Introduction and motivation
We live in the postmodern age of hyperreality. Vast majority of information most people receive through technological means without having any direct access to it’s source or any way to verify or deny what is written or shown on the screen.
A pretty girl sitting in a private jet might have paid a lot of money to fly somewhere warm or might have paid a company that keeps the plane on the tarmac a smaller amount of money to do a photo shoot. Her fans on Instagram have no idea which of these scenarios is true simply from seeing the picture of her relaxed in fancy seat, looking into distance through her sunglasses and sipping wine. That’s hyperreality. The same applies to supposedly very successful twenty-something guy in an expensive car that may or may not have been rented for a day to make an entire months worth of pictures.
In modern information society, attention is money. There are multiple ways to financially benefit from peoples attention through PPC advertisements, sponsorships, affiliate deals, promoting your own products or by simply selling your audiences attention to the highest bidder (influencer marketing). Some people have become millionaires or even billionaires by being able to play the social media game very well. Hyperreality is particularly prevalent on modern social media platforms that are deliberately engineered to extract as much time and attention (engagement) from it’s users as possible, then sell it through the integrated ad platform to the highest bidder. Objective truth be damned, Big Social executives will do anything they can to keep the people reading, clicking, watching. Large scale virtual reality as it was imagined by science fiction writers and technologists in 80s and 90s did not happen yet, but virtual social reality is already here. Lines betweeen true and fictional are increasingly getting blurry and muddled. In postmodern society, everyone is trying to influence the consensus reality to their benefit. Those able to do it efficiently and scalably emerge as winners.
If Big Social business objectives require, the organic reach is deliberately being limited to make more space for paid ads. Therefore promoting your brand on major social media platforms by playing it straight and relying on organic reach only works up to a point. It is still feasible on Tiktok, but less and less viable on Instagram and Facebook. The same applies to buying peoples attention on social media ad platforms such as Facebook. When social media ad spend is overall increasing auctions for human attention are getting more and more competitive, thus many business are finding that they are being priced out. In 2022, it is very easy to burn bunch of money on e.g. Facebook ads only to end up net negative by the end of the quarter.
So if we simply want to make some money, how do we proceed in this crazy world? If Big Social takes away the organic reach with one hand and charges more and more money for paid ads with another how do we turn a profit?
One of the feasible answers is growth hacking through gray hat social media automation. Let me elaborate what I mean by that. Growth hacking is a kind of marketing that focuses on low-cost, experimental, unorthodox ways to get new customers and retain them. Social media automation is the use of programmatic techniques and software tools to do things that someone would be doing manually. In the order of increasing sketchiness, we classify the automation techniques into three levels:
- White hat - fully legitimate and T&C-compliant automation, such as posting stuff on schedule through Buffer, ToS-compliant chatbots and so on. Legit stuff in the eyes of social media platform that nobody would legitimately complain about.
- Gray hat - automations that explicitly break T&C of social media platform. Somewhat sketchy stuff.
- Black hat - automations for malicious and criminal purposes, such as phishing, harassment or mass-reporting social media accounts on fraudulent basis. The evil stuff.
Assuming we don’t want to play the dangerous game of truly black hat automation on moral and risk management grounds, why do we pick the gray hat option? Why not go with white hat automation and do everything in entirely clean way? Well, frankly that’s because white hat automation has no teeth in 2022. We would be merely making an improvement over manual social media management, but not going far beyond that. Altough Big Social is proactively working against non-whitehat automation on technical and legal levels, gray hat automation at least gives us a fighting chance when trying to monetise these massives audiences on modern massive social networks.
Gray hat social media automation is not an easy thing to do in the post-Cambridge Analytica world, but may provide something like a cheat code to those trying to make money online. The idea is to turn the tables against Big Social by exploiting the very state of manipulative data-driven hyperreality that the Zuckerbergs of the world have created.
Like in object-orient programming, there are patterns - abstract principles of doing things that keep popping up when one works in the field or is researching it. We will discuss the patterns of gray hat social media automation that are more or less applicable to multiple platforms and can be thought of as theorethical concepts without scrutinizing how exactly they would be implemented.
But refrain from implementing some equivalent of FizzBuzz Enterprise Edition if you can.
Two major approaches
Broadly speaking, scraping HTML pages and submitting forms via POST requests is not how social media automation is done in modern times. There are two major technical approaches to implement gray hat automations for social media. Both of these must deal with automation countermeasures that platforms implement.
(Ab)using platform APIs
Another approach is running an entire browser and controlling it programmatically. In some cases browser is running in headless mode - without the GUI window. There’s several ways to do this. One is implementing basic bots that automation user could install as browser extensions. Another is using browser automation technologies such as Selenium, Playwright or Puppeteer to develop scripts implementing the automations. On surface this may seem to be easier than API-based automation, but it entails the risk of detection due to some tell-tale signs in the JS environment, as browser being automated will not behave exactly the same way as browser being user by a real user. InstaPy is a prominent open source project to implement IG automations by controlling the browser with Selenium.
Being able to quickly and automatically create social media accounts is highly beneficial for growth hacking. It entails automating registration flow with the help of proxies, captcha solvers and - if phone verification is requeired - SMS reception services such as SMSPVA or Valar SMS or your own setup, such as phone farm or product like Proxidize that can also be used to receive text messages for verification.
Scraping and data analysis
Social media scraping
Social media platforms can be scraped just like any other website, but it is generally more difficult due to automation countermeasures and many pages generally being behind login. It is advisable to take a look into Network tab in Chrome DevTools as sometimes information not visible on the frontend is available in API responses, such as email addresses for some business Instagram accounts.
Scraping social media pages can yield not only some contact information to be used later, but can also help to map out communities in graph-theorethical way. Speaking in graph theory terms, a social media profile is a vertex and connections between them can be either undirected edges (e.g. Facebook friends) or directed edges (one account following or mentioning another).
Social network analysis
From information gathered via social media scraping, we can model communities or audiences as graphs. By doing graph-theory computations, we can get insights and ideas on how further growth hacking operations should be performed. Perhaps it is not desirable to target major influencers or even their direct followers, but people connected indirectly to major influencer in a social graph (e.g. followers of direct followers, but not direct followers themselves) would be the best audience to hit with automated outreach. Merely plotting it with NetworkX and matplotlib might provide some insight.
See the following examples:
- Osintgram - tool to scrape Instagram for OSINT purposes.
- youtube-dl - tool to download videos from many sources, including social media platforms.
- TWINT - tool to scrape Twitter for OSINT.
In broader picture, social network analysis is entire branch of social science that deals with investigating communities through the lens of graph theory. If we have social media footprint of some community scraped and stored in a database in a structured way we can try to reverse engineer the social dynamics and use them to our benefit.
Automating content creation
Automated social media accounts should have at least some content. There are are ways to automate content generation and/or posting.
If you have content created for one purpose or platform you can use automation to repurpose it for other platform, for example by cutting up a longer video into shorter segments and uploading it to Tiktok for viewers with short attention span.
Another approach is develop code to read RSS feeds of prominent news sources, use something like Bannerbear API to turn them into headline images and post them with link to original source.
Yet another way is to simply repost well performing content from other accounts. Many Instagram theme pages were build by doing exactly this - a practice known as “cash cow” pages.
At this point, AI systems such as GPT-3 can help with content creation, but not replace a human working at reasonable quality. If someone is not supervising it, the text generation can easily go off the rails and produce complete nonsense. One can use GPT-3 to generate Buzzfeed-style listicles, but to go beyond that you would need to get good at prompt engineering - a practice of crafting good prompts that instruct the language model. That is a skill in itself. However, AI systems can definitely make the content creation easier.
Spamming and manufacturing engagement
Mass direct message sending
One spammy tactic is to send mass direct/private message to a pre-scraped list of potential customers. This is popular in NFT-related communities on Discord by using tools like discord-dm-go.
Whether or not it will annoy them is another matter.
There are multiple way to nudge social media users into seeing our account:
- Liking their post.
- Commenting on their post.
- Mentioning their account.
- In some cases viewing their content (watching stories on IG or taking a look at LinkedIN profile of user with paid plan).
All of these tactics can be done with automation.
Instapy quickstart script provides a following example:
session = InstaPy() with smart_run(session): # general settings session.set_dont_include(["friend1", "friend2", "friend3"]) # activity session.like_by_tags(["natgeo"], amount=10)
If we want to collect a following of real people in our target niche it might be desirable to do follow-unfollow tactic that entails automated following bunch of people within the niche every day, giving them a chance to follow our account and unfollowing them if they don’t reciprocate. This is fairly low cost way of not only bringing some traffic to the page, but also building an audience over time. However, we cannot be too agressive with this as platform have rate limits on how many actions we can perform daily (although they change over time). Furthermore, it is desirable to keep your following-to-follower ratio as low as possible, as it may become very obvious over time what we are doing here.
Generally speaking, one should use mobile 4G proxies for social media automation. Yes, that’s expensive, but there are ways to set them up with off-the-shelf hardware. In some cases residential proxies of sufficient quality are also enough.
You don’t always want to use one proxy per account, but the amount of accounts sharing single exit IP should not be high.
If you create or purchase an account you may not want to or be able to use it at full capacity from the beginning. You may need to let it go through a pre-warming sequence, which entails gradually increasing activity on the account to make it less likely to get banned and also to make social media platform increase the rate limits over time. This depends on exact social media platform, account standing and what kind of automations are planned for the account.
API scraping entails making a lot of API calls to backend systems to extract information. If social media platform is prone to cracking down on account based on this thing alone, it is desirable to have some accounts that are easily replaceable and meant to act as cannon fodder while gathering information for further parts of automated workflow. These are called scraper accounts.
Due to automation countermeasures, it is highly desirable to refrain from running any sketchy automations on valuable accounts that we are trying to establish as assets to make money long term. Yet it is still possible to use automation to grow these accounts through a technique called mother-child method. It is primarily used for growth on Instagram and uses 3 level-system of multiple accounts:
- Mother account that is not running any gray hat automations directly, but is used to build a primary audience and sell stuff.
- Child accounts that perform things like follow-unfollow, mass DMs, etc. to bring new people to the mother account and maybe to build secondary assets (that may be difficult because platform is likely to shut them down eventually).
- Scraper accounts that gather audience data for child accounts to act on. These are meant to be disposable and replaced when burned.
Abusing social media ads
Targetting scraped audiences
PPC platforms such as Facebook ads allow you to import list of people (e.g. CSV file with column for emails) that are supposed to be your customers. This can be abused to run ads towards a scraped list of people and possibly save some money on advertisment.
Nanotargeting is a practice of using PPC ad platform to narrow down the ad targetting to very small number of people that we want to influence, perhaps as little as one person. There’s a research finding that 4 rarest Facebook interests of a person makes them unique in the user base with 90% probability.
Manufacturing social proof
Bot or clickworker accounts can be used to artificially inflate the seeming popularity of social media account and fake social proof.
Fake likes, comments and views can also be used to fake social proof.
Astroturfing is a practice that attempts to influence public discourse on product or public issue by having paid shills or bots post on public forums and social media platforms as if they were honest, concerned citizens or members of the community.