What the assistant in the help centre is, where to open it, what it knows and what it cannot know, why every answer carries a source link, how to write a question it can answer, which subjects it hands straight to a person, and how it behaves on a phone versus a desktop.
Last updated: September 30, 2026
Every page of the help centre has a help assistant behind a round button in the bottom right corner. You describe your problem in your own words; the assistant reads the guides in this help centre, answers, and shows a link to the article it took the answer from. Instead of guessing which keyword the search box wants, you can simply say what went wrong.
This guide explains what the assistant is, what it cannot know, how far you can trust its answers, and when it puts you straight through to a person.
You are probably here for one of these:
The assistant lives in the help centre, in the bottom right corner of every page. Press the button and the chat panel opens.
The assistant requires a session. If you are not signed in, the panel draws a sign-in card instead of the question box and sends you to the panel's own sign-in screen; afterwards you land back on the help page you were reading. Coming from the panel, the Help button in the top bar carries your session with you and you are not asked for a password again.
💡 Tip: Open the assistant from the screen where you hit the problem. Which screen you came from travels with you, and the article about that screen is pushed to the front of the search results.
When the panel opens you see one of two things: the conversation you already have in progress (scrolled to the bottom), or an empty question box. You never have to go hunting for a conversation you left half finished.
The assistant knows the whole guide in this help centre. Two things happen on every question:
The answer is built from those pieces and nothing else. The assistant may not lean on general knowledge, on what other products do, or on a guess. Writing a screen name, a button name, a menu path, a number or a duration that is not in the guide is forbidden.
It also follows the thread of the conversation: the last few turns travel with the question, so when you say "that didn't work, I still get the same error" it knows what you mean.
This is the part people misread most often. The assistant cannot look at your account.
| What you ask | What the assistant does |
|---|---|
| "How many messages do I have left?" | Tells you where the quota is shown; cannot tell you the number. |
| "How much is my invoice this month?" | Will not state an amount; hands the subject to the support team. |
| "Is my Instagram account connected?" | Describes which screen shows the connection status. |
| "Why didn't my automation run last night?" | Cannot read your logs; walks you through where to look. |
| "Did my payment go through?" | Cannot see your payment status; puts you through to a person. |
This is a deliberate design choice, not a gap: none of your account data is ever sent to the assistant. The detail is in Assistant limits, privacy and retention.
The assistant also does not do the work for you. It will not build an automation, connect an account, change a setting, send a message or close a ticket. It explains; you act. Letting it perform irreversible operations off the back of a chat sentence would be a bad trade.
Finally, you cannot attach files to the assistant. If you need to send a screenshot, use the support ticket route; the attachment rules are in How to open a support ticket.
Underneath every answer you see at least one source article badge. Press it and the article opens; if the assistant meant a particular section, the link takes you straight to that heading.
This is not decoration, it is a safety gate. The assistant has to declare which articles it used, and the server checks that declaration:
So behind every answer on your screen there is at least one article you can open and read for yourself. Whenever a sentence looks off, the next step is obvious: press the badge and read the source.
⚠️ Note: An assistant answer is a summary of the guide, not a replacement for it. Before doing something consequential — deleting an account, changing a plan, running a bulk send — read the source article.
The assistant is obliged to admit when it does not know. A half-correct answer does more damage than no answer: you lose time on the wrong screen and end up writing to support anyway.
You get an "I couldn't find this" answer in three cases:
All three end the same way: the assistant does not disappear, it offers you the support ticket route. How that works is described in From the assistant to the support team.
Unanswered questions are recorded, and that record shows which subjects the guide is missing. So an "I don't know" is not wasted — it is the reason the next article gets written.
The assistant is noticeably better with a well-written question. Three rules are enough.
Also worth knowing:
💡 Tip: Use the Solved and Not solved buttons under each answer. "Solved" records that the article genuinely helped and closes the conversation; "Not solved" tells the assistant to try another route, and if you keep pressing it, puts you through to a person.
On some subjects the assistant is not allowed to answer at all. It gives you a one-line reason and points you at the support team.
| Subject | Why the assistant stays out of it |
|---|---|
| Prices and fee amounts | Amounts vary by campaign, country and plan; a wrong figure reads as a promise. |
| Refunds and withdrawal terms | These have legal consequences and cannot be read off a guide summary. |
| Statutory periods, contracts, personal-data questions | They need a binding interpretation. |
| Invoice amounts, payment status, balance, quota | Account-specific data the assistant cannot see. |
| Security and access exceptions | They require identity verification. |
| Any promise about a date or an outcome | The assistant cannot commit on anyone's behalf. |
When one of these comes up you do not have to wait around: a support ticket opens immediately and the reply also arrives by e-mail.
The assistant replies in the language you write in. The guide index is built per language, so a question in English is answered from the English articles and a question in Turkish from the Turkish ones. If you switch the help centre's language, new questions are answered in the new language and your existing conversation stays as it is.
The assistant's name and icon come from the installation settings. If the administrator has not given it a name, it appears under a generic label ("help assistant") with a generic icon. The name makes no difference to how it behaves.
No. The model behind the help assistant is separate from the AI credit your account spends on automations, and it takes nothing from it.
If you send exactly the same text again, the assistant does not think it through a second time — it repeats its last answer. If you want a different answer, phrase the question differently or say what was missing.
That is temporary: the model did not respond in time or the request timed out. Try again in a minute. If it persists, open a support ticket — the assistant offers that route inside the panel.
No. Only you and the support team can see it. Other members of the same portfolio have no access to your exchange.
With Close conversation inside the panel. The confirmation appears in the product's own dialog. A closed conversation is not reopened; your next question starts a new one and the old exchange stays in your record.
First press the source badge and read the article; more often than not the difference is that your situation departs from what the article describes. If the answer really is wrong, press Not solved and open a support ticket. You do not have to retype what you asked and what it replied — the transcript is attached to the ticket automatically.
There are three possibilities: you are not signed in to the help centre, your plan does not include the assistant, or the administrator has turned it off. In all three cases the ticket route stays open — see How to open a support ticket.
Was this article helpful?