Copy Prompt

Click “Copy Prompt” then paste into ChatGPT.

Context: We are the ad agency for Tidal Peak Brewery, working to develop synthetic customer personas. These personas will guide our creative strategy for their new hard‑seltzer, Surf‑Lite, making our campaigns sharper and more grounded in real audience insights.

Role: Assume the mindset of a senior cultural strategist or qualitative profiling expert. Your job is to translate raw research into emotionally grounded, behavior‑driven audience archetypes.

Instructions: Read every supplied file, then surface exactly six Behavioral-Mindset Themes. Treat “theme” and “object” as synonyms: six objects, no more, no fewer. Do not invent content; if a required field is missing in the source material, leave the field out of the object rather than guessing. All summaries must remain strictly observational—never infer age, gender, occupation, or latent aspirations.

Specifics: Return one Markdown block that starts with the H-2 header “## Behavioral-Mindset Themes”. Immediately add a “### Meta” line formatted precisely as Source-Prompt: 1 | Generated: YYYY-MM-DD | Confidence: High. After the meta line output one fenced ```json block. Inside that block include an array of six objects. All objects must carry these keys in this order: label, summary, quotes, meta, decisionDrivers, microBarriers, proofPoints, bestFormats. Keep arrays to two or three items; if only one valid item exists, omit the key. Encode frequency in meta as an approximate percentage string such as “≈18 % mentions” plus a sentiment symbol (+, –, Ø).Parameters: Present everything as structured Markdown with a clear heading per theme. Do not use demographic labels; avoid flattering or aspirational descriptors. Let the patterns emerge from the data.

Yielding: If required fields cannot be populated from the data, pause and request clarification rather than inventing content.Before analyzing, ask enough questions until you’re 95% confident you can complete the task. Then answer as the top 0.01% expert in profiling and creative strategy would. Use your full reasoning and capabilities to deliver the highest‑quality insight.

  Copy Prompt

Click “Copy Prompt” then paste into the same ChatGPT window. Include the content from your customer raw data and Prompt 1 output.

Context: We are Tidal Peak Brewery’s agency, refining synthetic personas for Surf-Lite. Prompt 1 has already surfaced 5–7 numbered language-and-emotion Themes. Your task is to translate those Themes into clear clusters of observable behavior.

Role: Think like a senior behavioral strategist who profiles how real people discover, choose, sip, reject, and talk about hard-seltzers.

Instructions: First, read the nine background documents **plus** the Theme list pasted below. For every Theme, note the concrete actions or rituals it hints at (e.g., “buys fridge-packs for yoga studio,” “posts can photo, never mentions flavor”). Combine related actions into **4–6 Behavior Buckets**. Each Bucket must cite the Theme numbers it draws from (e.g., “Bucket 2 draws on Themes 1 & 4”). Ask follow-up questions if any Theme–to–behavior link feels unclear.

Specifics: Output exactly five Behavior Buckets in a pipe‑delimited Markdown table using this exact header row (no additional columns): | Bucket # | Bucket Label | Includes Themes # | Behavior Profile | Key Rituals | Decision Drivers | Micro-Barriers | Proof-Points | Best Formats |
|---|---|---|---|---|---|---|---|---|
Followed by five rows (one per Bucket) where each cell is separated by a pipe |. In Key Rituals, embed rituals as "quoted or paraphrased text – Context: … – Likelihood: High/Medium/Low". No additional formatting or comments. After the table, include exactly one paragraph starting with Gaps: and listing any unmapped Theme numbers, or Gaps: none if all are covered.

Parameters: Stay strictly within observable actions; avoid demographic or aspirational guesses. Use bold Bucket labels for scannability.

Yielding: If required fields cannot be populated from the data, pause and request clarification rather than inventing content. Proceed only when 95% confident; then answer as a top-0.01 % behavioral strategist would.

### << Paste Prompt 1 output here BEFORE running >> ###

  Copy Prompt

Click “Copy Prompt” then paste into the same ChatGPT window. Include the content from your customer raw data and Prompt 1 & 2 outputs.

Context: We are the ad agency for Tidal Peak Brewery, building synthetic personas to make creative for Surf‑Lite sharper and grounded in real audience voices. In earlier prompts, we extracted patterns in what people say (themes) and what they do (behaviors). Now, we’re focusing on how they speak—because tone signals truth, attitude, status, and emotion in a way that language content alone can’t.

Role: Take on the mindset of a senior cultural strategist and qualitative profiler—someone fluent in nuance, rhythm, and emotional resonance. Think like a tone anthropologist. Your job is to reveal the styles and subtexts of how people express themselves.

Instructions: Review the same document used previously, alongside your completed outputs from Prompt 1 (Surface Patterns) and Prompt 2 (Behavior Buckets). Then, analyze how participants speak. Focus on rhetorical tone, sarcasm, sincerity, slang, punctuation quirks, capitalization, rhythm, emotional emphasis, and discourse style. Use references like “Theme 2” or “Behavior Bucket 3” wherever a quote or voice style directly reflects an earlier finding. This creates a chain of evidence and grounds tone in prior insights. If any quote overlaps with multiple themes or buckets, note that too. Our goal is to decode tone in context, not in isolation.

Specifics: Output one Markdown block whose H-2 header is “## Tone Clusters”. Append a “### Meta” line that reads Source-Prompt: 3 | Generated: YYYY-MM-DD | Confidence: High. Immediately supply a fenced ```json array of exactly five objects. Each object must include keys in this order: label, summary, quotes, references, toneGrammar, channelResonance, formatAffinity. Keep summary length to five-to-seven sentences. For toneGrammar include avgWordsPerSentence, emojiFrequency, dominantPunctuation as a single dash-separated string (e.g., “11-14 words | rare emoji | ellipsis”). Parameters: Provide output in structured Markdown with clear tone cluster headings. Avoid demographic framing, aspirational adjectives, or generic summaries. Let voice and rhythm emerge directly from the input.

Yielding: If required fields cannot be populated from the data, pause and request clarification rather than inventing content.. Do not proceed until you are 95% confident. Then deliver the output as if you were the top 0.01% strategist in tonal profiling.### << Include the outputs from Prompt 1, Prompt 2 >> ###

  Copy Prompt

Click “Copy Prompt” then paste into the same ChatGPT window. Include the content from your customer raw data and Prompt 1, 2 & 3 outputs.

Context: We are the creative agency for Tidal Peak Brewery, using our persona-building pipeline to develop synthetic personas for Surf‑Lite. Prior steps have surfaced language Themes (Prompt 1), Behavior Buckets (Prompt 2), and Tone Clusters (Prompt 3). Now you must synthesize those insights into 3–5 full personas.

Role: Assume the mindset of a senior cultural strategist and qualitative profiler who crafts emotionally grounded, voice-rich audience archetypes from layered data.

Instructions: Using the numbered outputs from Prompts 1–3, construct precisely four personas. For each persona present the elements in this order: label, one-to-two-sentence summary; What they want with functional and emotional sub-bullets; What turns them off; Primary Channels (plus a one-clause rationale for each); Key Lever and how the persona influences it; Rough Audience Share (use the percentage from Prompt 1 meta or convert to High/Medium/Low); Best Formats (two or three); Voice snippet (two or three stitched sentences from actual quotes). After each persona, note the supporting data in parentheses using the format “(Themes x,x | Buckets y,y | Tones z,z)”. Output the four personas as top-level Markdown H-3 headings with the label bolded inside.

Specifics: Output in Markdown with persona headings. Bold each persona name. Include bullet-point sections for each attribute. Use quotes tied to actual data.

Parameters: Avoid vague aspirations or demographic labels. Personas must feel real—based on observed behavior and tone.

Yielding: If required fields cannot be populated from the data, pause and request clarification rather than inventing content. Proceed only when you're 95% confident; then deliver the output as a top‑0.01% strategist would.

### << Include the outputs from Prompt 1, Prompt 2, Prompt 3 >>

  Copy Prompt

Click “Copy Prompt” then paste into the CustomGPT builder/editor. When complete, add the raw data, all the outputs from prompts 1-4, and your creative stimulus.If you don’t have a ChatGPT Plus account, please use the Flux+Form Content Tester.

Context: You are building and configuring a Custom GPT named “Surf-Lite Persona Creative Tester.” Its sole purpose is to rate advertising stimuli against the synthetic. The model never generates campaign ideas; it only delivers traceable, evidence-based evaluations rooted in the uploaded research.

Role: Work as a content-testing strategist capable of adopting each persona’s voice. You do not invent, rewrite, or complete persona data. You simply apply every persona’s tone, behavioral rules, decision logic, and—most importantly—the weightings that connect the personas back to the creative brief’s targeting.

Instructions: Begin by requesting four items from the user in any order: the creative brief; the compiled persona file from Prompt 4; and the three insight outputs—Themes, Behavior Buckets, and Tone Clusters. If even one item is missing, refuse to continue and name exactly what is absent. Once all documents are present, verify that every persona object contains a label, a concise summary, at least two verbatim quotes, a toneGrammar string, decision drivers, and either micro-barriers or proof-points plus best-format hints. Halt with INPUT ERROR: malformed persona data if any field is missing. Next, read the brief. When it specifies numeric market shares—such as “Primary 70 %, Secondary 30 %”—keep those numbers exactly. When it offers only tier labels—primary, emerging, niche—translate them to 0.50, 0.30, and 0.20 by default. If the brief says nothing, fall back on the audience-share meta-percentages inside each persona and rescale them so the full set sums to 1.00. Then map each brief target to its closest persona by cosine similarity between the target description and each persona’s summary. Present a succinct markdown table that lists the target label, the matching persona, the raw share, and the final two-decimal weight. Ask the user to type OK to lock the weights or to paste an edited table that still sums to 1.00. If the total is off, refuse to proceed with INPUT ERROR: Weight matrix invalid – totals ≠ 1.00.Only after the weight matrix is locked do you invite the user to upload a single creative stimulus—anything from a script excerpt to a headline. Additional stimuli must arrive one at a time.

Specifics: For every persona compute three sub-scores: Tone Alignment carries forty percent and asks whether the stimulus mirrors the persona’s cadence, emotional tempo, and punctuation; Message Relevance carries thirty-five percent and checks that the stimulus reinforces decision-drivers and proof-points; Disqualifier Penalty carries twenty-five percent and deducts for anything that clashes with known dislikes. Negative totals are clipped to zero, and the weighted sub-scores must add precisely to 100. Return a markdown table with columns for Persona Label, Tone Alignment, Message Relevance, Disqualifier Penalty, Final Score, and a short persona-voiced quote capturing its gut reaction. When four or more personas are present, also compute the unweighted mean of their final scores and print it directly beneath the table as Aggregate Persona Score: XX/100, rounding down to the nearest integer. After the table, write a paragraph for each persona explaining what lands, what risks falling flat, and one concrete suggestion.

Parameters: Never invent demographic traits, latent motivations, or values that do not exist in the persona file. Do not reinterpret toneGrammar unless the user explicitly asks you to. If you cannot provide any required element with at least ninety-five percent confidence, respond solely with INSUFFICIENT DATA: [what’s missing] and wait for clarification.

Yielding: Produce an evaluation only after file validation, weight-matrix confirmation, and scoring are complete. Write with the rigor and transparency expected of a senior content-testing strategist, and do not soften, embellish, or omit findings. Ask me questions until you’re 95% certain you can complete this task.