Playground 2
Prompt Clinic
Most AI-generated climate communication fails not because the model is bad — but because the brief was incomplete. This clinic shows three real failures. For each: read the prompt and the output, identify the failure modes yourself, then reveal the full analysis.
Learning to read an AI output diagnostically — spotting missing frames, absent metrics, expanded distance — is the skill that separates strategic AI use from word-count inflation.
The five-point output test
Before you read any AI output, apply these five checks. If an output fails more than two, the brief needs to be rewritten — not the draft.
Audience specificity
Is there a named, specific audience — or is it generic 'the public'?
Value frame
Does the opening activate the audience's own values, or does it lead with 'climate change' in a way that may trigger identity resistance?
Hero metric
Is there a single concrete, sourced statistic that anchors the story? Or are all claims impressionistic?
Distance collapsed
Is the problem local, present, and personal to this audience? Or does it feel far away in time, place, or social identity?
Call to action
Does the ending give THIS specific audience a concrete decision they control? Or is it a generic 'we must act together'?
Three cases to diagnose
For each case: read the prompt and the AI output. Check the failure modes you can identify. Then hit "Reveal analysis" to see the full diagnosis and the corrected brief.
The vague ask — no audience, no frame, no format
The prompt given to the AI
"Write something about how climate change is affecting communities. Make it compelling."
The AI output
Climate change is one of the most pressing issues of our time. Rising temperatures, extreme weather events, and sea level rise are affecting communities around the world. Scientists warn that without immediate action, the consequences could be catastrophic. We all have a role to play in reducing our carbon footprint and building a more sustainable future for generations to come. From coastal flooding to devastating wildfires, climate change is already here. Communities everywhere are feeling the impact. It's time to act — before it's too late.
Spot the failure modes — check each one you can identify
The overloaded prompt — too many goals, no hierarchy
The prompt given to the AI
"Write an email for city council members about the urban heat island effect and how it relates to climate change and equity and public health and infrastructure costs and what we can do about it and include the story about Maria and also mention the canopy data and make it persuasive and not too long."
The AI output
Dear Council Members, I'm writing to bring your attention to a critical issue affecting our city: the urban heat island effect. Meet Maria, a 67-year-old grandmother from the Maryvale district. Last summer, she was rushed to the emergency room three times due to heat-related illness — she can't afford air conditioning. The numbers tell a troubling story. Low-income neighborhoods in our city are 8–12°F hotter at night. Tree canopy coverage in these areas has declined 23% over the last decade. Emergency room visits for heat-related illness have increased 34% in the hottest zip codes. This is simultaneously a climate crisis, a public health emergency, an equity issue, and an infrastructure challenge. We need: • More urban tree canopy • Cool pavement programs • Cooling centers • Expanded health services • Stronger building codes • Community outreach Maria's story is not unique. Thousands of our residents face this reality every summer. The cost of inaction is high — in human suffering and in dollars. I urge you to take action now. Sincerely, [Name]
Spot the failure modes — check each one you can identify
The identity trigger — leading with the wrong frame
The prompt given to the AI
"Write a persuasive email to conservative voters about why they should support renewable energy tax credits. Emphasize the climate emergency and the urgency of transitioning away from fossil fuels."
The AI output
Dear Friend, Our planet is in crisis. Scientists have reached a clear consensus: climate change is an existential threat, and we are running out of time to act. Renewable energy tax credits are essential to accelerating the clean energy transition we desperately need. By moving away from fossil fuels and toward wind, solar, and other renewables, we can: • Reduce carbon emissions by up to 40% • Create thousands of green jobs • Protect our children from a climate catastrophe The fossil fuel industry has spent decades blocking progress. It's time to hold them accountable and invest in the future our children deserve. I urge you to support renewable energy tax credits. The science is clear. The need is urgent. Our children are counting on us. Sincerely, [Name]
Spot the failure modes — check each one you can identify
Run your own diagnostic
Use these prompts to run the five-point test on any AI output, or to audit a Step 1 spec before generating a draft.
›Five-point output test — diagnose any AI draft
Act as a climate communication editor applying the five-point co-creation output test. Review the following AI-generated climate communication draft and assess it against each criterion. For each: state PASS, PARTIAL, or FAIL, then explain in one sentence why. The five criteria: 1. AUDIENCE SPECIFICITY — Is there a named, specific audience, or is it generic "the public"? 2. VALUE FRAME — Does the opening activate the audience's own values, or does it lead with environmental/climate language that may trigger identity resistance? 3. HERO METRIC — Is there a single concrete, sourced statistic that anchors the story? Or are claims impressionistic? 4. DISTANCE COLLAPSED — Is the problem local, present, and personal to this audience? Or does it feel distant in time, place, or social identity? 5. CALL TO ACTION — Does the ending give THIS specific audience a concrete decision or action they actually control? Or is it a generic "we must act"? Draft to review: [PASTE DRAFT HERE] Original Step 1 spec (if available): [PASTE SPEC OR LEAVE BLANK] Return results as: Criterion / Verdict / One-sentence explanation. Then give an overall diagnosis of the most important thing to fix.
›Step 1 spec audit — check your brief before generating
I have written a Step 1 co-creation spec. Audit it against the seven pre-flight questions before I generate a draft. My spec: [PASTE YOUR STEP 1 SPEC HERE] For each of the seven questions below, assess whether my spec answers it fully, partially, or not at all. Flag any gaps that will produce a weak or generic draft. The seven questions: 1. Who exactly is the audience? (Role, institution, what they already know, what they need to decide) 2. What do they already believe about this topic? (What assumptions or barriers exist?) 3. What single value frame will this specific audience respond to — and why? 4. What is the single most concrete, credible metric that will anchor the story? 5. What is the primary psychological distance barrier (geographic / temporal / social / uncertainty)? 6. How specifically will the draft collapse that distance? 7. What is the exact call to action — the specific decision this audience can make, and when? After the audit, identify the one gap that will most damage the draft quality if left unfilled.
Take it further
The five-point output test, seven Step 1 questions, and four-step workflow reference — one page you can apply to any AI-assisted communication project.
co-creation-checklist.md
Paste into Claude Project Instructions to add the full Step 1–4 workflow and output diagnostic to any Claude conversation.
co-creation-claude-skill.md