3 Lessons Learned from HIV.gov’s Chatbot Pilot

Content From: HIV.govPublished: August 28, 20184 min read


Graphic of a hand holding a phone with a chat going on.

It's been one month since we launched our first Facebook chatbot—a tool that can simulate conversation to deliver tailored content to your audiences. We are always excited to try emerging digital tools because our experiences help us support our partners (including you!) to experiment with innovative ways of communicating with their audiences.

If you're considering using a chatbot for HIV/AIDS communication, we hope this blog post will help you prepare for your launch.

Before we dive into the details, if you need a refresher on what chatbots are and how they can be used for HIV/AIDS communication, you can read our previous blog post.

Why a chatbot?

As we discussed in that post, the use of messaging platforms has exploded. While social media timelines and email inboxes are often congested, messaging platforms, like a chatbot sent via Facebook Messenger or hosted on a website, are often not as crowded.

Another  feature of chatbots is that they can be personalized. Depending on someone's interest and responses, we can tailor our content so that we are sending our subscribers the information they want—easily and quickly.

How we used it

We launched our chatbot before the International AIDS Conference (AIDS 2018). We planned to offer extensive conference coverage, including blog posts and Facebook Live streams, and we thought this would be a great time to try a new platform for sharing.

There are many chatbot options and tools and we chose one that integrated with Facebook Messenger. We knew the audience we were trying to reach uses Facebook and would most likely use Messenger too.

But, like the old adage about a tree falling in the woods, if we sent chatbot content but didn't have any subscribers, no one would receive it. So before the launch, we encouraged our users to subscribe by promoting our chatbot plan across our social media channels, emails, and website/blog content.

Once we began building our audience, we shared background information about the conference and the guests we planned to interview in our Facebook Live streams. We asked our subscribers about their preferences; for example, did they want to receive reminders before each Facebook Live stream?

That question, in particular, served two purposes:

  1. It let us know if anyone was actually interested in Facebook Live reminders (they were); and
  2. It enabled us to tailor our content – we were able to "tag" or save subscribers' content preferences and thus we only sent them the content they actually wanted to receive. For example, we only sent Facebook Live reminders to people who expressed interest.

In addition to Facebook Live reminders, we also sent exclusive behind-the-scenes content, like short video clips and photos, from AIDS 2018. We also sent messages when we posted breaking news from our blog.

How did we do?

We weren't sure what to expect when we launched the chatbot. We were pleasantly surprised with our open and click rates: we had an average of 90% open rate and a 27% click through rate. We often sent multiple messages per day, and our engagement rates stayed stable throughout the week-long conference.

Lessons learned

It's safe to say we learned a lot during our pilot. The three most important lessons were:

  1. Chatbots can help us be more responsive to our audiences. What makes the chatbot different from our other platforms is its ability to ask our audience about their communication preferences. This means we don't have to create or send "one-size-fits-all" content—we can tailor content based on audience responses. Tagging is a critical piece of this.
  2. It's supposed to be a conversation, not a one-way blast. Writing for chatbots is a bit different than writing standard social media posts or email blasts. It's important to make sure the content is conversational—we tried to ask a question in every other message. There's a learning curve, and our team found it helpful to read the conversation out loud as we were writing to ensure that our messages sounded like a real conversation.
  3. Mix up text-only content with links, photos, and short video clips. Get creative! We had fun sending different types of content, including short videos. We also shared links to our other social media channels and blog. Tip: Be sure you understand the parameters of your chatbot tool and crop photos and video as needed.

Our team plans to share updates during the upcoming 2018 United States Conference on AIDS (USCA). Subscribe to our chatbot todayExit Disclaimer to follow along.

Are you interested in learning more about chatbots? Do you need help setting up your own chatbot? Sign up for Virtual Office Hours for a one-on-one coaching session from a member of our team.