Skip to main content
PBS Public Editor

Data Expert Answers Our Questions About 'Bot Farms'

Email share
Paul Bischoff, Editor, Comparitech
Comparitech

Comparitech found a bot farm with 14,000 fake Facebook accounts that was generating 200,000 political posts each month

The PBS Public Editor’s office recently talked with Paul Bischoff, a consumer data privacy expert and editor at Comparitech, a firm based in the United Kingdom. We wanted to know if PBS had recently been the target of bots, or if real humans generated the online noise that we were seeing about some subjects that seemed controversial. But were they? 

All responses were edited for clarity and brevity. 

Is a bot ‘artificial intelligence’

Paul Bischoff: Not per se. AI may be used, for example, in setting up phony accounts and loading them up with followers (many of which are other bots). Bots are used for influencing and for scraping data (personal data that can be used for advertisers). Where AI comes into play is the concealment of true intent, i.e., covering tracks, making an account look like a real person. 

How do you spot a bot? 

PB: They are getting more difficult to detect but they have some things in common. Their Twitter handles tend to be alphanumeric strings — not personal names — like, say Kdh00738901. Most accounts start with a string of letters followed by numbers. The account names (handles) are generated by a machine (bot). You can also look at the age of the account.

Bots are getting harder to detect. 

Are bot farms just enterprising hackers out to make a buck? 

PB: There are gray market services — for example, an influencer might say 'pay us $10 we’ll get you a thousand likes'  —  that kind of thing. There are many bot farms in the world operating on that basis, yes. They are offering services for sale, for a profit. 

Aren't random-looking alphanumeric strings of characters a dead giveaway? 

PB: Most people don't even look at the account name; they look at what is being tweeted and what the reaction has been. If you look at what is being tweeted, if you see a post that has 500 likes I think you're likely to jump on the bandwagon and say hey, this is legitimate as opposed to the one that has three likes.  

Actual content doesn't have to be super advanced — maybe in more targeted campaigns the tweets might be a little more sophisticated. Basically, the bots source their content — images and text —  somewhere (they don't write in prose) but that's kind of the black box for us.