practitioners.

The Metaphor, Fully Extended

The Antiques Appraiser	Practical Hallucination Mitigation Concept
A deliberate, practiced checklist for every examination	A deliberate, practiced checklist for every prompt
Catching things ad hoc judgment alone sometimes misses	Catching hallucination risk ad hoc prompting alone sometimes misses
A systematic process, not relying on memory or instinct alone	A systematic technique set, not relying on hope alone
Immediately applicable without needing new equipment	Immediately applicable without needing to fine-tune or rebuild anything

Quick Recap

• Practical, prompting-level techniques include instructing a model to express uncertainty and cite sources explicitly.
• Breaking complex questions into smaller, verifiable steps is another effective, immediately applicable technique.
• These connect directly to the prompt engineering techniques covered in this content library’s dedicated series.
• These techniques are increasingly taught as standard practice for everyday model use.

Where This Fits in the Series

Article 15 covered practical, immediately applicable techniques. Article 16 turns to the appraiser in the auction house: hallucination risk within multi-step agentic systems.