Source-Based Quote Extractor
What It Does
This Agent scans the readable body of an article and extracts only direct quotes enclosed in quotation marks. Each quote is paired with a Speaker value, based on the article’s attribution. If no speaker is provided, it defaults to Speaker unknown.
If no quotes exist, the Agent will simply return:
No direct quotes were found in this article.
Why It Matters
Accurate quoting is critical for journalism, research, and editorial workflows. Misattributed or paraphrased statements can damage credibility. This Agent ensures you always get verified, source-true quotes in a structured format — with zero guesswork or noise from site navigation, ads, or commentary.
When to Use It
Perfect for:
- Extracting quotes for news articles, newsletters, or analysis pieces
- Building source-verified quote archives or databases
- Supporting fact-checking and editorial review
- Pulling statements from press releases, interviews, or speeches
- Identifying exact words from politicians, executives, or cultural figures
When Not to Use It
Avoid using this Agent:
- For paraphrased insights or summaries (use Summarizer instead)
- For sentiment analysis or topic tagging (use Metadata Tagger instead)
- On content that isn’t article-based (e.g., PDFs, broken feeds, or placeholder text)
Inputs and Configuration
Basic Configuration
Field | Required? | Description |
---|---|---|
Source Content | ✅ | Raw readable text of the article (fetched beforehand). Must include body copy, not navigation or metadata. |
Advanced Configuration
Field | Required? | Description |
---|---|---|
Language Model | ✅ | Default = GPT-4.1 |
Brand Voice / Guidelines | Not used | This Agent ignores brand fields — it always outputs direct quotes in neutral structure. |
Brand, Tone, and Rules
- Quotes must be extracted exactly as written — no paraphrasing.
- Only extract text within quotation marks (“ ” or " ").
- Always attach a Speaker field:
- If name provided → use as-is
- If only a role → use the role
- If no attribution → “Speaker unknown”
Inclusion Rules
✅ Include:
- “We are proud to launch this initiative,” said CEO Jane Smith.
- “I didn’t vote for it,” the senator said.
❌ Exclude:
- Paraphrased summaries
- Headlines, captions, or metadata
- Social embeds or comments
Operational Details
Pre-Configured Defaults
- Model: GPT-4.1
- Speaker field always included (even if unknown)
- Filters out menus, ads, footers, and boilerplate text
- Skips malformed or blank content with a clear error
Safety & Fallbacks
If content is missing, empty, or unreadable:Error: Unable to process this URL. The content may be blank, inaccessible, or non-article text.
If no quotes found:No direct quotes were found in this article.
Never invents or rewrites names, roles, or quotes.
Output Format
Each quote is listed as:
Quote:
“Exact text of the quote.”
Speaker: Jane Doe
If multiple quotes: repeat this structure in sequence.
If none: return the no-quotes message.
Smart Behaviors & Validation
- Strips non-article elements (ads, nav bars, comments)
- Preserves punctuation and casing of quotes
- Always returns a usable, clean structure
Guidance and Integration
Tips & Tricks
- Use with Summarizer Agent for quick quote + recap packages.
- Combine with Metadata Tagger Agent to add context (who, where, what).
- Perfect companion for Sales Follow-Up Sequence or Article to Email Agent, where verified quotes can increase credibility.
- Great for internal editorial workflows that need clean pull quotes.
Example Use Case
Input:
Article about a new tech launch with three attributed quotes.
Output:
Quote:
“We are thrilled to bring this to market.”
Speaker: Jane Smith, CEO
Quote:
“The data shows clear benefits for customers.”
Speaker: CTO John Doe
Quote:
“We will continue investing in this area.”
Speaker: Company spokesperson
Troubleshooting
- No quotes returned? → Check if the article actually contains quotation marks.
- Speaker missing? → Defaults safely to “Speaker unknown.”
- Error message shown? → Source input may be paywalled, blank, or non-article text.
FAQs
Q: Does it ever paraphrase quotes?
No — it only returns verbatim text within quotation marks.
Q: Can it handle transcripts?
Yes — if formatted with quotation marks and speaker attribution.
Q: What happens if there are dozens of quotes?
All are extracted, with each listed in the same Quote + Speaker structure.
Q: Can it misattribute speakers?
No — if a name is not clearly tied, it defaults to “Speaker unknown.”