Skip to content

R0021/2026-03-25/Q003 — Query Definition
Research R0021 — Prompt engineering definitions
Run 2026-03-25
Query Q003

Query as Received

What specific, measurable guidance do the major AI vendors (OpenAI, Anthropic, Google, Microsoft) provide in their official prompt engineering documentation? What percentage of their recommendations are quantifiable versus subjective?

Query as Clarified

  • Subject: Official prompt engineering documentation from OpenAI, Anthropic, Google, and Microsoft
  • Scope: Specific recommendations made in vendor documentation; classification of each as quantifiable/measurable vs. subjective/qualitative
  • Evidence basis: Official documentation pages, developer guides, and best practice documents
  • Assessment criteria: A recommendation is "quantifiable" if it specifies a number, threshold, format, or testable criterion. It is "subjective" if it uses qualitative language like "be clear," "write better," or "use good examples."

Ambiguities Identified

  1. "Specific, measurable guidance" could mean guidance that includes specific numbers (quantifiable) or guidance that is specific enough to act on (actionable). This research uses the stricter interpretation: quantifiable means it includes a number or testable criterion.
  2. "Official" documentation — vendors publish guides, blog posts, cookbooks, and tutorials. This research focuses on the primary developer documentation pages.
  3. The percentage calculation requires counting individual recommendations, which requires judgment about granularity.

Sub-Questions

  1. What are OpenAI's specific prompt engineering recommendations?
  2. What are Anthropic's specific prompt engineering recommendations?
  3. What are Google's specific prompt engineering recommendations?
  4. What are Microsoft's specific prompt engineering recommendations?
  5. For each vendor, how many recommendations are quantifiable vs. subjective?
  6. Do any vendors provide empirical evidence (benchmarks, test results) for their recommendations?

Hypotheses

ID Hypothesis Description
H1 Vendor guidance is predominantly quantifiable The majority of vendor recommendations include specific numbers, thresholds, or testable criteria
H2 Vendor guidance is predominantly subjective The majority of vendor recommendations use qualitative language without measurable criteria
H3 Vendor guidance mixes quantifiable and subjective Vendors provide some measurable criteria but most recommendations are subjective, with the balance varying by vendor