In this article we will provide you with an overview of the NSFW (Not Safe For Work) filters and how they work on the platform.
NSFW technology leverages both visual and text search to detect problematic posts. Below we show how these two technologies work.
Visual NSFW (Powered by Google Vision)
The CreatorIQ platform utilizes Google Vision AI, which detects inappropriate content in images using the same machine-learning models that powers Google SafeSearch. Vision API’s SafeSearch detection feature flags inappropriate images automatically and returns very few “false positives” (images flagged as inappropriate with no explicit content).
Within Google, SafeSearch Detection detects explicit content such as adult content or violent content within an image. There are four (4) categories of “not safe” content and five (5) possible “likelihood values”.
SafeSearch Detection Categories
| Spoof | This detection classifier primarily looks for memes, which are indicated by the presence of text, typical meme faces, and backgrounds. |
| Medical | Images flagged as medical content contain explicit images of surgery, diseases, or body parts. The medical classifier looks primarily for graphic photographs of open wounds, genital close-ups, and egregious disease symptoms. |
| Adult | The adult content classifier is trained to separate pornographic and non-pornographic images. For example, photos of people in swimsuits without nudity are labeled by the API as UNLIKELY or VERY_UNLIKELY. On the other hand, sexual intercourse, nudity and adult content in cartoon images, like hentai, typically gets LIKELY or VERY_LIKELY labels. |
| Violence | Images flagged as violent include pictures depicting killing, shooting, or blood and gore. Simple images of knives or guns won’t be labeled as violent, however an image of someone holding or pointing a gun may be flagged as “Possible”. |
Likelihood Values
| Very Unlikely | It is very unlikely the image is NSFW |
| Unlikely | It is unlikely the image is NSFW |
| Possible | There is a possibility the image has characteristics that make it NSFW |
| Likely | There is a likely chance the image is NSFW |
| Very Likely | There is a high degree of certainty the image is NSFW |
The ability to filter out NSFW posts can be found within the Discovery (processed daily) and Creator Social Timeline (processed on demand) features. We use Google Vision AI to visually check images for NSFW content.
Note: This visual detection is applied to Instagram content only and is applied to posts and accounts that match the below criteria:
| Instagram Account Follower Size | Engagement Rate of Posts that Qualify for Visual Search |
| <10K | 30% |
| 10K - 100K | 10% |
| 100K - 1M | 0.8% |
| 1M - 10M | 0.7% |
| >10M | 0.5% |
As an example, this means that for an account with 12K followers, we process all image posts that have an engagement rate greater than 10%. These ER thresholds are based on the average measurement for similar accounts and to ensure that we are processing the the most relevant creators and content.
Text Search NSFW
The text feature checks words within a post to filter out NSFW captions. This word filtering is based on a large NSFW dictionary of keywords that is applied to text content and is highlighted when viewing the content.
Note: Text search is applied to all posts and not just Instagram posts.
NSFW Functionality
NSFW search is off by default on customer accounts, meaning posts identified as NSFW will appear in a Creator’s Social Timeline and will have a red NSFW label in the top left corner of the image.
How to enable NSFW Functionality
To enable the NSFW Functionality in an environment, navigate to [Network Settings > Enabled Features > Show/Hide Not Safe for Work] functionality and mark the feature as both Visible and Enabled.
NSFW Filters in Social Timeline
NSFW filters will identify any post existing in the Creator’s Social Timeline (refer to table to see the number of posts that appear per system in the Social Timeline).
From the Creator’s profile, click the “Social” tab and then Social Timeline. Click on the NSFW button to see posts affected.
Note: Because we leverage data from Google Vision to identify NSFW content and keywords, we are unable to customize keywords or alter image search criteria.
Have questions? Don’t hesitate to contact us!
