The Discourse blog runs its comments in full app mode
We run the comments on our own blog with full app mode.
Read Understanding Discourse Trust LevelsProducts used: Discourse embedding (full app mode)
Challenge
We wanted readers to join the discussion on posts that have collected years of replies, without sending them to another site first.
Solution
Readers reply, like, and quote posts directly under articles such as Understanding Discourse Trust Levels. New readers add to topics that started years ago, and every reply also appears on Discourse Meta.
Keep readers on the page
Let readers join the discussion without leaving the article
Readers reply, like, and quote in place
When you turn on full app mode, readers get the whole Discourse discussion under each article. They reply, like, and quote without opening a new tab or losing their place in the story.
Readers sign in with an account they already have
Connect your existing login through DiscourseConnect, or let readers sign in with Google, GitHub, Discord, and other social accounts. Readers who already belong to your community start commenting straight away.
Readers get fast pages
You can load comments only when readers scroll near them, and set the comment area to grow with the conversation, so a post with no replies doesn't leave a gap and a busy topic has room.
Look & feel
Style embedded comments with your own CSS through Discourse themes so comments match your fonts, colors, and layout.
Turn commenters into a community
Upgrade every article with a discussion that keeps working after the news cycle ends
Each article gets a matching topic
The first time someone opens an article, you get a topic on your community that links back to the original. Readers who arrive from the article and readers who arrive from the forum join the same topic.
People find your discussions in search
Each discussion lives at its own public address on your community, so people searching Google, or asking ChatGPT, Gemini, or Claude, can find what your readers said about a story long after you published it.
Readers stay for the next conversation
A reader who comes to comment on one story finds categories, other topics, and people to follow. Your regular commenters become members, and your members come back for the next post.
You own the data
Discourse is open source. You keep every post and every account, and you can export them whenever you like. We don't run ads in your comments, and we don't sell your readers' data.
Moderate at scale
Keep comments civil without hiring a moderation team
Readers earn trust over time
New accounts start with limits on links and images. As readers participate, they earn more privileges, and your most trusted regulars help you flag spam and keep topics tidy.
Your moderators catch abuse before readers see it
Moderators set Discourse AI to flag NSFW content and toxic posts, and can hold first posts from new accounts for approval.
Moderators work from one queue
Readers flag posts, and your moderators review every flag, held post, and new account in one place and act in a click.
AI spam detection
Stop spam before anyone reads it
Spammers go after open comment boxes first. When they get a few posts through, your readers stop replying, and your editors spend their mornings deleting links.
With AI spam detection, we use a language model to read every post from new commenters, decide whether it's spam, and pull it down before your readers see it. We build it into Discourse and turn it on by default for hosted Pro and Business communities.
What we check
We check posts from new and low-trust accounts: by default, readers at trust level 0 or 1 who have made three posts or fewer. We skip your staff and your established members. If a new account edits a post within 24 hours, we check it again, so a spammer can't post something harmless and add the links later.
What the AI reads
For each post, we give the model the post text, up to three images, the topic and category, the post it replies to, and the account's age, email address, post count, and rough location. The model decides whether the post is spam and explains why.
What happens to spam
When the model marks a post as spam, we hide the post, silence the account that wrote it, and send it to your review queue. If the spammer started a new topic, we hide the whole topic. Your moderators restore anything we got wrong in one click, and the account can post again.
What you control
You add your own rules in natural language, and even choose which trust levels we check. Hosted customers use the language model we host for them, while self-hosted teams can connect their own.
AI translations
Host one conversation in every language
You publish for readers around the world, but readers who don't speak the language of a topic scroll past the comments. Few of them reply, so each discussion only hears from part of your audience.
With AI translations, we translate every comment into each language you choose to support and show each reader the version in their own language. A reader in Madrid and a reader in Tokyo read and reply to the same topic, each in their own language.
What readers see
Signed-in readers see each topic in the language set in their account, and visitors pick theirs from the language switcher. Readers see a small marker next to any post that started in another language, and they can tap it to read the original. Readers who speak several languages can list them, and we leave posts in those languages as written.
How translations stay current
We translate each post once per language and save the result, so readers never wait for a translation to load. When an author edits a post, we translate it again.
How search engines find it
We give search engines a version of each topic in every language you support, tagged so Google can match each version to its language. Someone searching in German can find a topic that started in English.
What you control
You choose the languages, the categories we translate, and how many older posts we translate each hour. You can write your own translation instructions to keep your house terms and tone consistent across languages. Hosted communities can support two languages on Free, four on Pro, and up to ten on Enterprise.