Sharper, more usable ChatGPT outputs follow when users change how they ask, because Wired compiled 28 practical prompting techniques that show how prompt structure and recent app features affect result quality. The guide frames those techniques around role and persona prompts, few-shot examples, explicit output constraints, iterative refinement, and methods designed to cut hallucinations. Wired also demonstrates how the ChatGPT mobile app's camera and file uploads let users add images to prompts, and it flags which higher-precision tactics are gated behind ChatGPT Plus. That makes many of the recommendations actionable right away for people with the app and for paying subscribers.

The immediate payoff is simpler, more predictable outputs, because Wired groups the tips into repeatable families that steer tone, format and factual rigor.

How the techniques work

Wired organizes the 28 tips into clear behaviors you can repeat. Assigning the model a role or persona, for example "senior software engineer" or "professional copy editor," narrows the model's lexical and logical patterns and raises consistency for technical or editorial tasks. The guide emphasizes few-shot prompting, where you provide two or three worked examples before asking for a new item. Those examples set an explicit template for format and content, which Wired shows reduces the need for follow-up edits.

Other tactics are about structure: explicit output constraints such as fixed length, JSON-only replies, numbered steps or memo headers force the model into a predictable shape. Wired includes copyable templates for tone, length and structure so a user can paste a pattern and get back a uniform result. The piece also favors iterative refinement: instead of restarting a conversation, Wired pairs each tactic with a suggested follow-up phrasing so users can refine outputs in place.

Wired gives specific advice aimed at reducing hallucinations. The guide endorses chain-of-thought style instructions that ask the model to produce intermediate reasoning steps before the final answer, and it recommends prompts that instruct the model to "explain your reasoning" or to "admit when you don't know the answer." Those habits surface uncertainty and cut down on plausible sounding but incorrect claims.

Mastheads

Features and limits in the ChatGPT app

Many of the tips lean on recent OpenAI app features. Wired shows how to use the ChatGPT app's features to add external content to prompts for identification, translation or remediation tasks, and how attaching a file can let the model work from existing content.

It gives short, practical examples such as requesting an "80-20" Pareto-style summary to surface the most important points quickly, and it maps prompt patterns to predictable output formats like JSON, numbered lists or formal memos.

Wired cautions that not every tactic is available to every user. The guide notes where examples rely on ChatGPT Plus features or other paid tiers. That distinction matters because some higher-capacity behaviors and templates are marked as requiring paid access. Wired pairs each tactic with an explicit prompt pattern and a suggested follow-up phrasing so readers can tell whether to refine an existing conversation or start a new one.

The need for careful prompting is shown by third-party reliability figures Wired cites. A Legal Guardian Digital study reported Perplexity AI's hallucination rate at 13 percent versus an industry average of 22 percent. Those gaps are the practical problem the techniques try to shrink: disciplined prompts and explicit reasoning requests do not eliminate errors, but they make mistakes easier to spot and correct.

Wired also pulls from community-tested prompt collections. The piece parallels other prompt guides that stress consistency patterns such as role assignment and few-shot examples, and it prefers micro-patterns that reduce hallucination risk and increase formatting fidelity rather than tricks that merely change tone. That makes many of the tips portable across other large language models while keeping the advice grounded in ChatGPT's app features.

OpenAI's own product changes matter to how quickly users can adopt these practices. Recent product notes and reporting show the company has been rolling out a scheduling hub for premium users. That rollout timing shapes which users can try the scheduling features immediately and which will need to wait or subscribe.

Practically speaking, a reader with the ChatGPT app and access to paid features can try most of Wired's examples today. The step-by-step examples in the guide map prompt patterns to predictable formats and offer follow-ups so readers can refine outputs without starting over. That makes the techniques usable, not just theoretical.

Related Articles

Most of Wired's 28 prompting techniques are usable today in the ChatGPT app; higher-capacity templates are clearly marked as paid-tier features. Originally reported by wired.com.

This article was created with AI assistance.