ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›AI MEANINGS›WHAT IS PROMPT ENGINEERING?
AI MEANINGS · MEANINGBEGINNER

What is prompt engineering? Writing and testing AI instructions

Prompt engineering means writing and testing the instructions you give an AI so it reliably gives what you need. What it means, with a simple example.

READ6 min
WORDS1,110
SECTIONS7
TYPEMEANING
CHECKED1 OCT 26

Prompt engineering means writing and testing the instructions you give an AI so that it reliably produces what you need. It is less about clever wording and more about being clear, giving the right background, and checking the results.

◈ IN PLAIN TERMS

Think of writing a brief for a new builder. A vague note gets you a wall in the wrong place. A clear brief with a sketch, measurements and a photo of what you like gets you the extension you pictured. Prompt engineering is writing that brief, then checking the work and adjusting the brief.

In a sentence: “Before we put the chatbot live, we spent a week on prompt engineering, testing its instructions against fifty real customer questions.”

What the word means

The plain point: the word “engineering” is there because it is meant to be repeatable, not lucky.

OpenAI defines it as writing instructions so that a model “consistently generates content that meets your requirements.”1

A prompt is the message you give the AI. See prompt meaning. Prompt engineering is the skill of getting that message right, and proving it works.

Where the word comes from

Researchers were using the term by 2021. A survey of prompting methods by Pengfei Liu and five colleagues, posted in July 2021, has a section called “Prompt Engineering” that opens: “Prompt engineering is the process of creating a prompting function” that “results in the most effective performance on the downstream task”. In plain words: finding the way of wording the input that gets the best results on a job.2 The sources checked for this page do not say who first used the term.

A simple example

The plain point: it is a loop: write, test, look, adjust.

  1. Write: “Summarise this customer email in one sentence and label it urgent or not urgent.”
  2. Test: run it on twenty real emails, not one.
  3. Look: it marks every complaint as urgent, even mild ones.
  4. Adjust: add what “urgent” means for you, with two examples, and test again.

Anthropic’s guide makes the same point: decide what success looks like and have a way to test it before you start tweaking.3

TAKEAWAY

One good answer proves nothing. A prompt is engineered when it works on many real examples.

Why it matters to you

The plain point: for everyday chats, clear writing is enough. For anything you repeat, the testing part pays off.

A ONE-OFF CHAT · A PROMPT YOU REUSE
Clear writing is enough for the first. The second is where the engineering starts.
A ONE-OFF CHAT
Yes: Plain, clear writing is enough
Yes: No magic phrases needed
A PROMPT YOU REUSE
Yes: Decide what success looks like first
Yes: Test on many real examples, not one
Yes: Adjust, then test again
Yes: Retest when the model changes
Reasoning, 1 Oct 2026 — summarises this page’s example (write, test, look, adjust) and its section on why it matters.
  • Magic phrases are mostly myths. Prompting myths checks the popular tricks against what the AI companies publish.
  • The prompt is not always the fix. Anthropic says some problems are easier to solve by choosing a different model.3
  • Models change, so prompts need retesting. OpenAI recommends tests that keep checking a prompt when you switch model versions.1 How to test whether a prompt change actually helped shows how.

Often confused with

  • Context engineering: building everything the AI can see (files, examples, tools, memory), not just the instruction. This site treats it as the skill that grew out of prompt engineering. See contextual prompting meaning and context engineering.
  • Fine-tuning: changing the model itself by training it on your examples. Much heavier. See fine-tuning meaning.
  • Jailbreaking: using prompts to get round an AI’s safety rules. A different aim altogether: breaking the rules rather than getting good work.

Questions people ask

What do you need before you start prompt engineering?

A clear idea of what a good result looks like. Anthropic’s guide starts with “A clear definition of the success criteria for your use case”.3

How do you test whether a prompt works?

Run it on a fixed set of examples and check the results, again whenever you change the prompt or the model. OpenAI recommends “Building tests and evaluation suites that measure prompt behavior”.1

Can prompt engineering fix every problem?

No. Anthropic says plainly: “Not every success criteria or failing eval is best solved by prompt engineering.” Sometimes a different model, or giving the AI better documents, is the real fix.3

Learn more on this site

QUICK CHECK · 3 QUESTIONS

Pick an answer to see if you got it. Nothing is recorded.

1. What is prompt engineering?

  • Finding secret magic phrases that unlock better answers
  • Building the computer chips that run AI
  • Writing and testing an AI’s instructions so it reliably gives what you need
Show the answer

Answer: Writing and testing an AI’s instructions so it reliably gives what you need. It is less about clever wording and more about being clear, giving background, and checking the results.

2. Why does one good answer prove nothing?

  • A prompt is only engineered when it works on many real examples
  • The AI always gives the same answer anyway
  • Good answers are usually copied from the web
Show the answer

Answer: A prompt is only engineered when it works on many real examples. The page’s loop is write, test on many real examples, look, then adjust and test again.

3. Why should you retest a prompt you reuse?

  • Prompts expire after a month
  • Models change, so a prompt that worked may behave differently
  • Retesting makes the prompt shorter
Show the answer

Answer: Models change, so a prompt that worked may behave differently. OpenAI recommends tests that keep checking a prompt when you switch model versions.

Related terms: Machine translation · Multimodal · Prompt chaining · Reasoning model · Recommendation algorithm · Role prompting

SOURCES
  1. 1OpenAI, Prompt engineering, undated guide, read at source 1 Oct 2026“Prompt engineering is the process of writing effective instructions for a model, such that it consistently generates content that meets your requirements.” ↑a“Building tests and evaluation suites that measure prompt behavior so you can monitor performance as you iterate, or when you change and upgrade model versions”. ↑b“Building tests and evaluation suites that measure prompt behavior so you can monitor performance as you iterate, or when you change and upgrade model versions”. ↑c
  2. 2Liu, Yuan, Fu, Jiang, Hayashi and Neubig, Pre-train, Prompt, and Predict: A Systematic Survey of Prompting Methods in Natural Language Processing, arXiv:2107.13586, 28 Jul 2021 (only version), read at source 13 Oct 2026, section 4“Prompt engineering is the process of creating a prompting function fprompt(x) that results in the most effective performance on the downstream task.” ↑
  3. 3Anthropic, Prompt engineering overview, undated documentation, read at source 1 Oct 2026“A clear definition of the success criteria for your use case”. ↑a“Not every success criteria or failing eval is best solved by prompt engineering.” ↑b“A clear definition of the success criteria for your use case”. ↑c“Not every success criteria or failing eval is best solved by prompt engineering.” ↑d

The comparison on this page (the builder’s brief) is this site’s own way of explaining, not a claim from these sources.

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE