Start with the evidence

Your first comparison

1. Open the workspace and sign in with ChatGPT.
2. Choose Compare. Paste two statements, import text files, or import a supported Qatar foreign-ministry statement URL.
3. Choose Arabic over time or Arabic and English. Add publication dates and source links, and check the imported text.
4. Save the comparison. Your notes, original text and source information stay in your account.
5. Optionally request an AI draft using the processing consent checkbox. Return to Reports to inspect its status and quoted evidence.
6. Download a briefing and verify the original sources before using the assessment.

What runs automatically?

A daily task is configured for 08:00 Europe/Madrid (CET/CEST). Collection checks up to eight links per language index and eight recent captured URLs. The task also processes up to five submitted AI comparisons. A backlog or failed task can delay results.

Collection automatically queues eligible Arabic revisions and publisher-linked Arabic–English pairs among the 200 most recent snapshots, with up to 20 new jobs per run. Texts over 12,000 characters are skipped. Draft assessments appear in Sources; unmatched first captures cannot establish a change. You can also submit private comparisons. There is no instant question-and-answer service or email alert service in this release.

Reading the Sources screen

First capture means the text was added to the archive; it does not mean the position changed. Wording changed means a later collected version differs. Neither label proves a policy reversal.

Complete means the attempted checks finished without recorded errors, not that every publication was captured. Partial means some checks failed. Failed means no statement checks succeeded. Open the run details to see the errors.

A publisher status-page redirect means the source did not provide the requested statement. Existing evidence is retained, and that attempt is recorded as unsuccessful.

Access and privacy

No payment is required. Saved reviews and AI results are account-scoped; the official-source archive is shared with signed-in users. Submitted analysis texts can be read by the operator and scheduled task. Do not submit confidential material.