Community Central
Community Central

You may have recently heard of a feature called Quick Answers that we began rolling out in earnest back in mid-August, after a few months of testing over the summer. There were some pretty clear bumps along the way with the initial AI-generated content, and one of the biggest pieces of feedback we got from the community was that the ability for the community to edit Quick Answers content—something that we had planned to deliver later on—was wanted and needed sooner rather than later.

We can officially announce that edit options for Quick Answers are going to be made available to admins, thread moderators, content moderators, and users with rollback rights. This could potentially be before the end of October, prior to any further significant rollouts of AI-driven Quick Answers content (there will be a couple of additional experimental releases ahead of time, which we’ll go into more detail later in the blog).

Today’s blog will take you through what the Quick Answers feature is (for those unfamiliar so far), how and why we’re leveraging Generative AI to help scale the feature, and what the initial editing options will look like once they’re rolled out.

What are Quick Answers?[]

The Quick Answers module as it appears on the Korra page on Avatar Wiki.

Quick Answers is a new feature that we’ve been experimenting with for several months, after initially teasing it back in April at Community Connect as part of our “Canonical Pages: The Ultimate Portal to the World of Fandom” session.

The concept of Quick Answers is pretty straightforward: to format information from wiki pages into a Q&A (questions & answers) format that would appear as a module on the wiki page it’s based on. These modules would be directly relevant to, and answer questions about, the subject from the page. For example, the image on the right (representing what’s currently visible only to logged out users) is from the Korra page on Avatar Wiki with questions and answers specifically about the character of Korra.

The goal of this feature is to have the content formatted in such a way that Google can easily pick it up and digest it in a way that provides value to readers on Google’s Search Engine Results Pages (SERPs), thereby enhancing the page’s SEO value to Google, which in turns makes it easier for readers to discover your pages on Google.

Of course, having something like this live across the site at a meaningful enough scale that Google elevates the Fandom platform even further would require a lot of manual work. At least, it would have prior to the advent of Generative AI in late 2022 and throughout 2023. Previous attempts at features that required significant manual content creation generally failed, so we decided that, to achieve the scale needed to have an impact, we would leverage Generative AI to create the initial content.

How is Generative AI being used?[]

Over the summer, we began experimenting with Generative AI models, via a GenAI partner, to leverage information from Fandom pages to create an initial set of answers to questions that were pulled from related Google searches. The questions and answers that were live during the experiment were reviewed by Fandom staff for accuracy, and then posted once they were reviewed.

Harry Potter Wiki’s Quick Answers appearing on a Google Search Engine Results Page (SERP).

This experiment showed promising results, including a bump in traffic to pages that had Quick Answers on them. You can see in the image on the right how that was starting to get picked up by Google’s SERPs. In the case on the right, using the Albus Dumbledore page on Harry Potter Wiki, Google even included them in the Q&A format that was displayed on the page itself.

Once we saw the successful results, we decided to prioritize scaling this feature out even further, with a set of questions and answers generated for 6,500 character pages across the platform. The goal of this rollout, which happened in mid-August, included live testing for the AI model’s accuracy as well as seeing how the rollout impacted traffic numbers on an even bigger scale. Once released, we saw the traffic bump we were looking for.

Unfortunately, testing in the wild also showed serious accuracy issues. We fielded a number of reports from the community about not just accuracy, but also the appropriateness of some of the answers. We opted to turn off Quick Answers across the platform so we could assess the situation and figure out our next steps. A review of the issues showed us that while many if not most of the questions and answers were accurate, the inaccuracies that we saw were glaring and at times inappropriate enough that it was clear that the initial method of generating the questions and answers was not the right approach.

So what was that method? Our initial method, “method A,” pulled the questions from Google and generated answers based not just on the wiki page the Quick Answers would appear on, but also from Discussions posts and other types of content. Given that opinion or otherwise subjective content was being used, that skewed many of the answers. Fortunately, we had already been working on a different model, “method B,” where both the questions and the answers were generated directly from the wiki pages themselves. As we generated that content, method B’s approach improved the accuracy of the answers.

From there, we re-released the initial test Quick Answers from earlier in the summer and a handful generated from method B. Testing and learning is ultimately part of feature development, so it’s not the worst thing in the world that this happened. We tried something, we learned from it (both what went right and what went wrong), we adapted, and we tried something else. And we’re continuing to fine-tune the model to work on improving it even further.

So, in addition to feedback about editors wanting to be able to edit the content, we also decided to add this to help with accuracy. Having you involved will only make the content better, since you’re even closer to the content and information than Google is and can recognize what the best, most impactful questions and answers are better and faster than Google can. Using AI allows us to scale the feature and take the initial content creation burden off of the community, but we know that you’ll be able to make it even better. We had been planning to do this later on down the line, but given the first impression made with method A, it made sense to move it up.

How can communities edit Quick Answers?[]

Knowing that, once we rolled back the prior method A release, we immediately began discussing what a community editing feature could look like. There were a few different options that we ran through, but here is the gist of what we decided:

  • Admins, thread moderators, content moderators, and users with rollback rights will be able to edit the questions and answers within Quick Answers on a wiki. We chose to add the option for editors with these user rights since, at least for now, we largely consider this a moderator and a content review feature. That’s not to say it can’t be opened up further in the future, but this felt right as a starting point.
    • Fandom Staff, Wiki Representatives, Wiki Specialists, and SOAP members will also be able to edit the content in Quick Answers.
    • The central page became a necessary component because Quick Answers will not appear on the page they’re about for logged in users, at least not yet. We wanted to keep the same experience for editors that they have now, though in the future we can consider revisiting this.
  • Each question and answer can be edited individually via a central listing page that we’ll be adding, which will be accessible via the Admin Dashboard. I’ll get into the details of that central listing page in a few moments.
  • Questions will each have a 100 character limit, and answers will have a 1,000 character limit. As a follow up feature (so this won’t be in the initial release), editors with those user rights will be able to remove questions as well as add entirely new questions. That way if there’s a question you don’t like, you can swap it out for a new one.
  • Users who aren’t in those user rights groups will be able to report questions and answers that they think are inaccurate. From there, the reports can be reviewed and, if needed, actioned upon in the central listing page.

So what will that central landing page look like? Great question! Here are the details:

  • You’ll see a list of all questions and answers present on the wiki, and there will be an editing entry point for each one so you can make changes as needed. The edit mode will be presented on the same page, so you don’t have to change to another page to edit.
  • You’ll be able to sort the list of questions and answers in a few different ways:
    • You’ll be able to sort by reported Quick Answers, so you can more easily find which pages need more immediate attention.
    • You’ll be able to sort the list by Quick Answers creation date, so you can review new Quick Answers you haven’t seen yet
    • You’ll be able to sort the list by Quick Answers edit date, so you can see which questions and answers have been altered by other editors
    • You’ll be able to sort the list alphabetically by page title, in case you’re looking for a specific title that starts with a specific letter. Note that the default sorting will be A-Z.
  • The central page will display the following information for each Quick Answers:
    • The article it’s associated with.
    • Whether it’s currently reported.
    • An option to add another report, in case an editor with user rights does not think a particular question or answer is okay and wants to signal that to others instead of making an edit directly.
    • A button to dismiss reports. Here’s a few specifics on how that will work:
      • When you edit a flagged question or answer and save your changes, the flags on that Quick Answers automatically are resolved and no longer shows as reported to other editors.
      • Users who reported a Quick Answers are able to report it again once a new version of it is saved, since reports are bound to specific versions.
      • The dismiss button resolved all reports on the current version of a Quick Answers at once.

Just as important as what’s included in these initial editing tools are what we’re not able to include right now. The following features are out of scope for this first release:

  • Quick Answers will not be displayed for logged in users on individual article pages, nor will there be editing entry points on individual article pages. The central listing page solves for this.
  • Quick Answers editing history or version control.
  • Text formatting, linking, or inserting any kind of media.
  • The ability for editors to add Quick Answers to wikis or pages that don’t currently have any.
  • Non-English translations, as we’re focused on English language content for now.

That’s not to say that none of these features can come later, but given available development resources and other items on the User Generated Content (UGC) team’s fourth quarter roadmap (focusing on editor retention), this is what we were able to commit to for now so we could get these tools out there for you as soon as possible.

What’s next for Quick Answers?[]

The UGC team has already started working on the Quick Answers editing tools, which we hope to release potentially before the end of October (though that could change, of course). We’ll have another announcement closer to the release of the tools, and we’ll keep you posted on our progress in the Technical Update blogs we post every other week (the next one is scheduled for October 6th).

We’re also going to continue experimenting with refined methods for AI generation of Quick Answers content. In the next couple of weeks, we’ll be releasing Quick Answers content on approximately 150 total character pages with questions generated off of Google keyword searches where pages are ranking lower, and with answers generated from the page the Quick Answers content is about.

Once the editing tools are released, and assuming we see success with the newer, more refined methods of AI content generation, we plan to begin releasing Quick Answers again more widely so we can all start seeing the benefits of better SEO and more readers on your wikis. We’ll be sure to let you all know when we have more information about that release.

In the meantime, if you have any questions, feel free to let us know in the comments, reach out to your community’s Wiki Representative if you have one, or contact our Customer Support team. We’re happy to answer any questions you may have!


BrandonRhea
Fandom Staff
Hey I'm Brandon, VP of Community at Fandom.
I'm a huge fan of Star Wars, Star Trek, and Marvel.
Want to stay up to date on the latest feature releases and news from Fandom?
Click here to follow the Fandom staff blog.

Click here to sign up for the From the Desk of Community email newsletter.

Want to get real-time access to fellow editors and staff?
Join our Official Discord server for registered editors!