Does Landbot Remember Between Conversations?

Partly. Landbot remembers Fields: the values your flow saves about a visitor, such as a name, an email, or a flag that marks them as a returning visitor. It does not remember the conversation itself. A new conversation starts the flow from the beginning, the AI Agent block works from a short window of recent messages, and anything you did not save into a Field is gone.

What does Landbot actually keep between conversations?

Landbot keeps data in Fields, its name for variables. Your flow reads and writes Fields as it runs: an answer to a question block, a value you set with a Set a Field block, or data you pass in from your page when the widget loads.

Some Fields are system Fields that Landbot fills in on its own. Two matter for returning visitors:

• Created: the date the visitor first chatted with the bot.
• Last seen: the date of their most recent visit.

Both follow the visitor ID, and Landbot's own Conditions guide notes that the visitor ID depends on the bot's Second visit settings. That is the whole native memory: an ID, a few dates, and whatever Fields your flow deliberately saved.

What happens when a visitor comes back?

On WhatsApp, Landbot's help center states the default plainly: once a conversation reaches the last block, any new message restarts the bot from the beginning. Without a workaround, a returning user goes through the exact same flow every time, including questions they already answered.

The web widget behaves the same way at the conversation level. You can pass values in when the widget loads, and you can update them during a chat, but starting a fresh widget instance means a fresh conversation. Landbot's developer docs even describe destroying the current instance and creating a new one when you want a bot to begin again with a clean slate.

Nothing in either channel replays the last conversation to the bot. The transcript exists for you to review. It is not memory the bot can use.

Does the Landbot AI Agent block remember past conversations?

No. The AI Agent block runs a conversational loop inside the current conversation. In its setup, you choose whether the agent sees the last few messages of this conversation, and you give it a Knowledge Base made of a document, typed text, or a website.

That Knowledge Base is reference material you supply. It is not built from past chats. There is no step where last week's conversation with the same customer becomes context for this week's conversation, unless you stored what mattered from that chat somewhere the block can reach.

So the AI Agent can be good inside one conversation and still greet the same person as a stranger tomorrow.

How do you make a Landbot bot recognize returning visitors?

Landbot's own documented pattern uses one Field as a memory flag. It takes four steps:

1. Capture the first message. Put a block at the very start of the flow that saves what the visitor sends into a Field.
2. Check for a returning flag. Add a Conditions block that tests whether a Field such as @returning_visitor already has a value. Empty means first visit. Filled means they have been here before.
3. Route the two groups separately. Send first-time visitors through your main flow and returning visitors through a shorter path that skips what they already did.
4. Set the flag at the end. Add a Set a Field block as the last step of the main flow that sets @returning_visitor to a value such as yes. Drop-offs who never reach the end are not marked, so they correctly get the main flow next time.

You can extend the same pattern: store a plan name, a support outcome, or a preference in its own Field, then branch on it next time. Each fact you want the bot to act on has to be a Field you planned in advance.

Where does the Fields approach stop working?

Fields remember single values, not understanding. A flag can tell the bot that a customer visited before. It cannot tell the bot what the customer was trying to solve, what was already tried, or what they decided.

Three limits show up in practice:

• You must predict what to save. If a fact was never written to a Field, it cannot be recovered later.
• The memory stays inside Landbot. A Field saved by your website bot is not available to an agent you run in another tool, or to a scheduled workflow reviewing the same customer tomorrow.
• There is no recall by meaning. The bot can check whether a Field equals a value. It cannot search past conversations for the one detail that matters now.

For memory that has to survive across conversations, channels, and tools, the store has to live outside the bot flow. A hosted memory layer works this way: the agent saves each exchange as it happens and loads the relevant context before it answers, using semantic and keyword retrieval rather than exact Field matches. Vilix AI is one example. A built agent connects to it with an API key carried as a Bearer token to its MCP endpoint, so the same memory follows the customer conversation whether it happens in Landbot, on WhatsApp, or in a different tool entirely.

Should you use Fields or an outside memory layer?

Use Fields when the fact is small, known in advance, and only needed inside Landbot: a returning-visitor flag, a language choice, a plan name. That is exactly what they are built for, and the setup takes minutes.

Use an outside memory layer when the bot needs the content of past conversations: what was discussed, what was promised, and where things were left off. That is the difference between recognizing a visitor and remembering them.

FAQ

Does Landbot remember between conversations? It remembers Fields saved about a visitor, plus system Fields such as Created and Last seen tied to the visitor ID. It does not remember the conversation itself, and a new conversation restarts the flow.

Does Landbot remember returning visitors on WhatsApp? Only if you build for it. By default a new WhatsApp message after a finished conversation restarts the bot. Setting a Field at the end of the flow and checking it at the start lets you route returning visitors differently.

Does the Landbot AI Agent remember previous chats? No. The AI Agent block works inside the current conversation, with a short window of recent messages and a Knowledge Base you provide. Past conversations are not part of its context.

Can Landbot greet returning visitors by name? Yes, if the name was saved in a Field during an earlier visit and the visitor is recognized by the same visitor ID. The bot can then insert that Field into its greeting.