Skip to content

R0023/2026-03-25/Q002/SRC03/E01

Research R0023 — Counterproductive advice and prompt lifecycle
Run 2026-03-25
Query Q002
Source SRC03
Evidence SRC03-E01
Type Analytical

OpenAI's prompt engineering guide does not identify individual authors; it represents an organizational product with commercial incentives.

URL: https://platform.openai.com/docs/guides/prompt-engineering

Extract

The OpenAI prompt engineering guide is published as organizational documentation without individual author attribution. It provides six strategies for getting better results, including writing clear instructions, providing reference text, splitting complex tasks, giving the model time to "think" (CoT), using external tools, and testing changes systematically. It recommends techniques like persona assignment ("You are a world-class Python developer") that empirical research (Wharton GAIL Report 4) has shown to be counterproductive for factual accuracy.

Relevance to Hypotheses

Hypothesis Relationship Strength
H1 N/A Does not address whether the guides are written by researchers
H2 Supports Vendor guides are not individually authored by researchers — they are organizational products
H3 Supports Vendor guides represent a distinct category: neither researcher-authored nor content-creator-authored, but corporate product documentation

Context

The absence of individual authorship on vendor guides is itself a finding. It makes accountability and credential verification impossible, and it means the guide's content may be shaped by marketing, product, and documentation teams rather than research teams.