Overview

Chirp AI is good at writing replies, but it can only be as accurate as the information it has about your operation. Knowledge snippets are how you give it that information — short, targeted pieces of property and policy knowledge that it pulls from when relevant. This lesson covers what snippets are, when to write them, and how Chirp uses them.

What a knowledge snippet is

A snippet is a small, focused piece of reference information — usually one or two short paragraphs — that answers a recurring question or documents a recurring fact.

Examples:

Parking at The Lakehouse: Two off-street spots in the driveway. Street parking is permitted on the east side of the road only. RVs and trailers are not allowed.

Late check-out policy (all properties): Late check-out (up to 1 PM) is $50 if the next stay allows. Submit requests at least 24 hours in advance via message. After 1 PM is not available.

Hot tub at Cedar Cabin: Open year-round. Maintenance happens every Tuesday between 10 AM and 2 PM — guests should avoid use during that window.

Each snippet is a fact. It's not a template (templates are prewritten messages). It's not a procedure (procedures live in your team's playbook). It's just the underlying info — written so Chirp AI can use it when crafting a reply, and so your team can see it when reading a conversation.

How Chirp AI uses snippets

When a guest message arrives, Chirp AI:

1. Reads the message
2. Searches your snippet library for relevance
3. Surfaces relevant snippets in the conversation's context panel
4. Uses the content of those snippets when drafting any reply suggestion

This is what makes AI replies grounded in your facts instead of generic. Without snippets, an AI might say something plausible-sounding but wrong about your hot tub schedule. With snippets, it says the right thing.

You'll see in the Quality Signals of an AI suggestion that one of the checks is "Factually grounded" — that signal is partly powered by whether Chirp had snippet content to anchor the reply.

Snippet scope

Each snippet has a scope — how broadly it applies:

• Company-wide — applies to all properties (e.g., refund policy)
• Property group — applies to a defined group (e.g., all Lake District cabins)
• Single property — applies to one specific listing

Use the narrowest scope that's accurate. A "company-wide" snippet about parking would be wrong for half your portfolio; a property-specific snippet about the same thing is right every time.

What to write snippets about

Good candidates:

• Anything you find yourself answering more than twice a month
• Anything that varies by property (parking, wifi, hot tub, pet policy, accessibility)
• Anything that changes occasionally and matters when wrong (cleaning schedules, late check-out fees, security deposit policy)
• Anything Chirp AI keeps getting wrong in suggestions — adding a snippet is the fastest way to fix it

Skip:

• Things that are already in the property listing/description (Chirp can read those)
• One-off explanations specific to a single guest (those go in internal notes, not snippets)
• Long policy documents (break them into focused snippets — one fact per snippet works best)

Writing a good snippet

A few patterns we've seen work:

• Lead with the answer. "Parking: two off-street spots." Not "We have a parking situation that..."
• Be specific. "Up to 1 PM, $50 fee" beats "small fee for late check-out."
• Note exceptions explicitly. "...except for groups of 8 or more."
• Date-stamp seasonal info if it'll change. "Pool open through October 15."
• Don't write in your guest voice. Snippets are notes for the AI and your team, not messages to guests. Chirp AI handles the tone when drafting.

Managing snippets

Snippet management lives in the Knowledge Hub, alongside your guidebooks. From there you can:

• Browse all snippets
• Filter by scope (company / group / property)
• Search by content
• Create, edit, archive snippets

Auto-extracted snippets

Every time a team member edits an AI draft before sending, Chirp AI compares what was originally suggested with what was actually sent. If the edit reveals something the AI didn't know — a factual correction, a policy detail, an operational note, a communication style preference — the system proposes a new knowledge snippet to capture it.

Snippets extracted this way are always generalized: no guest names, dates, booking amounts, or anything specific to one reservation. They're scoped as either property-specific or company-wide based on whether the knowledge applies broadly.

In Conversation Settings, you can enable "Require review for auto-extracted knowledge" — and we recommend it. With this on, proposed snippets appear in a review queue in the Knowledge Hub before being added. Your team approves, edits, or discards each one. This keeps the library accurate without requiring you to write everything from scratch.

Over time, auto-extraction builds up your knowledge base for things you'd never think to document proactively. More snippets means better AI suggestions — and if you run Autopilot, a higher auto-reply rate without changing any settings.

For larger teams, snippet hygiene is a low-grade ongoing task — every couple of months, scan for stale or duplicate entries and clean up. Outdated snippets are worse than missing snippets, because Chirp will trust them.

Snippets vs. templates: a quick reminder

Up next

Rules for Conversations — automate the things you'd otherwise do by hand on every conversation that meets a condition.