In-app support chatbot

What it can answer

The support chatbot draws on two sources at once:

• WebGPT documentation — the same articles you're reading now. Ask "how do I connect a WordPress site?" or "what does Trusted Sources do?" and you'll get an answer drawn from the docs.
• Your own account data — when a question is about you specifically, the bot calls a server-side lookup. It can answer:
  • "What plan am I on, and when does it renew?"
  • "How many articles can I create today? How much have I used?"
  • "Which WordPress sites are connected to my account?"
  • "What articles have I made recently?" (last 20)
  • "What chatbots have I created?" (Independent plans)
  • "How many conversations has my chatbot had?" (Independent plans)
  • "Are my content autopilots running? Which ones are failing?"
  • "Why did my autopilot fail yesterday?" — pulls the last 5 runs of a specific bot and shows the failure reason for any that didn't succeed
  • "Are all my integrations healthy?" — rolls up Cloudflare, Akamai, Fastly, Google Analytics, WordPress sites, and platform connections, with a sample of any that need attention
  • "Is anything wrong with my account?" — surfaces active warnings: credit card expiring, payment failed, trial ending, AI vendor key invalid, WordPress site unhealthy, knowledge-base documents stuck indexing, autopilots failing repeatedly

The bot detects what page you're on and uses that as context. Asking "how do I add images?" from the article editor gets you an answer about image insertion in articles; asking it from your account page would prioritize different information.

What it cannot do

The support chatbot is read-only by design. It cannot:

• Create, edit, or publish articles for you
• Connect or disconnect WordPress sites
• Change your plan, cancel your subscription, or modify billing
• See data belonging to other users or other accounts
• Run AI Visibility scans, theme builder generations, or other paid features

For anything that would change something, the bot will explain where in the app to go and you take the action there.

Languages

The chatbot detects the language you write in and responds in the same language. Supported: English, Hebrew, Russian, Chinese, Spanish, Portuguese, German, and Arabic — the same set as the rest of the WebGPT interface. Right-to-left rendering works correctly for Hebrew and Arabic.

Privacy and safety

A few things happen behind the scenes to keep account lookups safe:

• Your account ID is never trusted from the chatbot. Every lookup re-confirms your identity from your logged-in session — the chatbot itself cannot ask the server for another user's data, even if someone tried to trick it via a message.
• Sensitive fields are stripped before any result is shown to the AI or written to logs — internal IDs, API key fragments, password hashes, and similar are filtered out automatically.
• Each tool call is logged. The platform keeps an audit row for every lookup so abuse can be detected and so support can debug failed lookups.
• A per-conversation rate limit caps how many lookups can happen in a single chat — prevents runaway loops without affecting normal use.

Hiding the chatbot

The chatbot widget is shown by default for every user in the account. If you find it distracting, you can hide it for yourself only — the toggle does not affect other team members in multi-user accounts.

Toggle it off from Account settings in the Contact & notification preferences section. The widget disappears immediately on the current tab and on any other open tab once the heartbeat refreshes (within seconds).

To bring it back, toggle it on again from the same place.

Different from the website chatbot

The chatbot on the public welcome page is a separate bot. It answers questions from anonymous visitors who haven't signed up yet — it has no access to any account data and is focused on helping people understand what WebGPT does before they create an account. The in-app support chatbot is what you see once you're logged in.