When the assistant shows you the page

"Where do I click?" is the most common question there is, and a link is a poor answer to it. You press the link, you arrive on a page with eleven controls on it, and the explanation you were reading is on the page you just left.

So the assistant can do something better than send you there. It can open the page and then point — at the box, at the selector, at the button that submits — one at a time, with a sentence for each.

What happens when the assistant shows me a page?

It is a tour of a page you are already allowed to be on.

The assistant stops on each control in turn. For every stop you get a short heading and a sentence or two saying what that control is for and what it decides. Three to five stops is typical, and the whole thing takes about as long as reading this paragraph.

That is the entire feature. It explains what is in front of you, and it does nothing else.

What it looks like while it runs

The rest of the page dims, and one control stays lit — ringed and bright against everything around it. A small card appears next to it with the heading, the sentence, and a row of dots showing how many stops there are and which one you are on.

A few other things happen that are worth knowing so they do not surprise you:

• The page scrolls to the control. If the thing being pointed at is below the fold, the page moves so you can see it.
• The first stop is usually centred, with nothing highlighted — that is the "here is what we are about to do" card. Some walkthroughs end on one too, explaining what happens after you leave the page.
• Sometimes a middle stop is centred too. Some pages only reveal a control once you have answered the one before it, so if the thing it wants to point at is not on screen yet, the card sits in the middle and the words tell you what to look for instead.

Use Next and Previous to move between the stops, or the right and left arrow keys. Previous is not shown on the first stop, because there is nowhere to go back to. The last stop's button says Done and closes it.

It cannot fill anything in, and it never presses anything

This is worth stating plainly, because "let me show you" and "let me do it for you" are easy to confuse.

A walkthrough only draws on top of your screen. It types nothing, selects nothing, submits nothing, and changes nothing about your account. When it points at the amount box, you type the amount. When it points at the button that submits, you decide whether to press it.

That boundary is deliberate and it is not a setting anybody can change. The assistant does not fill in forms on anybody's behalf, on any platform, in any configuration — because the only way you actually see what you are agreeing to is to do it yourself on the real page, with the platform's own confirmations in front of you.

You can still use the page underneath

The dimming is only a tint. The page below it is live the whole time — you can scroll, click, and type into the very box being pointed at while it is still being pointed at.

That is on purpose. A walkthrough that locked the screen would mean reading it, closing it, and then trying to remember stop three. Instead you can follow along and do each thing as it is explained.

How do I stop a walkthrough?

Two ways, and neither costs you anything:

• Press Escape.
• Press the ✕ in the corner of the card.

There is no confirmation and nothing is lost. You are left on the same page, with whatever you had already typed still there.

Can I see the walkthrough again?

Nothing remembers that you have seen one. Start it again and it runs from the first stop, as many times as you like — there is no "you have already seen this" message, because that is exactly the wrong answer to somebody who is asking again.

Why did a walkthrough button appear?

A walkthrough is offered as a Show me how button, and it turns up in two places.

Under an answer in the conversation. This is the usual one. If what you asked is better shown than described, the assistant puts a small card under its reply naming the walkthrough and telling you how long it is — "I will open the page and point out four things, one at a time" — with the button beneath. You do not have to be on the right page for this: pressing the button is what takes you there, and the walkthrough starts once the page has loaded. No process needs to be running, and the card can appear wherever on the site you happen to be.

Inside the panel of a multi-step process. When the assistant is carrying you through a process — see When the assistant walks you through something — the panel that follows you offers the same button for the step you are on. That copy appears only once you are actually standing on the page the step belongs to, because at that point you are already there and "take me there" would do nothing.

Nothing ever starts on its own. Arriving somewhere and having your screen dim without being asked would be alarming, and you may well already know what you are doing — so pressing the button is always the last step.

Why does this screen look different?

A few of the busier screens — the trading terminals and the swap screen — shipped with their own built-in tour long before the assistant existed. Those tours were written by the people who built those screens and know things a general walkthrough does not.

On those pages the button starts that screen's own tour rather than drawing a second one. The card says so before you press it — it offers to open the page and start the walkthrough it comes with, rather than counting stops — and the tour itself looks a little different from the walkthroughs described above. That is expected rather than a fault.

If you never see one

Then the people who run your platform have either switched walkthroughs off or narrowed them to a shorter list. It is their decision, and you can always ask for a person instead — see The support assistant, and asking for a person.