Conclusions
Detects boilerplate conclusion phrases like 'In conclusion' and 'In summary' that AI models rely on heavily to wrap up text in a predictable, template-driven way.
Technical description
Identifies formulaic conclusion patterns by matching against a dictionary of boilerplate closers ('In conclusion', 'In summary', 'To summarize', 'Overall, this study demonstrates'). Analyzes the final paragraph's structure for template patterns: restatement of aims, summary of findings, future work suggestion, and significance claim. Measures how closely the conclusion follows the most common AI-generated conclusion template.
How it works
Layer 1 (deterministic): Identifies the conclusion section or final paragraphs. Matches against boilerplate conclusion openers. Checks for the generic conclusion template (restate → summarize → significance → future). Measures similarity to the most common AI conclusion patterns. Flags documents with multiple boilerplate closers.
Why this matters
AI models almost always end text with 'In conclusion' or 'In summary' followed by a generic restatement. Human authors vary their conclusion strategies — some end with specific implications, others with open questions, and some with direct calls to action. The formulaic nature of AI conclusions is one of the easiest signals to detect.
Score thresholds
- 0-1
- Original conclusion strategy with specific implications
- 2-3
- Some formulaic elements mixed with original content
- 4-5
- Entirely boilerplate conclusion following AI template
Limitations
Many journals and style guides recommend starting conclusions with 'In conclusion'. Short abstracts may legitimately use formulaic closers. The indicator works best on full papers, not isolated paragraphs.