Brand settings

The brand's name, working language, company details, public web address, social writing voice, brand images, incoming email filtering, the hours your team answers chats, AI usage limits and the danger zone — the eight tabs of Brand → Brand settings.

Where it is

Brand → Brand settings, the last of the brand's pages. The screen is headed Brand details & settings and is split into eight tabs:

General · Public address · Social · Email · Team hours · Advanced · Usage limits · Danger zone

The tabs are presentation only. It is one form with one save control in the title bar — a change you make on one tab is saved together with a change you made on another. With nothing to save the title bar shows a quiet Saved; once something differs from what is stored it shows Unsaved changes, Discard and Save changes, and leaving the page with unsaved edits asks first. On a wide screen the tabs are a menu down the left side, with Danger zone set apart in red at the bottom; on a narrow screen they collapse into a dropdown. Opening the page requires the Brands & chat widgets permission.

General

Who this brand is. The name is what you see in the dashboard; the rest is what your AI knows about you.

FieldWhat it does
Brand nameShown across the dashboard and in the sidebar. It cannot be empty.
LanguageThe language this brand works in, and the one its chat widget opens in. Anything you add without a language of its own falls back to this.
Company nameYaplet uses it when it introduces itself and whenever it answers as you.
WebsiteYour main site. Yaplet may point visitors to it.
What the company doesA sentence or two (up to 600 characters), so Yaplet does not have to guess from the name. Used in chat, calls, inbox suggestions and social posts.
Changing Language also changes the language your chat widget opens in, immediately — there is nothing to change afterwards in the widget designer. It is the starting language, not a lock: a visitor can pick another one with the widget's globe button, and their pick wins.The widget's interface speaks 38 languages (English, German, French, Spanish, Chinese, Japanese, Hungarian and 31 more — the full list is in the Widget Languages article). You may set the brand to any language at all — knowledge search and what the AI writes follow it either way — but if it is outside those 38, the widget's own buttons and tab names fall back to English, and a line under the picker says so.Change it after you have written the widget's own texts, and the text boxes on the widget's Home tab show what is stored for the new language, usually nothing. Your texts stay filed under the old language: visitors reading that language, or one you translated them into, still see them, and everyone else sees Yaplet's built-in wording. The Languages tab shows a warning until you write the texts in the new language on the Home tab (they are translated into the other languages from there) or change Language back here. See Widget Languages.

Fill from website

Fill from website, at the top right of the General section, opens AI setup. Give it your website address and it fills in the company name, description, website, language, brand icon and brand logo. If your plan includes Yaplet, a checkbox — Also read my website into the AI's knowledge — adds your whole website to what the AI knows in the same run: the first pages are learned right away and the rest follow in the background, up to 2,000 pages, and the site is then kept up to date like any website you add on the Knowledge page; on a plan without Yaplet that checkbox is not shown at all. AI setup is paid like other AI work, from your free 14 days' credits, bought credit or your plan; if none of them can pay, it does not run and the payment window opens.

If those details already have something in them, it asks you to choose Keep what's there (only empty fields get filled) or Replace it. Keep is the default.

Brand images

Two images, both on the General tab:

  • Brand icon — square. Appears beside the brand in the menu and in the brand switcher.
  • Brand logo — wide. Shown in the header of your public knowledge base, documentation and roadmap.

Each one is one main image plus an optional alternative tied to a single color mode. It is not a light/dark pair: the main image is used in both light and dark mode, and that is all most brands need. Only if your image disappears against one of the two backgrounds — an all-white or all-black mark — do you switch on Add a second version for light or dark, choose whether that second version belongs to Dark mode or Light mode, and pick it. The main image then covers the other mode, and a What gets shown preview puts both side by side so you can see the result. The alternative's own picker sits on the background it will be seen against, so a white logo is not judged against a white tile.

Switching the second version back off throws it away rather than hiding it, so a stored image can never come back unexpectedly.

Images are chosen from your organization's shared image library — the same library behind the chat widget's images and the social watermark. Uploads are PNG, JPEG or WebP, up to 10 MB, and they join the library for reuse. See Your image library.

Public address

The <name>.yaplet.help address your customers reach this brand's public pages on. Your knowledge base, your documentation and your public roadmap are all served from it.

  • Web address — the part before .yaplet.help; letters, numbers and hyphens only. Suggest from name suggests a free address based on the brand name.
  • What's served there — what visitors find there right now, naming the knowledge base and the documentation set currently served, or None when nothing is behind the address yet. Which ones are served is decided on Knowledge; Manage in Knowledge goes straight there, and Open site opens the public site.
Changing the address breaks every link anyone already has. Old article and documentation addresses stop working immediately, and the old address becomes free for someone else to take. Saving a changed address asks you to confirm and names both the old and the new one first. If someone claimed the new address while you were typing, the save is refused rather than quietly moving you to a numbered variant.

Social

The first section, Brand voice & visuals, is how this brand sounds and looks in social posts. It applies to every caption and image generated for the brand, in the composer and on Autopilot, and it does not change how your AI answers chats or calls.

  • Social writing voice — applied to every social caption written for this brand.
  • Image style — one of the presets (photorealistic, minimal, illustration and so on), or No preset.
  • Image guidance — extra direction added to every image: colors, mood, things to avoid.
  • Default watermark — placed on generated images unless a post picks something else. Chosen from your images; a processed copy is made, so changing the original later leaves the watermark alone.

Below that, Connected pages lists the Facebook and Instagram accounts that post as this brand and whose messages land in its inbox. It is read-only here — which page belongs to which brand is chosen on Meta Integration, in the same place that decides where that page's messages go.

Email

How this brand treats incoming email on its connected addresses — yaplet.io and custom-domain alike.

  • Spam filter — on by default. Incoming email flagged as spam, or coming from a provably forged sender, is discarded before it ever reaches the inbox: no conversation is created and nobody is notified. Emails carrying a virus are always discarded, regardless of this switch.
  • AI answer filter — off by default. When on, a quick AI check reads each email before Yaplet would reply and decides whether it deserves an answer at all — customer questions do; newsletters, receipts and cold sales pitches don't. Filtered emails stay visible in the inbox, silently, with no automatic reply and no notification. The Filter instructions field takes your own plain-language rules on top of the built-in ones — for example, "Never reply to invoices."

Both settings are described in full in Yaplet.io Emails.

The AI answer filter only matters while Yaplet answers email. When this brand's agent is switched off, or its Answers on → Email switch on Brand → AI agent → General is off, incoming email goes straight to your team and there is nothing for the filter to filter. The tab says so at the top, with an Open Chat AI settings link to the AI agent page.

Team hours

When your team is usually online to answer chats. These are not your opening hours: set the hours your team answers chats, not when your business or shop is open — keep your opening hours in your knowledge, where the AI can read them.

  • Time zone — the time zone the hours below are in. It starts as your browser's time zone. Visitors are told the times in their own time zone.
  • Weekly schedule — one row per day, Monday to Sunday. Switch a day on to set from and to; a day that is off reads Offline. A day you switch on starts at 09:00–17:00, and Copy Monday to weekdays copies Monday's hours to the other weekdays. An end of 00:00 means midnight. A day's hours cannot run past midnight, so a late shift is entered as two days: until midnight, then from 00:00 on the next day.

If a day's end is not later than its start, both times turn red and Save is refused with "On some days the end is not later than the start." until you fix them.

What the hours do:

  • Yaplet mentions them only when asked. When a visitor asks when they can reach a person, whether anyone is there, or asks for a person, Yaplet answers with your hours — always as usually, never as a promise, and in the visitor's own time zone when it is known. It never says the team is away while someone is online. With nobody online and outside your hours, it says the team is away and when it is usually back. With nobody online inside your hours, it only says nobody from the team is free right now — it never says someone should be there. See Human hand-off.
  • The visitor is told when the team is usually back. When a visitor asks for a person, nobody is online and it is outside your hours, the website chat's card for leaving a message for the team adds a line like "Usually back Monday 09:00". On Messenger and Instagram, which cannot show that card, the "all our agents are currently offline" message gains a sentence like "Our team is usually back Monday 09:00." See Email, Ratings & Agent Handoff.
  • Phone agents use them too. When a call starts, a phone agent reads the date and time in this time zone, and it knows your team hours. It brings them up only when a caller asks when they can reach a person, and says usually, never as opening hours.
  • They are information only. They do not switch anyone online or offline: whether someone counts as online still comes from each agent's own Online status.

Switch every day off and the hours are not set — nothing anywhere mentions hours. A connected AI app can read and change them too: see Widget Tools.

Advanced

Two settings that control how this brand's AI reads your knowledge.

Search depth

How deeply the AI searches your knowledge for each message. Two levels:

  • Standard (default) — searches your content and puts the most relevant results first.
  • Deep — also searches two rewordings of the question, so it finds content that is phrased differently. Slightly slower and costs a little more.

At both depths the AI handles each separate topic of a visitor's message on its own and says plainly which topics it has no information on instead of guessing. The setting applies to chat, to phone calls and to the AI reply suggestions in the inbox alike.

Privacy masking

Hides email addresses, phone numbers and card numbers in what visitors write. Off by default. It covers chat, phone calls, visitor memory and AI-written ticket titles. It does not apply to social posts.

One switch, two different behaviors, and you should know both:

  • In chat, it redacts what is sent to the AI provider — the personal details never leave Yaplet.
  • On a phone call, it redacts what is stored. The caller's speech reaches the AI in real time, so masking can only clean the saved transcript afterwards.
Privacy masking is best-effort, not a guarantee, and it is not encryption. If the masking service is slow or unavailable, the text goes out unmasked and nothing records that it happened. Even when it is on, it does not cover the page addresses a visitor was on — which routinely carry email addresses and order numbers in the query string — or images they upload. Conversations recorded before you switched it on stay exactly as they were.

Usage limits

How much of this brand's AI one person can use in a rolling 24 hours — in chat and on the phone. Six numbers, every one counted separately per identity. Leave a field empty for the default. There is no off switch: enter a large number (up to 100,000) instead. The same Save button saves these. The limits belong to the brand, so changing them needs the Brands & chat widgets permission.

AI chat answers

FieldCountsDefault
Per website visitorone visitor of the brand's chat widget30
Per email senderone email address (every email answer re-reads the whole thread, so this one is lower)10
Per Facebook / Instagram userone Messenger or Instagram user30

A workflow message with Let the AI say it on counts as one answer too. Once the person has used their answers, such a message goes out exactly as you typed it instead.

What happens at the limit:

  1. The message that would need answer number 31 gets one short closing message instead — written in the visitor's own language, and it never mentions limits or quotas.
  2. The AI is paused for that person for 24 hours. Their conversation gets an agents-only line — "AI paused — daily AI limit reached (30 answers in 24 hours). The AI resumes at …" — visible in the inbox only, never to the visitor. Conversation starters are hidden for them while the pause runs; they can still type.
  3. Right after that closing message, the conversation becomes an agent request: your team is notified, and the visitor sees Yaplet's own handoff line in their language — or, when nobody from your team is available, the card for leaving a message for the team (on Messenger and Instagram, the offline message). From then on it is an ordinary conversation for your people.
Raising the number does not shorten a pause that is already running — it always lasts 24 hours from the moment it started. The limit protects your AI budget from one person (or one script) asking endlessly; a normal support chat never gets near it.

Phone calls

FieldCountsDefault
Minutes per callerone phone number, per rolling 24 hours15 min
Calls per callerone phone number, per rolling 24 hours5
Minutes for withheld numbersone shared allowance for every caller whose number is hidden or unknown30 min

A caller over the limit (or with less than a minute left) is declined before the call is answered — no greeting, no charge — and the agent's call list shows Declined — daily limit. A call in progress whose minutes run out ends gracefully: the agent finishes its sentence, says a short goodbye in the call's language that cannot be talked over, and hangs up. Details on Pricing and limits.

Danger zone

Deleting a brand permanently removes everything it owns: its chat widget, its knowledge, its products, its AI agents and its workflows. It also takes the brand out of AI Social Posting, so the autopilot that posted for it goes with it — the posts already published and the connected Facebook and Instagram pages are not deleted, but they are detached from the brand. There is no undo and no way to get any of it back.

  • Delete brand is always here, whether or not the brand has a chat widget. It opens a dialog that counts what will actually be removed before you commit, and arms its button only once you have typed the brand's name.
  • The chat widget is deleted along with the brand, and it is listed first in that count. Deleting only the widget — leaving the brand and its public pages alive — is a separate action in the widget designer's own Danger zone.
  • Phone numbers are the exception: they are not deleted. They become unassigned and keep billing until they are released at Settings → Connections → Phone numbers, which only the organization's owner can do.
The widget is destructive in its own right, so read that line in the dialog carefully. It takes its whole history with it: the conversations and their messages, the visitors behind them, the session replays, and the banner campaigns that ran in it. The chat bubble disappears from every page the snippet is installed on.

For the full list of what a brand owns and what deletion destroys, read Brand.

© 2026 Yaplet. All rights reserved.