How to use Discourse for LLM search

Discourse turns every community discussion into a durable search asset that agents can find and cite.

Search has three layers

Your knowledge has to be publicly indexed and cleanly structured

If your community's best solutions are buried in a Discord server or locked in Slack, AI can't find or cite them. To be visible today, your knowledge has to be publicly indexed and cleanly structured.

Discourse turns every community topic into a durable search asset: searchable through Google, retrievable by AI assistants, and surfaced through Discourse's own built-in search. Get the answer right, structure it well, keep it updated and fresh, and AI search agents will keep using that answer indefinitely.

Classic Google

Each Discourse topic gets its own URL, clean HTML semantics, and crawlable metadata, so Google can index every topic.

AI assistants

ChatGPT, Claude, and Perplexity answer from the public web. Public Discourse forums are part of it.

Discourse's own search

Semantic search and Ask AI find the right topic inside your community, even when the question is phrased differently.

Quality controls

Turn community answers into citation-ready knowledge

Searchability only matters if the content is accurate. When an AI cites a wrong or outdated answer, it does more damage than no answer at all. Discourse includes quality controls that maintain a high signal.

The goal is a body of community knowledge you'd be comfortable having an AI cite to a prospective customer. Discourse's moderation tools make that possible.

A ChatGPT answer about Unreal Engine tutorials, with a sources popover citing three Epic Developer Community topics

Trust levels

Discourse's trust system gates permissions based on participation history. New accounts can't immediately post links, move topics, or take moderator actions. Users earn access through good-faith engagement, which keeps spam and low-quality posts out of your indexed content.

Moderation tools

Staff can hide, split, merge, or lock topics. Wiki posts let trusted members maintain a single authoritative answer rather than scattering corrections across follow-up replies.

Solution marking

When staff or trusted members mark an answer as a solution, Discourse surfaces it at the top of the topic and structures it so search engines and LLMs can recognize the resolution. A topic with a marked solution is easier for a model to parse than one that trails off.

Translations

If your community runs in multiple languages, Discourse supports localization, and automated translation is available on every plan. Your tier sets how many languages you can run. More indexed answers in more languages make your community searchable to more people.

Built for search

Make answers searchable by humans

Each Discourse topic gets its own URL, clean HTML semantics, and crawlable metadata. On top of it, Discourse adds:

Discourse's Ask AI screen with the question 'How can I make a theme for my new discourse site?' and a sidebar of past conversations

Semantic search

Users find the right topic even when they phrase the question differently from the person who asked it first. Ask AI goes a step further, answering the question directly from your community's content rather than handing back the right topic to read.

Conversation summaries

Discourse can generate AI-powered summaries of long topics, so readers get the answer without wading through 80 replies.

Structured data markup

Discourse outputs QAPage schema for solved topics, which signals to Google and AI crawlers what type of content this is and which reply resolves the question. Core marks up ordinary posts too, with DiscussionForumPosting, Comment, InteractionCounter, LikeAction, Person, and Organization.

AI searchability

Public Discourse forums get indexed by the crawlers that power LLM retrieval. When someone asks Perplexity or ChatGPT a product question your community answered months ago, a well-structured Discourse topic is the kind of source those models reach for.

Best practices

Habits that make LLMs love your content

Structure gets your community indexed. These habits make each answer worth citing, so the topic a model reaches for is the one that solves the problem.

Walled gardens

Why hiding your content costs you citations

Slack and Discord operate as walled gardens, completely hiding community knowledge from the open web. Because search engines can't index these internal conversations, organic discovery is impossible. Instead of finding existing answers with a simple web search, users must create accounts, join specific servers, and ask the same questions again and again.

At Discourse, we keep every chat as open as you want it to be, and make sure every search agent can tap into your knowledge, elevating your brand in the process.

No organic discovery

Search engines can't index conversations inside Slack and Discord, so the answers your community already wrote never show up in a web search or an AI assistant's reply.

No true interoperability

These platforms force reliance on proprietary clients and brittle third-party integrations.

History held hostage

Free tiers often hold historical knowledge hostage, such as Slack's 90-day message limit, and exporting chat logs is structurally messy or restricted.

A higher barrier to entry

By gating discussions behind mandatory logins and prioritizing synchronous chat, these platforms raise the barrier to entry and alienate asynchronous contributors.