TrustAn answer you can
An answer you can
check, not just read.
Every claim on this page describes what the product actually does today. Where Simpledot cannot promise something, it says so.
Answer quality
Compared, judged, and shown with its confidence
A single model answering alone has no way to notice its own mistake. On eligible paid requests Simpledot runs several models and lets a judge pick — then shows you the reasoning instead of hiding it.
- Verdict panel: which answer won, confidence level, and whether a second engine agreed or disagreed.
- Disagreement is surfaced, not smoothed over — when engines split, you see that they split.
- Exact-math engine handles EMI, SIP, FD, GST, percentages and date arithmetic deterministically; the model is told the verified figure instead of estimating it.
- Free chat uses one model and the interface says so. No fake 'multi-AI' badge on a single-model answer.
High-stakes guard
Tax, medical, legal and money answers are labelled
These are the questions where a confident wrong answer costs real money or health. Simpledot detects them and refuses to let them look more certain than they are.
- Label under the answer: 'answered from model knowledge — unverified' or 'grounded in N live sources'.
- 'Verify with live sources' re-asks the same question with live web research and asks for dated official sources, flagging anything outdated.
- The model is instructed to say 'not sure' instead of guessing on these topics, and to prefer rules with an effective date.
- Short definitional lookups ('what is GST?') are not badged — the guard is for advice, not vocabulary.
Sensitive mode
Identifiers are masked on your device
Indian users regularly need to paste a PAN, an account number or a phone number to get a useful answer. Sensitive mode makes that safe by making the real value never leave the browser.
- Masks Aadhaar, PAN, card numbers (Luhn-checked), phone numbers, emails, IFSC, UPI IDs and passport numbers before sending.
- Placeholders like [PHONE-1] keep the question answerable; the chat shows 'Masked on your device: 1 phone number, 1 PAN'.
- Private chats: answered normally, stored nowhere — not history, not cache, not Brain.
- Both are per-device settings under Settings → Privacy and are off by default so nothing surprises you.
Your data
Deletion is a button, not a ticket
Memory is visible and editable in the Brain, and the whole account can be removed without talking to anyone.
- Clear all chats, or delete the account and all data, from Settings → Data & usage.
- Account deletion removes conversations, Brain, preferences, workspace runs, usage and the sign-in itself.
- Invoices are kept for tax records but detached from your identity.
- An active subscription blocks deletion until cancelled, so a paid plan is never lost by accident.
Security
Boring, documented, enforced
Row-level security on every user table, bot protection on sign-up and chat, cross-origin request refusal on the API, and private-IP-safe fetching for any URL a user or a citation hands us.
- Supabase Auth with row-level security — one user cannot read another's rows even if the application had a bug.
- Vercel BotID on sign-up, login and chat; rate limits on every costly route.
- Security headers (HSTS, nosniff, referrer policy, frame protection) are set at the edge; Content-Security-Policy runs in report-only mode while it is tightened.
- Full details and a disclosure contact on the Security page.
Reliability
Public status, real probes, honest refunds
If something breaks you should be able to see it without asking us.
- The status page runs live probes against each dependency with a 30-day uptime and incident log.
- Provider failover: when one AI provider fails, eligible requests move to another supported model.
- 7-day refund window on paid plans, in writing on the Refunds page.
Straight answers
- How does Simpledot decide which AI answer to show?
- On eligible paid requests Simpledot sends the question to several AI models in parallel, a separate judge model compares the responses, and the Verdict panel shows the chosen answer, a confidence level and whether a second engine agreed. Free chat uses a single model and says so.
- How do I know whether an answer was checked against live sources?
- Tax, medical, legal, money and safety questions are detected automatically. Under the answer Simpledot shows either 'answered from model knowledge — unverified' or 'grounded in N live sources' with the source count. A one-tap 'Verify with live sources' button re-asks the same question forcing live web research with dated official sources.
- Does Simpledot guarantee correct answers?
- No. AI models can misunderstand a question or agree on a wrong claim. Comparison, the Verdict and the verified/unverified label are aids that make errors easier to catch, not a guarantee. Confirm important financial, medical and legal decisions with a qualified professional.
- What does Sensitive mode do?
- With Sensitive mode on (Settings → Privacy), Aadhaar numbers, PAN, card numbers (Luhn-checked), phone numbers, emails, IFSC codes, UPI IDs and passport numbers are replaced with placeholders such as [PAN-1] in your browser before the prompt is sent. The original values never reach Simpledot's servers or any AI provider, and the chat shows exactly what was masked.
- Can I chat without anything being saved?
- Yes. Private chats (Settings → Privacy) answer normally but write nothing to your chat history, the shared answer cache or your Simpledot Brain. Each private turn is marked 'Private — not saved' in the conversation.
- How do I delete my data?
- Settings → Data & usage has 'Clear all chats' and 'Delete account and all data'. Account deletion removes your conversations, Brain memory, preferences, workspace runs and usage records and deletes your sign-in. Invoices are retained for tax records but detached from your identity. An active subscription must be cancelled first so a paid plan is never deleted by accident.
- Which AI providers process my prompts?
- Depending on the task and plan, prompts may be processed by OpenAI, Google, Anthropic, Perplexity and ElevenLabs (for Vaani voice). Simpledot does not sell conversation data. The Privacy Policy lists the current providers and what each receives.
- What happens when Simpledot is down or a provider fails?
- The public status page runs live probes against each dependency and keeps a 30-day uptime and incident log. When one AI provider fails, eligible requests fall back to another supported model rather than returning an error.
Related: Security · Privacy Policy · Live status · Refunds