For a few years now I have been seeing the numbers drop on about a dozen of the Discourse fora I follow - some I have actually stopped checking myself because the activity got so low. My gut feel is that people are moving away from fora . .
Reminds me of something we discussed some time ago:
Right, so far it seems this is being called AIO (AI optimization) and GEO (generative engine optimization)
Hi Helga, this is my personal opinion but the issue of a support forum is the “cue” the unknown wait for a response deters a userbase. Most Saas services offer support tickets instead of a community because of that regard and emphasize heavily on an alert base notification for submission as users can indicate a level of priority i.e. low/high/critical which a community forum doesn’t have.
AI adoption has curved the flow and the depth of a response in quick time which a large userbase are finding most useful, however it still lacks the persona that one gets from others who’ve faced similar issues, the conversational factor tends to ensure peace of mind in a user, so a support forum will still remain valid.
A suggestion if I may, consider with your team, creating an active digital support agent that does query the base information that’s required in small cases either through discourse ai bot or using an mcp, and a remote agent for instance a response could be:
Hi Helga, I got alerted of the issue you just shared with CompanyName I’ve got a log of similar cases and can possible give you some suggestions and notify CompanyName team right away if you’d like, CompanyName team can also scheule a one on one conversation just let me know how urgent the matter is.
This is all a suggestion, given I’m not aware of what type of company you are a part of. And how you feel about automated responses. Hope it all works out for you.
BTW follow my topic, I will share some support integrations I’m working on that would enhance any self hosted discourse forum.
Hello ![]()
Personally, if I were running a support forum for a specific product or SaaS, I would seriously consider putting the entire forum behind a login requirement at this point.
Here is my logic:
The main SEO and organic traffic should be driven by the product’s official website and documentation anyway. The forum’s primary job is to serve actual customers. By making the forum private:
- You block AI scrapers and search engines from indexing your community’s collective knowledge for free, meaning users must log in and join your space to get those specific answers.
- It creates an exclusive environment where paying customers feel safer sharing screenshots, logs, or specific setups without them being indexed publicly on Google.
Of course, this might increase the load on official support channels for basic questions that could have been resolved via a quick Google search, but as AI keeps stealing forum content to answer questions on third-party interfaces, forcing a login might be the only way to keep your community active and protect your data.
I also experience a huge decrease but I have no product so not the same thing. I try to find a way to fix this. ![]()
The unfortunate downside is that you start losing some influence in search engines and AI this way… with Discourse for example, if someone asked “can I build a forum that does X”, it’s likely an answer will be available via Meta.
Without that sort of content being indexed, someone might end up with a conclusion that points to a competitor simply because their answer is more readily available (which has been the SEO battle before AI).
So if you’re not careful, instead of people finding answers about your offering via an AI discussion and skipping your website… you might find that what you’re offering isn’t relevant enough to be mentioned at all.
It’s a conundrum! At the core of it “what can your forum give me that other places can not” has always been a good question, but at least without direct participation you could get an idea of how useful content was by measuring visits.
With AI capturing much more of that non-participatory traffic, emphasis kind of has to shift away from measuring it as value and towards what brings people back to participate directly. Building a reputation that carries some influence, seeing your feature request get implemented, sharing something you’ve made, helping someone get unstuck in very specific situations… there’s still some stuff that AI can’t provide, and maybe we can find some more to lean on.
This is such a sharp point. It touches on what is probably the biggest dilemma for community builders today.
We are trying to play a new game with old rules, and if we don’t adapt, we’ve already lost.
My approach (few months interval, niche, hobby forum): If you can’t beat them, feed them (and get cited)
Personally, I’ve decided not to fight this wave. I have specifically optimized my own forum to be as AI-friendly and AI-compatible as possible.
Instead of hiding my content behind walls for AI, my goal is to make it incredibly easy for AI crawlers to parse, understand, and trust our data. I am actively positioning the forum so that whenever an AI compiles an answer about our niche, our links are the ones it cites as the authoritative source. If the AI is going to answer the question anyway, I want to make sure my platform is the one getting the citation and the referral traffic.
Honestly, so far, not much success…
We are unfortunately seeing fewer and fewer visitors coming in currently.
I don’t think the issue is that AI models aren’t citing us, they actually do, quite often.
In my opinion, the average AI user has zero incentive to click through to the source. If they need more details, they won’t visit the website; they will just ask the AI to elaborate. For most users, simply seeing that the AI’s answer is backed by a real, quoted website is enough to satisfy their need for credibility but they stay inside the chat.
This is a massive issue because even if we “win” the citation game, we still lose the actual visitor.
Still, I’m giving it some more time to see if things turn around. I’d rather try to play along with this new world than fight it.
The SEO/privacy trade-off isn’t new. Publishing publicly has always meant making content discoverable and so freely available in exchange for reach.
Google used embeddings 10 years ago to help map content. AI leverages the same technique but having been trained on large corpuses can also regurgitate content without users having to visit the source.
As a result, the unique value of forums shifts away from being repositories of facts and toward being places for conversation, trust, shared experience, and community.
I trust that people still want to “get it from the horse’s mouth” and are still social animals.
Talking to an AI is useful but a spiritual dead end.
It depends on what you consider success, though, eh?
How are you thinking about that aspect of things at this point?
Everything else you’re doing so far still sounds pretty compelling to me.
That’s a very good question. If I step back, it’s clear that the definition of “success” has fundamentally changed for us.
To give you some context: years ago, we used to get nearly 50 new registrations (now mostly under 20) a day, and we’ve reached almost 30,000 members over the past 6 years. Out of those thousands of people, a dedicated core group naturally formed, the team that keeps the site alive and remains active in the community to this day.
For me, the number one success is when this core group expands, and it’s also a success when a great, helpful conversation sparks on the forum, allowing us to discuss deep, technical topics in a genuinely warm and friendly atmosphere.
What breaks my heart, though, is that the people who get stuck in the AI chat are completely locked out of this experience.
There is a very important -if not the most important- difference between Google (or search engines) and AI.
When a person goes to a search engine, they know exactly and basically do so with the goal of clicking through to the source! Someone who goes to an AI chat does so to get the information right then and there!
All they see is what the AI tells them, if it even cites an example: “According to someone on XY, you should do this and that…” If the user just wants a quick fix, they won’t go any further. They will never step inside our forum to actually experience what would eventually make them stay, become active, and in time, become part of that core group I mentioned at the beginning.
Sure, “brand awareness” is probably there. If the AI mentions us often enough, people might remember the name, and if they come across us elsewhere, it will click: Ah, I’ve heard of them…. But that is not a community yet.
I think true success in this new era would be a win-win scenario where the AI actively encourages users to click through to the source for the human element.
If the AI offered something to the user like: “For the full discussion and practical advice from the community, click here,” it would keep the entire ecosystem healthy. By directing users back to us, over time they too could enrich the forum with real, new human experiences, which in turn would continuously feed and train the AI.
Since we have no direct influence over how LLMs choose to display links, I recently started experimenting with a practical workaround: I built a plugin that, among other things, generates multilingual llms.txt and llms-full.txt on my forum, in which I formulated guidelines specifically for AI crawlers, leveraging the fact that AI models generally try to avoid safety liabilities in niche, highly regulated, or technical DIY topics.
I am very curious to see if prompting the LLMs directly through llms.txt will have any “psychological” effect on how they structure their answers and citations. It’s a small experiment, but at this point, I’d rather try every avenue to bridge the gap between AI-based search and real human communities.
I discussed this with a customer recently. They asked for advice because the change in search behaviour is having a detrimental effect on their success metrics. They want to feed agentic search with legit and up to date information about their product, but ungating it has tanked their sign up rates. Here is an abridged version of my thoughts.
I believe un-gating is the right move if you are optimising for growth, in which case you should be measuring and gauging success based on reach.
Have you read Before You "Build a Community," Decide: Library or Coffee Shop??
I would take this as an opportunity to change how you measure and communicate ROI. If you stop thinking of your community as a coffee shop and start pitching it as a coffee shop inside a very successful library, you might be able to paint a different picture.
If you take that approach you can start reporting the community as a funnel, measuring reach (search impressions) and value consumed (accepted answers viewed). Then you should normalise participation numbers when you report e.g. new contributors per 1,000 organic visitors. If you just report raw numbers they will be harder to interpret as you un-gate content because the denominator changes. If more anonymous people can consume content without registering, new registrations and MAU may drop even if the community is providing more value than it was.
We have Gated Topics in Category but I wouldn’t gate solved support content or evergreen troubleshooting topics. That is the content most likely to win SEO/AIO reach and deflect support, as well as building trust with new members.
Your observation that closing Announcements helped with traction is useful info to have. Based on that I’d gate product and company-value content like announcements, release notes, news, roadmap, Ask [company] Anything etc. You could try some other traditional engagement tactics in private categories too. Live AMAs, QA roundtables, career clinics, office hours, that kind of high intent programming that is worth signing up for (or at least enough to get people curious).
I completely agree that we need new metrics, that reframe is genuinely useful. My concern is that changing the dashboard alone doesn’t necessarily solve the underlying problem.
If AI increasingly satisfies users without them ever visiting the source, then “reach” and “community growth” begin to diverge. We may successfully build the world’s best library while finding fewer and fewer people ever walk into the coffee shop.
Measuring the library differently is useful. The harder question is how we reconnect the library to the coffee shop.
One thing I’ve realized is that the “Coffee Shop” isn’t only valuable for retaining experienced members. It also acts as a vital onboarding space.
Many new users are intimidated by highly technical discussions because they’re afraid of asking “stupid” questions in front of experts. Our off-topic category has always been the place where they could simply interact as people first. Once they realized that the most knowledgeable members were approachable and friendly, participating in technical discussions became much less intimidating. In that sense, the coffee shop doesn’t just retain the core community, it actively creates it.
That idea has also shaped how I’ve been building my own forum.
Rather than trying to fight AI, I’ve been experimenting with making the “Library” as AI-friendly as possible while preserving the unique value of the human experience.
One practical experiment has been building a plugin that trims topic content for anonymous visitors, while verified search engines and AI crawlers still receive the complete content for indexing and citation. The goal isn’t to hide information from AI or search engines. It’s to keep the Library fully discoverable while making the Coffee Shop itself something you only truly experience once you walk through the door.
What I’m really hoping for is an ecosystem where AI doesn’t just answer questions, but also recognizes when the real value lies in the human conversation and naturally points people toward it.
The mission of communities hasn’t changed. We still need to build great coffee shops inside our libraries. What has changed is the path people take to get there. Search engines naturally led people through the library before they reached the coffee shop. AI often satisfies them before they ever arrive.
If AI becomes the world’s librarian, our challenge isn’t to reinvent the coffee shop. It’s to make sure the librarian still points people toward the door.
I think that theme component uses JavaScript to “gate” the hidden content. My instinct is that AI crawlers would bypass that anyway, if they can access the website in the first place.
I would quite like there to be a plugin that really hides some content from unregistered users. Given what you said, maybe it could just show any solution post as well as the original post.
I concur, however:
That’s covered by long-standing Category permissions. Just use Category security features which have recently received some UI ![]()
As far as I can tell this allows for See/Reply/Create permissions per group.
Maybe I’m asking for extra permissions – “See replies” and “See solution”.
Yeah, don’t believe that’s possible
@Don’s solution at How I Revamped Our Forum's Visitor Flow might be a useful thing to note here.
