In the last couple of weeks I’ve been at #commscamp13 and #ukgc13 – at each there where plenty of people at the event and away from it who were tweeting using the hashtags – this sort of backchannel has become an important part of the event experience since the early days when people started appearing in conference halls with internet devices, a way to say “hello, I’m here” and “I just spoke to X – she’s amazing!” or “what Y just said reminded me of http://…” etc.  it also lets people “outside the room” join in to an extent.

On both of these recent occasions the channel has been disrupted by an invasion of spambots, to the point that the stream became unusable temporarily for those actually interested in the topic.

These bots latch onto the trending hashtag, saying something inane like “I can’t believe this!” plus a spammy link and flood the stream.

I know nothing about the creation of twitterbots, but I assume the accounts are programmatically created and they sit listening for trending hashtags and then throw themselves into the stream with the hope that from the large number of people taking part, there’ll be one or two who end up clicking on the link or even (yikes!) following the bot themselves.  Today I’ve seen them swarming around the #uksnow hashtag, the other day, when the trumpeter Kenny Ball died, his name was mentioned in spambot tweets.  It’s ridiculous.

At both the events recently, I’ve witnessed the bot tweets have died out over time, during the day, but it’s not clear whether this is simply because the tag became unusable and so stopped trending from non-bot-traffic or because Twitter noticed and did something (but see point 1. below) or because people following the stream reported the bots quickly.  I’m guessing it’s a combination.

So we’re all asking, “What can we do?”

0) Use another platform for the backchannel.  hmm…  any suggestions that have the ease of use and userbase of Twitter?

1) Get Twitter to sort it out.  Good luck with that one.  I’m sceptical of the idea that Twitter are actually able to do anything – the bots are all over the #sxsw stream right now for example.  Surely, if Twitter could kill this sort of thing easily, they would be doing it routinely on the tag associated with the biggest social media event of the year. Twitter also have shown themselves to be repeatedly clueless about what’s really happening on their network.

2) Build better filters – as Twitter increase the amount of metadata associated with tweets, perhaps there’ll be ways of identifying the little monsters quickly and removing them from searches.

3) Start a bot war – fight bots with bots. The most tempting, it’s kind of like point 1, but we do it rather than leave it to Twitter, but who’s the we that’s going to do this?  And like any war, it might remove the present symptoms, but does it just lead to retaliation and escalation?

Sorry, that’s all I’ve got right now.


6 thoughts on “TwitterBot-ageddon”

  1. Lloyd,

    good post. It did get a bit confusing as the official and unofficial Tweets combined. I was not sure whether #DigitalWomen and #WomenInDigital were discussions on virtual lady’s of the night.

  2. I noticed that the stream was particularly bad when I was following on Hootsuite but much less of a problem when using the actual Twitter App. This suggests that there is scope for competition between the Twitter “integrators” (for want of a better word) on quality of spam filtering, and this may actual create a useful competitive dynamic (re your 3 above). Hootsuite is dead to me for conferences now, but Twitter isn’t.

  3. Of course, if I am right, ie that twitter themselves are able to filter out the rubbish for users of their apps then it also implies that they are letting the unexpurgated spammy stream feed in to their competitors. In one way that is fair enough, I guess.

    1. Jonathan, thank you, I think this is to do with the metadata that they’re adding that’s talked about here:

      Looking at a search on in the browser, you get options for Top/All/People you follow – I assume that “Top” (the default) is the one that filters out the nasties and that it’s connected to the filtering_value attribute mentioned in the post above. I’m guessing then that folk like hootsuite haven’t yet integrated this new metadata to their searches, but that it is available and will hopefully be in new releases of the apps.

      But that’s a lot of assuming 🙂

Comments are closed.