Blog: The Code of Making an Instagram Influencer
The fear of technological automation leading to mass unemployment has existed since technology began to be used in the workplace. And while research shows this fear has been largely inflated, it is still a concern that programmers and engineers must address when developing these technologies. Many industries, from healthcare to truck driving, have been discussed in the media as on the chopping block by AI’s hand. One industry, however, that has not been examined is the world of Instagram influencers.
I worked as a Social Media Marketer for approximately two months at a small e-commerce fashion startup. Among a few other tasks, my primary role was to take over their influencer outreach. I spent hours reaching out to hundreds of influencers, making passioned appeals as to why they should want to work with and promote our company. It was often a frustrating process as much of my outreach was fruitless. However, we did it because it works. Influencer marketing has become a standard in marketing an online business. For each dollar spent on influencer marketing, marketers see an average of $7.65 in earned media value returned. By 2020, the industry of influencer marketing is expected to grow to a 2.3 billion U.S. dollar business.
And it is by no means only a favorable market for companies: extremely high profile influencers like Kylie Jenner make an estimated $1 million for a single sponsored post. Even ‘micro-influencers’, those with around 10,000 to 50,000 followers, can expect a few thousand dollars for a post. This is a lucrative, and growing, business.
When I began programming and learning more about Artificial Intelligence, I never connected it to my past in social media. From experience, what makes influencers successful is a psychological brew of how credible they can be to their audience — the relatability of someone who shows seemingly every aspect of their daily life down to the food they eat, expertise on fashion, food, beauty, and halo effect, that attractive people are smarter and more likable. This intoxicating combination is deeply connected to our vulnerabilities as, at times, insecure human beings and seems so far from what a bot could attempt to replicate.
There have been some more artistic attempts at making influencer bots. The most well known example being an influencer known as Lil Miquela.
When Lil Miquela first came online in 2016, she garnered a lot of controversy with many speculating that she was real and some viewing her as part art project and social experiment. Her page, with over 1.5 million followers, shows an Instagram it-girl: pretty, well-dressed, and hanging out with your favorite celebrities. The only thing that separates her from the top influencers is that…well, she’s not real. The page raises questions of what is ‘real’ and ‘not real’ in the world of social media, but besides her image everything else is still made by a human.
What would it mean if everything on a Instagram page was tech generated? If the photos, captions, and following were all auto-generated all with the goal of gaining the biggest influence possible? This is exactly what Chris Buetti explored in his article “How I Eat For Free in NYC Using Python, Automation, Artificial Intelligence, and Instagram.” I have to be honest here and say that the first half of his title is what caused me to click on the article, but the latter half is what made me stay. Buetti created an entirely automated Instagram page that pulled pictures from around New York City and ended up amassing over 25,000 followers (at the time his article was written). The beauty of gaining this amount of followers was this allowed him to reach out to local restaurants and ask for a free meal in exchange for being featured on the page. The reaching out was automated, too, of course.
This article, to me, was genius on my screen. I’ve lived in New York for the past four years and over time have gone through waves of Seamless binges before my debit card cries out in exhaustion. The idea of having a seemingly endless supply of free meals at my disposal, all done through code, seemed inconceivable. And the best part was Buetti’s emphasis that anyone, with varying levels of automation, could do it for themselves.
Buetti used Python for most of his project. I assume Python works best for the project as it is a language that is most commonly used in Machine Learning .Over time, I hope to replicate his entire page, but I wanted to start somewhere I was comfortable, so I picked captions.
First, I created an array of captions and hashtags. Buetti’s list was probably a hundred or more, but I wanted to start small and see how it worked. Then I made a function that would iterate through this list and create unique captions with each hashtag.
There we go! I now have a very simple caption function that will randomly choose from my list and create unique captions I could use on my photos. Of course, I now need to incorporate giving credit to the page I get the photo from using REGEX and machine learning, but this is one step on the way to making my automated page.
Next steps are working on the Machine Learning Model that will select high quality images from Instagram and re-post them on the page. I hope to continue and end up with a fully functioning automated Instagram and eventually become an influencer that companies will want to work with.