Opening Question

Before we talk about how to prompt well — let's look at where things are going wrong right now.


When AI-generated content disappoints you, what's usually the problem?

It's too generic — sounds like it could be written for anyone
The tone is wrong — too formal, too casual, or just off
It's missing key details I already gave it
The format doesn't match what I actually needed

Discussion: Every answer on this list points to a gap in the prompt — not a limitation of the AI. Today we learn how to close those gaps before we ever hit send.

The Prompt Framework

Seven elements. Every strong prompt for a creative professional task contains most or all of them. The more you include, the less the AI has to guess.


1
Role — Tell the AI who it is

Before it generates a single word, give the AI a professional identity. This changes its vocabulary, its assumptions, its level of formality, and even the sources it draws on. A news reporter and a brand copywriter will write the same event very differently — so tell it which one to be.

"You are a PR coordinator at a regional university communications office."
2
Context — What's the situation?

Give the AI the background it would need to brief a new hire on this project. What's the organization? What's the campaign or story? What's already happened that this content responds to? The more specific the situation, the more specific the output.

"We are promoting our annual spring speaker series on AI in media, running March through April. We've hosted three events so far; this caption is for the fourth."
3
Goal — What do you actually need?

State your deliverable clearly. Don't say "write something about" — say exactly what you want: a press release, a 90-second podcast intro, three Instagram captions, a homepage hero paragraph. Vague goals produce vague outputs.

"Write a 300-word press release announcing this event to local media."
4
Audience — Who is this for?

Audience definition is one of the highest-leverage elements in any prompt. The same message written for a city council editor and a college freshman will sound completely different. Tell the AI who will read or hear this — their age, role, expertise level, and relationship to your organization.

"Audience: local newspaper editors and TV assignment desks, not the general public."
5
Requirements — What must be included?

Specific facts, people, dates, data points, or structural elements that must appear in the output. If the AI doesn't know a detail, tell it — or instruct it to leave a placeholder so you can fill it in. Never assume the AI will invent accurate specifics.

"Include the event date (October 3, 2026), location (Degenstein Campus Center), and a placeholder for a dean quote marked [DEAN QUOTE TK]."
6
Constraints — What should it avoid?

Word limits, words or phrases to avoid, topics to exclude, things that are off-brand or legally sensitive. Constraints save you from the AI defaulting to its most generic, safest output. If you've ever received a response full of "exciting opportunities" and "thrilled to announce" — a constraint would have fixed that.

"Maximum 150 characters. Do not use the words 'exciting,' 'thrilled,' or 'proud to announce.'"
7
Format — How should it be structured?

Tell the AI exactly what the output should look like: AP style, bullet points, a script with speaker cues, a headline followed by three paragraphs, a table. Without a format spec, the AI picks one for you — and it's usually the wrong one for how you'll actually use the content.

"Format: AP style press release with a headline, dateline, body paragraphs, and a standard university boilerplate at the end."

The "Ask Me Questions First" Technique

Before the AI writes anything, have it ask you 3–5 clarifying questions. This is one of the most powerful moves in professional prompting. It forces the AI to surface gaps in your brief before it starts generating — and it often reveals what you forgot to include.

Add this line at the end of any prompt:

Before you begin, ask me 3 to 5 questions that would help you produce a better result. Wait for my answers before writing anything.

The questions the AI asks will tell you exactly where your brief was incomplete. Answer them, then let it write.

Weak vs. Strong

The same task. The same AI. Two very different prompts — and two very different results. Look for which framework elements are missing in the weak version.


Journalism — Covering a Town Hall Meeting
Weak Prompt
Write a story about the town hall meeting.
Strong Prompt
You are a local news reporter for a small-town Pennsylvania newspaper. Write a 400-word news story about last night's Selinsgrove Borough Council meeting, where residents debated a proposed zoning change that would allow a new apartment complex near the elementary school. Audience: adult residents aged 30–65 who read print and digital news. Tone: neutral and factual, AP style. Include a headline and dateline. Mark any quotes needed from residents as [RESIDENT QUOTE TK] — do not invent quotes.
The weak prompt gives the AI no location, no issue, no audience, no length, and no style. The output will be generic to the point of being unusable. The strong prompt gives everything a reporter's editor would demand before publication.
Public Relations — Campus Event Announcement
Weak Prompt
Write a press release about a campus event.
Strong Prompt
You are the communications director at Susquehanna University. Write a 300-word press release announcing Media Day 2026, an annual showcase where Communication students present capstone projects to local media professionals. Date: October 3, 2026, 10am–2pm, Degenstein Campus Center. Audience: local newspaper editors and TV news directors. Tone: professional, enthusiastic, newsy. Format: AP style with headline, dateline, and a standard university boilerplate. Mark the dean quote as [DEAN QUOTE TK].
Missing from the weak prompt: who's announcing it, what the event is, who it's for, its date and location, the audience for the release, and how to format it. Each missing element is something the AI will invent — or ignore.
Social Media — Instagram Campaign
Weak Prompt
Write Instagram captions for our club.
Strong Prompt
You are the social media manager for the SU Communication Society, a student organization at Susquehanna University. Write 3 Instagram captions promoting our spring speaker series on AI in media. Each caption must be under 150 characters, include one relevant hashtag, and end with a call to action (link in bio). Audience: college students ages 18–22. Tone: energetic and insider — not corporate. Do not use the words "exciting," "thrilled," or "proud to announce."
The weak version will produce three interchangeable captions that sound like every other campus org. The constraints in the strong version — character limit, no filler language, specific CTA — are what make the output actually usable.
Podcasting — Episode Intro Script
Weak Prompt
Write an intro for my podcast episode.
Strong Prompt
You are the host of "Signal & Noise," a podcast about media careers for college students. Write a 90-second spoken intro (approximately 225 words) for Episode 14: "When AI Writes Your Story — And Gets It Wrong." The episode features a guest who is an investigative journalist at a regional newspaper. Open with a hook about a real AI journalism blunder, then introduce the guest, then preview what listeners will learn. Tone: conversational, sharp, slightly skeptical of AI hype. No filler phrases like "Today we're going to talk about."
Word count, show name, episode number, guest description, structural requirements, tone, and forbidden phrases — every one of these eliminates a decision the AI would otherwise make badly on your behalf.

Getting Better Output

The first response is a draft, not a deliverable. Iteration is how professionals use AI — not prompting once and accepting whatever comes back.


1

Refine the Ask

If the output is close but not right, don't rewrite the entire prompt. Identify the specific problem and address only that. "Make the opening more direct." "Cut this to 200 words without losing the key facts." "The tone is too formal — make it sound like a person, not a brochure." Targeted follow-up is faster and more effective than starting over.

2

Add an Example

If the AI keeps missing the style or format you want, show it. Paste in a sentence, a paragraph, or even a full piece and say: "Match this tone" or "Follow this structure." A single good example outperforms a paragraph of description. This technique is called few-shot prompting, and it is one of the most reliable tools you have.

Here is an example of the tone I want: [paste example]. Now rewrite the opening paragraph to match this voice.
3

Change the Format

Sometimes the content is right but the structure makes it hard to use. Ask the AI to reorganize, reformat, or restructure without rewriting the substance: "Give me this as a bulleted list." "Put the most important point first." "Convert this to a script with speaker cues." Format changes cost nothing but a follow-up prompt.


Common Mistakes to Avoid

⚠️

Vague Task

"Write something about our event" forces the AI to invent everything — the angle, the audience, the length, the purpose. Be specific about what you need and why.

⚠️

Missing Audience

Without knowing who will read or hear this, the AI defaults to a generic adult audience. That's rarely who you're actually writing for.

⚠️

No Format Specification

The AI will choose a format — usually the most common one for that type of content. If you need something specific (AP style, a script, three short paragraphs), say so.

⚠️

Skipping Constraints

No word limit, no tone guidelines, no list of things to avoid — and you'll get the AI's defaults. Defaults are what all generated content sounds like before you make it yours.

⚠️

Asking for Too Much at Once

Packing five deliverables into one prompt splits the AI's attention. Build each piece separately, then assemble. One prompt, one focused output.

The Prompt Lab

Pick one real-world task. Write your first instinct prompt. Then rebuild it using the seven-element framework and the "Ask me questions" technique. Submit both versions with a short reflection.


1

Choose Your Task

Select the one you'll work on today. You'll use this same task for both your weak and strong prompt.

2

Your First Prompt

Write the prompt you would have written before today's class. Don't overthink it — just write what you would naturally type. This is your "before."

Now copy this into ChatGPT, Claude, or Gemini and read what comes back. Note where it fell short.

3

Ask Me Questions First

Add the following line to the end of a revised prompt and send it. Read the AI's questions carefully — they'll show you what your brief was missing.

Before you begin, ask me 3 to 5 questions that would help you produce a better result. Wait for my answers before writing anything.

Answer the questions, then let it generate. That answered version becomes the basis of your strong prompt below.

4

Your Revised Prompt

Rewrite your prompt using the seven-element framework: role, context, goal, audience, requirements, constraints, format. Use what the AI's questions revealed.

5

Reflection & Submission

In 2–3 sentences: what specifically changed between your first and second prompt, and how did the AI's output improve as a result?

0 characters

Glossary & Ethics Moment

Key terms from Lesson 2, plus a discussion on who AI writes well for — and who gets left behind.


Prompt Craft Terms
Prompt

The input you give an AI — your question, instruction, or request. The quality of a prompt directly determines the quality of the output. Prompting is a learnable professional skill, not a technical one.

Prompt Engineering

The practice of deliberately structuring your inputs to an AI to get better, more useful outputs. Prompt engineering is not about writing code — it's about communicating with precision. Every element of the seven-part framework is a form of prompt engineering.

Iteration

The process of refining AI output through multiple follow-up prompts rather than accepting the first response. Professional AI workflows almost always involve iteration — treating the first output as a rough draft, then prompting further to improve, correct, or redirect it.

Few-Shot Prompting

A technique where you provide one or more examples of what you want, and the AI uses them as a pattern to follow. Instead of describing a style in words, you show it. "Match the tone of this paragraph" is a form of few-shot prompting — and it is often more effective than any description you could write.

Zero-Shot Prompting

Asking the AI to perform a task with no examples — relying entirely on your written instructions. Most everyday prompting is zero-shot. It works well for clear, well-defined tasks. It struggles when the style, format, or nuance is hard to describe but easy to show.

System Prompt

A set of instructions given to an AI before the conversation begins, establishing its role, behavior, and constraints for the entire session. When a company deploys an AI assistant with a specific personality or set of limits, those are defined in a system prompt. As a user, you're working within whatever system prompt the platform has set — though you can usually override much of it with a strong enough role definition in your own prompt.

Human-in-the-Loop

A workflow design principle where a human reviews, verifies, and makes final decisions on all AI-generated content before it is published or submitted. The AI drafts; the human decides. This is the professional standard — not optional. No framework, no matter how well-constructed, replaces editorial judgment.

Ethics Moment — Who Does AI Write Well For?
Discussion prompt: Use the same strong prompt from the Weak vs. Strong section, but change the audience. Specify "suburban homeowners in their 50s" for one version and "first-generation college students" for another. Run both. Compare the outputs — not for accuracy, but for depth, warmth, and assumption. Who does the AI seem to know better?
Bias in Training Data

LLMs learn from whatever text they were trained on. That text reflects the demographics, assumptions, and blind spots of its authors — which skews heavily toward English-language, Western, educated, and affluent sources. An AI asked to write for a specific community it has little training data on will produce output that feels generic, flattened, or subtly off. Recognizing this isn't a reason to avoid AI — it's a reason to edit critically and know your audience better than the model does.

Default Voice

The style and perspective an AI falls back on when no specific audience, role, or tone is specified. The default voice tends to be generic, formal, American-English, and corporate-adjacent. It is the voice of AI slop. Strong prompting — especially audience definition and constraints — is what moves the output away from the default and toward something that actually serves your readers.

AI Disclosure

A statement attached to any work product created with AI assistance, describing which tools were used and for what purpose. In this course, disclosure is required on all graded work. In professional practice, disclosure standards vary by outlet — but the direction of the industry is toward more transparency, not less. Getting comfortable with disclosure now puts you ahead of colleagues who are still hiding their AI use.

Lesson 2 · Planning and Prompting for Creative Professionals