Check an AI answer for made-up numbers
Paste what an AI assistant said and the report or table it should have answered from. Every number in the answer that isn't in the data is marked: a total it added up, a share it worked out, a figure it rounded or invented. It runs in your browser; nothing is sent anywhere.
Pujan Motiwala · Checked against the assistants' own code on

How the check works
- Any format. ₹8,72,020, 872020 and 8.7202 lakh are the same number; so are 1.2 crore and 1,20,00,000. Commas, the rupee sign and spaces are ignored.
- No rounding. “About ₹12.4 lakh” against ₹12,43,500 is marked. A rounded figure is one the assistant made (a mild hallucination), and the owner should see the real one.
- Whole numbers only. 25 doesn't count as found because 1,250 is in the data.
- Single digits are skipped (“3 invoices”, “on 2 Oct”): too common to tell anything.
Where the rule comes from
It is a grounding check, the one Atelier AI, the assistant in Savistar & Saag, an interior design studio and furniture workshop, runs before it shows a written line: if the sentence contains a number that isn't in the table, the sentence is dropped and only the table is shown. The figures themselves come from fixed look-ups in code, never from the model. How that works is in an assistant that only answers from your own records. To use the check on a vendor, run it on the answers you collect with the vendor test.
Paste an assistant's answer and the report it should come from, and see every number that isn't in the report.
Ask Atelier AI two questions on the interiors demo · how Ahrom Labs builds assistants
Official sources
- Digital Personal Data Protection Act, 2023, section 8. Ministry of Law and Justice, Gazette of India, October 25, 2026.
What this page is not
A guide written by Ahrom Labs from the assistants it runs. What each assistant answers is stated as it stood on the date checked; a vendor's assistant should be judged by the tests, not by this page. Not legal advice on data protection. Last updated .
Talk to us about your systems.
Tell us what's held together with workarounds right now. We reply to every message ourselves.