This guide covers how to run an analysis in Undersignal™, how to read the report, and how your material is handled. For the scoring framework itself, see Methodology.
The wordmark takes you to your dashboard, your account home. Methodology and Help sit in the top bar wherever you are. When you are running or viewing a report, the bar also carries New Analysis, the sun and moon to switch light and dark, and a gear that opens Settings.
The gear opens four controls:
There are three ways to start:
Undersignal analyzes up to roughly 50,000 characters. Longer inputs are read from their opening and closing portions rather than in full, so for a long document, paste the section you want assessed rather than the whole thing.
Every analysis you run is saved to your account and reopens exactly as it was generated, so the report you cite is the report on file. To return to a result, open it from your history rather than running the input again. A URL is the cleaner anchor when someone else needs to verify your work: it is keyed to the published content, so anyone analyzing the same source reaches the same report.
Export PDF (or Print) produces a formatted file to share or save off platform. Download JSON gives you the structured data behind the report, including every dimension score and the source and output fingerprints. To share a report, export the PDF and send the file.
The report reads top to bottom, from the headline judgment to the evidence beneath it.
The score runs from 1 to 10. It measures how heavily a piece is constructed to move the reader, not whether it is true. A higher number means more deliberate persuasion mechanics: the rhetorical and structural choices that steer how a reader responds. The report places your score on a scale from Minimal to Critical and against the average for its content type, because a 6 in opinion writing means something different from a 6 in straight news.
Beside the score is a construction label, ordered from least to most deliberate:
The score measures intensity. The label names the kind. Read them together.
Three quick reads sit directly below the score:
The Executive Summary is the brief: the verdict, what the piece does, who benefits, how transparent it is, and why it did not score higher. The Decision Snapshot turns that into action, with separate reads for finance, communications, and policy desks, plus how far the piece is likely to spread and what to watch next.
Argument Devices is the core of the report. Each rhetorical move is tied to a specific quote, named by mechanism, and rated for confidence, so no claim floats free of the text. From there the lens widens: the Cross-Ecosystem Map estimates how the same facts are likely to be reframed for different audiences, Who Benefits follows the incentives, Missing Perspectives marks what is absent, and Dimension Analysis breaks the score into its 14 signals across 6 clusters so you can see which signals drove the score.
The Source Certificate and Reproducibility Certificate record exactly what was analyzed and how, which is what lets the report hold up when you put it in front of someone. The Canonical Evidence Record preserves the underlying data. For the framework behind the scoring, see Methodology.
Your material stays yours.
This matters most for the people who paste statements that have not been released yet: your drafts never leave your account and never enter the corpus.
Is my pasted content used to train or benchmark Undersignal?
No. Pasted and uploaded text stays private to your account and is never added to the corpus. Content you analyze from a public URL may be, because it is already public.
Can anyone else see my reports?
No. Your reports are visible only to you.
How do I get the exact same report back?
Open it from your history. Saved reports never change, so reopening a result is always identical. A different score only comes from editing the text or submitting something new.
What does my report compare against?
A reference corpus of analyzed content of the same type. Public content analyzed by URL can become part of it; pasted and uploaded text never does.
How many reports can I run and keep?
Founding membership has no per-report limit for normal use, and your reports are saved with no limit on how many you keep.
How consistent is the scoring?
A saved report never changes: open it from your history and it is identical every time, because it is stored exactly as produced rather than regenerated. Resubmitting the same text pulls that saved report back rather than re-scoring it. Editing the text, or submitting something new, is a different analysis with its own score, measured against the same fixed rubric version.
For help or questions, email hello@undersignal.ai. For account or billing, email accounts@undersignal.ai.