detector

Your Capstone Proposal Is a 1.0. Fancy Words Fixed Nothing.

The detector reads sentence shape, not vocabulary. Your thesaurus habit changes nothing; rhythm and concrete detail move the numbers.

On October 7, 2026, we ran a controlled experiment. Same base text: a generic AI-generated capstone proposal paragraph. Seven rewrite conditions. Two independent detectors, scores 0–1, lower means more human. Results: baseline scored 1.0 and 0.9999. Adding fancy thesaurus words: 1.0 and 0.9999. Stacking transition phrases like 'Furthermore' and 'Moreover': 0.995 and 0.9999. Those are the two moves most international students make at 2 a.m. with a capstone proposal doc open. They change nothing. What moved the numbers: deliberate short-long sentence rhythm hit 0.612 and 0.5334. Mixed sentence-length variation: 0.551 and 0.6614. First-person with one concrete project detail: 0.693 and 0.9146. And deliberate 'imperfection'—fragments, hedges, little surface mistakes? 0.986 and 0.9705. Essentially zero. Let me walk through what this means for your proposal, section by section.

International students write capstone proposals like they are performing 'academic English' at a tribunal. 'It is imperative to elucidate the multifaceted implications of the aforementioned phenomenon.' You swap 'show' for 'elucidate,' 'important' for 'imperative.' The detector reads your text the same way. Score: 1.0. Unchanged. I get it. Your instructor's past papers read like that. Your culture taught you formal vocabulary signals seriousness. But a capstone proposal is a project pitch. Your supervisor wants: what problem, what method, what timeline, what you actually produce. Fancy words do not buy credibility. They add syllables without adding meaning. The detector cannot distinguish 'elucidate' from 'show' because both sit inside an identical template structure. Your instinct to 'sound academic' is the single most expensive habit in capstone writing.

Then come the transition crutches. 'Furthermore.' 'Moreover.' 'In addition.' 'Consequently.' You learned these in your first-year writing course and they became armor. In a methodology section: 'Furthermore, a semi-structured interview protocol will be employed.' 'Moreover, thematic analysis will be conducted.' The detector registered 0.995 versus a 1.0 baseline. That is not improvement. That is noise. Transition words do not create rhythm. They create a metronome. Every sentence gets the same beat: transition, passive verb, noun phrase. The scoring tool reads metronomes as machine text because metronomes ARE machine text. Delete half your transitions. Replace them with sentence-length shifts. I am not asking you to be grammatically perfect. I am asking you to stop sounding like a template.

What actually moved the detectors was rhythm. Not vocabulary. Not hedges. Not 'imperfections.' Mixed sentence-length variation: 0.551, 0.6614. Deliberate short-long beats: 0.612, 0.5334. These numbers say something blunt: the detector reads sentence structure, not word choice. In your capstone proposal, this looks like: one long sentence laying out your methodology, then a short one. Two words. Then back to a 40-word construction explaining your data-collection timeline. The contrast is what registers. Try this on your problem statement right now. Count your sentences. Are they all 25–35 words? Fix that. Break one at 8. Let another run 45. The scoring tool reacts to the shape of your paragraph, not the words inside it. This is the lever. This is exactly what the easydue humanizer restructures when you feed it a flat paragraph—it changes your sentence lengths, not your vocabulary.

Now the most annoying finding. 'Add deliberate imperfection.' 'Hedge your claims.' 'Throw in a fragment so it reads human.' This advice circulates in every forum thread about academic writing. We tested it. Fragments, hedges, small surface mistakes: 0.986 and 0.9705. The detector barely blinked. Why? Because fragments and hedges are local edits. They do not change the paragraph's structural shape. You put 'somewhat' before a claim and the architecture is identical. Meanwhile, one first-person sentence with a concrete project detail—'I spent three weeks in the lab observing X'—moved one detector to 0.693. Concrete beats vague. Specificity is a structural change, not a word swap. Your capstone proposal should name one thing you did, saw, or will physically touch. That is how you signal a human is writing.

Here is how you rewrite your proposal's three highest-risk sections. Problem statement. ❌: 'This project endeavors to contribute to the extant body of knowledge regarding sustainable urban drainage in Southeast Asian megacities.' ✅: 'I grew up in Jakarta. The 2023 floods shut down my neighborhood for eleven days. No engineering literature I found addresses drainage retrofit in informal settlements specifically. That is the gap.' Methodology. ❌: 'In order to investigate the aforementioned variables, a mixed-methods approach will be employed, further supplemented by triangulation.' ✅: 'Surveys first. Two hundred households, stratified by flood frequency. Then six interviews. If the numbers show X, I drill into why. If they do not, I stop and rethink.' Timeline. ❌: 'The project will be conducted over approximately twelve weeks, with deliverables at regular intervals.' ✅: 'Week 1–3: site access, ethics clearance. Week 4: pilot survey. Week 5–8: full data collection. Week 9–10: analysis. Week 11: draft. Week 12: revision with my supervisor. I need the ethics form signed before week 4.' Run your draft through easydue after rewriting to check whether you actually varied the rhythm or just made everything shorter. The shape is what matters.

One more number. When we ran the easydue humanizer on strong mode over the raw baseline: detector A dropped from 1.0 to 0.539. Detector B dropped from 0.9999 to 0.0261. Same text, two detectors, radically different scores. What does that mean? The two scoring systems measure different things. Optimizing for one may not help the other. The humanizer restructures sentence shapes—length variation, rhythm breaks, concrete anchoring. It does not swap 'utilize' for 'use' and call it a day. But the gap between 0.539 and 0.0261 tells you: no tool is a silver bullet. Read your own proposal for rhythm before you trust any score. The best capstone proposal sounds like a person explaining their project to their supervisor in a 15-minute meeting. Write it that way. Then use the tool to check the shape. Not to fake it.