How does Instagram censor hashtags?

Nicolas Suzor
4 min readSep 18, 2016
Cara Delevingne

Instagram comes under frequent criticism for the way that it blocks or removes content on the platform. Apart from removing individual images, Instgram frequently blocks entire hashtags — and its choices about which hashtags to block sometimes seem bizarre.

The site provides very little information about how it makes decisions about unacceptable content — and unsurprisingly, this often leads to worries that its systems could be biased or prejudiced.

Our research tracks how Instagram makes decisions about what content is available. As a first step, we focus on identifying which hashtags are blocked and how.

In this post, I just want to briefly show some of our initial observations. This is all work-in-progress, and these are very preliminary findings. They shouldn’t be relied upon — but I think they show that there are some really interesting questions to be asked about how Instagram moderates content at a hashtag level.

How are hashtags blocked?

We see that Instagram blocks hashtags in at least four different ways:

  • (semi-)Permanent blocks: some hashtags are very blocked. They return a 404 error page:
An example of a blocked hashtag page
  • New posts moderated: Some tags are apparently only temporarily blocked. For these hashtags, Instagram shows ‘top posts’ but not ‘new posts’:
An example of a moderated hashtag page.
  • “No posts yet” error message: for some tags, Instagram simply says that there are no relevant photos — but it’s clear that many photos have in fact been tagged. This one is strange; it’s not clear why Instagram does this.
An example of a hashtag with many tagged photos that returns a “No posts yet” error message.

It is hard to see any discernible pattern in the tags that we have seen that are falsely returning ‘No posts yet’ message. Here’s the most active hashtags that have been blocked in this way that we have found so far:

  • Content warning: the least restrictive category of blocking, Instagram attaches a click-through content warning to some posts:
An example of a content warning on a hashtag

Instagram seems to use a content advisory warning in three main cases: eating disorders, suicide or self-harm tags, and tags used to facilitate private sales of firearms:

Why are hashtags blocked?

When we start to look a bit deeper at what hashtags are blocked, it is difficult to work out what exactly Instagram is doing and why.

One of the first things we noticed is that there is little consistency in which tags are blocked. Most of the permanently blocked tags are keywords that are likely to be sexually explicit, very generic (#IG is banned, for example), or are offensive slurs. But some of the permanently banned tags are not consistently treated. For example, #Anorexia has a content warning, but #thinspiration and #proana are hard-blocked.

You can also see that the state of some hashtags changes quite rapidly. Here’s a small sample we have been following over the last two weeks:

You can see here that some hashtags change state quite quickly. Most of the other hashtags we have been following, however, do not change very often.

Next steps

We would love to hear from you if you are interested in this research. We will be tracking how Instagram censors hashtags to start to ask a few interesting questions.

  • Measuring bias: is there any truth in allegations that Instagram exhibits signs of systematic editorial bias against particular types of content? For example, does it disproportionately censor tags that might be used to promote positive body image, while not censoring tags that are used to shame womens’ bodies?
  • Tracking legitimacy: Instagram plays an important role in mediating how we communicate with each other. Like other social media platforms, many users believe that Instagram should at least enforce its rules about content fairly. We will try to identify how consistently Instagram’s rules are applied. Because Instagram blocks at a hashtag level, it will often block legitimate content in its efforts to block prohibited material. We try to understand how often this happens, how much of a problem it causes, and how mistakes are corrected when they are found.

If you are interested in talking about this research, or have ideas about some collaborations, please let us know!

--

--

Nicolas Suzor

I study the governance of the internet. Law Professor @QUTLaw and @QUTDMRC; Member of @OversightBoard. All views are my own. Author: Lawless (July 2019).