Headline Finder

Description

Generates headline proposals from virtual personas built on real European demographic data (Eurobarometer). Provide article text or a URL, set optional demographic filters, and get 15 headline candidates (3 per persona) — each with an editorial explanation grounded in the persona's profile. Output is a structured markdown report with thematic analysis, editorial recommendations, and panel demographics.

Core Capabilities

• Creative direction control: steer tone with a free-text style prompt (e.g., "sensational", "political", "data-driven", "question-based")
• Thematic analysis: identifies recurring editorial patterns across all headlines and maps which personas gravitated toward each
• Editorial recommendations: actionable advice on which angles to lead with, A/B test, or reserve for specific channels
• Bilingual output: English or German, with automatic language detection from source text

Good Query Examples

• Text: [800-word article about EU carbon tax] | Country: Germany (West) | Occupation: Middle Management | Age: 30-50 | Style: "political" → specific demographic filters with clear creative direction
• Text: [full curling scandal article] | Gender: Men Only | Age: 18-45 | Style: "sensational" → gender + age filter with tabloid direction
• URL: https://example.com/tech-review | Country: United Kingdom | Style: "data-driven" → article URL with country filter and style
• URL: https://youtube.com/watch?v=abc123 | Style: "emotional" → YouTube video as content source
• Text: [800 words of article text] | all other fields left at defaults → no filters, no style — uses full panel with maximum demographic diversity
• Text: [article about Austrian rail delays] | Style: "journalistic" → style set, no demographic filters

Bad Query Examples

• Text and URL both empty → no content provided. Fix: paste article text or enter a URL
• Text: [3 short sentences, under 50 characters] → too short for personas to work with. Fix: paste the full article or a substantial excerpt
• Text: [article in French] → output will be in English or German, not French. The agent will warn about the unsupported language and ask whether to proceed. Fix: accept EN/DE output, or provide EN/DE source text
• Style: "for American women in New York" → the country dropdown only has European countries. Fix: pick a European country, or leave it on "All Countries" for a pan-European panel
• User wants headlines split by male vs. female reactions → Gender filter picks one option per run ("All Genders" gives a mixed panel, not separate breakdowns). Fix: run twice — once with "Men Only", once with "Women Only" — and compare reports
• User expects 50+ headlines → fixed at 15 per run (5 personas × 3). Fix: run multiple times with different filters or styles

Use-Case Ideas

• Editorial headline testing — identify which angles resonate across demographic profiles before publication
• A/B test seeding — use thematic analysis to pick 2-3 framing strategies for split testing
• Audience-specific framing — compare how German managers vs. British students would frame the same story
• Creative direction exploration — run the same article with different style prompts to compare tonal shifts
• Content marketing brainstorm — get headline ideas grounded in demographic perspectives rather than internal assumptions
• Localization signal — compare English and German output for the same article to spot framing differences
• Newsletter subject lines — use headline candidates as email subject line options

Limitations

• European personas only — Eurobarometer panel covers 34 countries. No US, Canada, Asia, or other non-European markets
• English and German output only — non-EN/DE source text is accepted but output language will be EN or DE
• Single filter set per run — one country, one gender, one occupation, one age range. To compare segments, run separately
• No psychographic filtering — only demographic filters (country, gender, age, occupation). Use Ask the Crowd for Big 5, IQ, income, or education filters
• Persona drift on longer text — dense source material may cause personas to interpolate beyond the article. Shorter text stays more faithful
• YouTube URLs use metadata only — the API extracts the video title and description, not the spoken content or captions. Videos with link-only descriptions (common) will produce headlines based almost entirely on the title. For richer results, paste article text instead
• Style prompt is suggestive, not deterministic — creative direction influences but does not guarantee tone

Input/Output Spec

Input:

• text (required*): Article text, 50-10,000 characters
• url (required*): URL of an article or YouTube video
• style (optional): Creative direction, free text up to 200 characters
• gender (optional): All Genders (default), Men Only, Women Only
• country (optional): 34 European countries (e.g., Germany West, France, UK, Poland). Default: all
• occupation (optional): 18 Eurobarometer categories (e.g., Student, Middle Management, Skilled Manual Worker, Retired). Default: all
• min_age / max_age (optional): 18-100. Default: full range

*Exactly one of text or url must be provided.

Output:

• Markdown report with:
• Headlines at a Glance: numbered table of all 15 headlines with persona name, age, gender, and occupation
• Headline Explanations: each headline with editorial rationale, grouped by persona
• Thematic Analysis: recurring patterns with headline cross-references
• Editorial Recommendations: actionable suggestions on which angles to prioritize
• Panel Demographics: age spread, gender mix, occupations, living areas

Disambiguation Signals

This agent generates headline candidates from article content. It is content-first and editorial — "What headlines would different European audience segments write for this article?"

• Route to Ask the Crowd when: the user wants opinions, survey data, agreement scores, or A/B testing of existing copy — not headline generation. Also when psychographic filtering is needed
• Route to Let the Crowd Decide when: the user has 3 existing options and wants a winner picked by a virtual panel
• Route to News Research when: the user wants to find and analyze news articles, not generate headlines for existing content
• Route to Advanced Web Research when: the user wants broad topic research or competitive intelligence
• Route to GWI Spark when: the user wants real consumer survey data about audience demographics or media behavior