It Will Not Invent Evidence
The engine never fabricates statistics, years, study findings, examples, organizations, quotations, sources, or links to make a rewrite sound more persuasive. When a stronger version of your argument would require evidence your draft does not supply, the engine marks the exact spot with[verified evidence needed] instead of inventing a fact. That marker is a precise instruction to you, not a gap the engine fills on your behalf.
It Will Not Pretend a Source Is Verified
The engine never claims a citation has been checked when it has not. When a factual statement needs checking, it tells you exactly what to verify and continues evaluating the logic of your argument independently. You will never see a source described as confirmed unless the source pass found a match — and even then, the match is flagged as tentative.It Will Not Fake Counterargument Support
The same rule applies to the attack tree. The engine will not claim that studies, institutions, or outcomes exist unless your draft supplied them. When building counterarguments, it describes the kind of evidence an opponent would need rather than pretending that evidence has already been found. That distinction keeps the attack tree useful without making it a source of invented ammunition.It Will Not Pad the Report
Sections contain as many distinct, useful points as your draft actually supports — not a fixed quota of filler observations. The same weakness is not repeated across multiple sections in slightly different words. If a section is short, it is because your draft gave the engine fewer real issues to surface in that dimension, not because the engine ran out of things to say.These guardrails are why a low source score does not crush your reasoning score. By keeping “what is proven” and “what still needs proof” clearly separate, the engine can give you an honest picture of your argument’s logic even when the citation work is unfinished.