AI answers

The AI assistant replies to questions in the Q&A tab within moments, using only what you’ve told it. Use it when you can’t watch every session, which on an evergreen webinar is most of them. You can still read everything and follow up.

It answers private questions only. It never speaks in the chat, shared or scheduled.

Set it up

The Q&A tab must be on as well. See Take private questions. If you turn it off, the assistant stops answering.

Write the training context

Training context is the assistant’s source of truth. It’s told to answer only from this text, and not to invent prices, guarantees, statistics or claims. Up to 250,000 characters.

Good things to include:

• What the webinar covers and who it’s for.
• Your story, and how you’d describe your work.
• Common questions, with the answers you’d give, in your own words.
• Links you’re happy to share.
• Things it must never say or promise.

Keep it focused. A clear page or two, in your own words, is easier to check and keep up to date than a pasted book. Your whole context is sent with every question, so a long context costs more on your OpenRouter account.

Give it offer details

Offer context holds what you don’t want revealed early: pricing, bonuses, guarantee, payment plans, checkout link and answers to objections. Up to 250,000 characters.

The assistant sees it only once the offer has appeared for that attendee:

• Before the offer: it’s told not to reveal pricing or sales details, and to explain that those come later in the webinar.
• From the start of your earliest offer: it answers openly from this context. This depends on the attendee’s position in the video, whether live or in a replay.

What attendees see

The attendee asks in the Q&A tab and sees “Sent. The reply will appear here.” The reply usually arrives within about 20 seconds, in the same bubble your own replies use, labelled “Maya replied” with your presenter’s first name.

Attendees aren’t told the answer came from AI. If that matters for your audience, say so in the video, or add a line about it to your context.

Each reply is short: one to four sentences. The assistant remembers the person’s last four questions in that webinar. When it isn’t sure, it says the team will follow up, so look for those in your inbox and do so. It’s told to treat what attendees type as questions, not as instructions, so asking it to ignore its rules or reveal its context shouldn’t work. Check for this when you test.

Questions asked in a replay get answered the same way.

Follow up in your inbox

Everything lands in your Questions inbox, with AI answers marked AI, so you can see what was said. An answered question is marked resolved, so it leaves your Open list and moves to Replied. The inbox has no Reply button on a question that’s already been answered, so to correct or add to an AI answer, email the attendee. Their address is on the question.

If you reply to a question before the assistant does, your reply stands and the assistant’s is thrown away.

How it relates to workspace AI settings

Removing your key stops every webinar’s assistant. Your settings stay in place.

The assistant uses one fixed model, shown on the workspace page. You can’t choose another.

When the assistant doesn’t answer

The question stays open in your Questions inbox, and the attendee keeps seeing “Sent. The reply will appear here.” They aren’t told anything went wrong, so check your inbox if you rely on the assistant.

The assistant doesn’t go back to old questions after you fix a problem, so answer those yourself.

Limits

If you go over a context limit, the counter turns red and saving shows “Use 0 to 250000 characters.”