AI can be convincingly wrong. Have every answer checked before you act on it.

Have the facts, the reasoning and the sources checked separately, and not by the same AI that gave the answer. Vera does this automatically with several AI models and shows you every improvement, every disagreement and every source. You then decide what to do with the answer.

In Vera, one model writes the answer and other models check it: one checks the facts, one reads it critically and one looks for current sources. That way made-up facts, wrong assumptions and missing sources stand out sooner. Vera does not guarantee that every error is found.

A single AI model often judges its own answer, and then easily overlooks its own mistakes. In Vera, other models check the answer. If they disagree, you see it, instead of one smooth answer with the doubt polished away.

Yes. Vera has an AI answer checked by the models you choose for the conversation. Each model has its own task, such as checking the facts, reading critically or finding sources. What they find shows up in the final answer.

If you switch on the source check, Perplexity looks for current sources for the main claims. The sources and any improvements appear in the overview, so you can see what the final answer is based on.

No. ‘Checked’ means the answer went through the checks you set, and that you can see the improvements, sources and disagreements. Vera does not guarantee that an answer is correct and does not replace your own judgement.

Yes, with the Semantic Privacy Shield. Names and other sensitive data are replaced with placeholders in your shielded Vera environment, and Vera checks the text before an external AI model receives it. If that check fails, no document text is sent.

Then Vera shows it, with the arguments on both sides. That difference is useful information: it shows where the uncertainty lies, so you can weigh it yourself.

Yes, in your own shielded environment, which nobody else can access. Documents in Vera Office stay there until someone with the right permissions deletes them. A PDF you share in the chat that goes through the Privacy Shield is also kept in your environment, together with the shielded version.

Yes. For each conversation you choose which models take part and how thoroughly a question is checked. For each step you see what it costs.

You pay per checked question, not per credit. A question that goes through all the checks counts as one, however many AI models work on it. Starter is free for the first three months, with 10 checked questions per month; after that you choose a paid plan, from € 79.00 per month excluding VAT. If you run out of questions before the end of the month, you buy a bundle or choose a larger plan. All pricing.

Then checking stops and you see a clear message. Nothing is added automatically and you will not get an unexpected bill afterwards. You choose: wait for the new month, buy a bundle of questions or move to a larger plan. Prices are based on average use in your profession, so light and heavy use balance each other out.