Primary tool · no signup

Compare what the answers show.

Bring public final answers, visible source URLs and the metadata needed to compare them. Leave with a transparent evidence package you can inspect and export.

Processing
In this browser tab
Inputs
1–5 observations
Output
JSON + CSV

Browser-local evidence workbench · method 0.1

Build an observation package.

Nothing entered here is sent to ai-fanout.com or an AI provider. The page analyzes only the text, metadata and visible source URLs you supply.

Local onlyNo account · no upload
01

Define the comparison

Coverage uses literal, case-insensitive phrase matching. It does not judge factual accuracy, quality or meaning.

02

Record the visible observations

01Observation 1
02Observation 2
2 of 5 observations

Use the synthetic demo to inspect the method before entering your own material.
05

What the measures mean

Visible source domain
A normalized host parsed from a URL that the user recorded as visibly cited or linked.
Recurring domain
A visible source host present in at least two included observations. Recurrence is not a quality score.
Answer overlap
The mean pairwise Jaccard overlap of normalized content-word sets. It describes shared vocabulary, not semantic equivalence.
Literal coverage
The share of observation-by-criterion cells containing the exact normalized phrase supplied by the user.

Before repeated collection

Freeze the rules that make comparison possible.

Build an observation protocol