I’m not going to start off this article like every other article out there by explaining what the “shadow ban” is because 1) it doesn’t exist and 2) if you’re here you know what it is already and 3) what you’re about to read below hasn’t been discussed yet on any of these other articles yet.
Here’s the answer guys.
Artificial Intelligence Search is to blame for the infamous Instagram “shadow ban”.
Since February, you’ve been furious with your engagement on Instagram and have read every shadow ban article out there, including, running your account through a “tool” to see if your posts have been deemed one of the unlucky many.
You’ve made your account private for a few days. You’ve switched up your hashtags, even stopped using so many hashtags in your posts’ descriptions (captions) or comments (worse advice ever- use them while you can!). You’ve just about given up hope on it all and have tried every trick/solution the Instagram Experts have told you to do. And yet nothing is working!
It’s driving us all nuts. Believe me, I know. It’s not possible to fix something that doesn’t even exist. But guess what: There is no such thing as the “shadow ban.”
Here’s what’s going on… While you all were freaking out over the “shadow ban” phenomena, Facebook introduced it’s artificial intelligence machine learned search capabilities. It was released in early February and just like any new feature, they have been A/B testing it over the past few months.
Before February, Facebook had been making big strides towards this advancement. Here’s some of the history:
Back in April 2013, Instagram moved all of it’s data into Facebook’s data center, which means, in order for Instagram’s infrastructure to rely on the same machines and data storage as Facebook’s, they had to adapt to Facebook’s infrastructure. It also means that any major changes to Facebook’s search infrastructure need to be adopted by Instagram in order to keep internal resources under control.
On April 5, 2016 Facebook announced a big step in photo recognition processes called automatic alternative text. This gives a machine the ability to read objects in your photos and categorize your photos from the machine learned information. This was just the start of something BIG Facebook was working on.
Now, let’s look at Instagram’s platform and the areas that an account or post/content can be discovered, or searched, and how they are populated currently:
The areas to focus on in this article are 3, 4 and 5. AKA, where people have experienced issues.
Let’s break this down some more. When you load your photo, Instagram captures metadata about that image.
“For example, Metadata can describe how, when and by whom a piece of User Content was collected and how that content is formatted.” – Instagram.com
Instagram can read that your image was taken with a Canon camera vs. an iPhone. Instagram can NOW read the objects and orientation of those objects in the photo, too.
Instagram has repeatedly stated that you need to focus on content, right? They weren’t kidding. Now it’s more important than EVER to use the right content and have quality photos.
Straight from TIME:
“Facebook says the new tech will improve its user experience in two ways. First, it’ll make it possible to search Facebook for photos based on what’s in them, rather than just by date taken, tags, or location. If you’re trying to find a photo of a paella dish you cooked last year, for example, you’ll be able to simply type “paella” in the Facebook search bar. This, Facebook hopes, will help its users quickly find images without having to remember when they were taken or how they were tagged.”
First off, you need to know that Facebook’s computer vision platform is called Lumos.
Now, let’s look at this photo recognition technology in action. If you load a photo of a pregnancy announcement as this example displays, Lumos only detects the objects: 3 People, People Smiling, People Standing, 1 Tree (though there are many), Outdoor, and. Yet you are likely using tags that refer to an entirely different category: for example, #babybump, #pregnant, #pregnancyannouncement, etc. Instagram isn’t going to show your post under that specific hashtag. I don’t think Instagram is testing this on all “categories” aka hashtags, but select ones for now that relate to the objects Lumos can identify currently. And if history repeats itself, it’s A/B testing by region too, usually, Australia first.
Ultimately, I predict that the manual hashtag tagging and search capabilities will disappear altogether in the coming months (just like the fate of the Map Search on the profile and Trending on the Explore page did last year) and we will find the search on Instagram to be identical to Facebook’s.
They look at the meta data of the photo, the objects in the photo, orientation of those objects and the photo description/caption (spruce up on your keywords, friends!) mainly for classification in search. If your description/caption and keywords for the photo don’t match the content on the photo, then Facebook and Instagram aren’t going to weigh it high in search, and may not show it all. To illustrate, the above posted on Facebook with a description with “announcement.” Lumos deemed the current object keywords, did not show up in a search for “pregnancy announcement.” However, once the description was edited to reflect “pregnancy announcement”, it immediately showed up in the image search on Facebook. Same applies for Instagram as they roll out the changes and/or A/B test it.
Your post(s) may be some of the few to be lucky enough to be tested in this new test environment. Here is why in my mind you are considered LUCKY now, even though your posts might be “tanking” and you’re feeling frustration.
You have a leg up against everyone else on figuring out how the artificial intelligence’s machine learned search capabilities will apply to your account’s content and niche NOW. You can work on finding that delicate balance between content and PRO (Photo Recognition Optimization – if you want to call it that – I just made that up) of the image directly. Someone without the tests being done on their account can’t A/B test easily to figure it out without getting super nerdy and techy.
So get to A/B testing my friends. Take the easy route while you can and you’ll be well ahead of it all when they officially roll it out across the entire platform while everyone else is scrambling.
*We will have a dedicated thread in the Instagram Posse for those experiencing it to A/B test with each other. Because testing alone for something like this without bouncing ideas and hypothesis off each other is, well, rather lonely.
Join Our Free Resources Area and get Guide 01 of Instagram Decoded FREE!
Now, here’s what you SHOULD be focusing on to work with the changes:
1). Great content. I know many people don’t want to hear this because they think it’s just fluff. Now you know, it’s not. Instagram even spelled this out for everyone back in their post about the “hashtag” issues in late February. “Make sure the imagery AND copy latter up to your main goal”. Think about composition of the photo, clarity of the photo (the better the image, the easier Lumos can decipher what’s in it to categorize it best), keeping your photos, branding and objects in niche. Don’t post a flat lay of a beer can and a pacifier together. Post a flat lay of a beer can and a cuzzi together with your logo on it. Post a photo of a pacifier and a rattle together.
Below are some examples of photos where Lumos is picking up a bunch of objects (which would result in more visibility/categorization opportunities). Click on the photo to view the full Instagram post and look for yourself if the objects defined in the picture by Facebook’s aligns with the posts’ descriptions and hashtags or not.
Here is an example of two photos Lumos can’t read any objects in. (Click the image to see the original Instagram post).
I believe this Corgi photo is not being categorized because Lumos does not recognize two ears. The dog is mostly covered in clothes/blanket obscuring it to the AI. So without 80% certainty, it’s not sure and so it won’t categorize it. Lumos does not show a category if it doesn’t have 80% certainty.
2). Provide a description that incorporates the right SEO/keywords that apply to the photo. If you need help with this, consider giving WordStream’s Free Keyword Tool a whirl. Type in a keyword you think applies and see what it suggests to use. (Fun Fact: I used to work for WordStream! It was my favorite company I ever worked for!) Here’s a simple example:, don’t talk about your girlfriend with the nickname of “ ” on a photo of your girlfriend’s face. If Facebook’s Lumos doesn’t see an actual baby in your photo, guess what, you’re “SHADOW BANNED” Dun dun dun. More like the post is just categorized incorrectly or not shown at all in the search because it saw you miscategorized the post or ABUSED the hashtag (current testing in our Niche Instagram Facebook Groups).
We’ve recently created the niche Facebook groups and have already been able to assist many in getting their photos to show up in hashtag search again. There have been multiple reasons why posts haven’t been shown already uncovered!
3). Continue using all 30 hashtags until they are phased out (Umm, HELLO. Instagram tells you to do this. I don’t know why you would consider doing something that is the opposite of what they tell you to do). Don’t pass up the opportunity to still be found on Instagram based on your own categorizations vs. machine learned categorizations.
4). Create unique content. Don’t replicate other content you see on Instagram. Create content that stands out. They can now tell if the images are very similar and will downgrade them in the search results. For instance, the 50 billion shots of someone in bed with a cup of coffee, a book, legs and socks and birds-eye camera shot will not win you any publicity. How overdone is this angle/picture/concept?!
OK, I gave you the main information you need to stop having daily nightmares about the shadow ban and to get working on your strategy for moving forward. So, get on it, POSSE!
And don’t forget to join your niche’s Faccebook group to join in on the A/B testing per niche. Because every niche will definitely be vastly different and need it’s own PRO strategy for this. Band together with your other niche influencers to facilitate A/B testing to nail this down faster. (For instance, a makeup influencer is going to have a really hard time figuring out how to tell Instagram/Facebook’s Lumos to look for green eyeshadow), when, eyeshadow may not even be an object Lumos recognizes yet. Fun fact- it does understand “selfie” though!
We’ll also work on compiling a list of known “objects” recognized by Lumos as we see them come in through in our niche A/B testing in the group. Ideally, we’ll compile a list of objects and how Instagram/Facebook categorizes them as well.
For instance, does a piece of asparagus get categorized as “” or “plant”. Would you rather find out from someone else’s previous test or waste a post of your own trying to figure that one out? Obviously, this will take a ton of experimentation and community coordination, but that’s what we’re all about! 🙂 Community over competition, right?!
Even with the recent changes, Instagram is still a layered approach strategy for growth and engagement. It will evolve over time and for now, our strategies and experiences outlined in our Instagram Decoded guides still are in line with recent changes (including the algorithm changes). We teach you how to grow your account with real followers that pertain to your niche and goals. Check out the success and testimonials from some of our over 2,000 Instagram Decoded purchasers have to say!
For those that want to geek out further on this because you find it really interesting, Part 2 of this post will be coming out shortly and announced in the Facebook group, Instagram Posse. The next post delves deeper into some more history on Facebook’s evolving photo recognition functionalities and some tools I found helpful during my own A/B testing phases to figure this all out… Facebook has been telling us this was coming for over a year and we all ignored it!
About the Author:
Liz Dean leads an active Facebook group called Instagram Posse, which is home to over 22,000 Instagram Influencers and Marketers. Liz comes from a background of software engineering and client support. She’s a total tech geek at heart with a side of social community lovin’! Liz also teaches online courses for Instagrammers, Brands, Companies and Marketers on both Instagram and Facebook Groups.