When Machines Start Writing Back: My Problem With DataPop & Automated Content Creation


I had the opportunity to meet a few of the folks from DataPop at SES San Francisco this summer. Not one to judge after just one meeting, I knew I needed to speak with them again. I had a bone to pick.

It wasn’t the first time I’d heard of the relatively new LA-based company, who call what they do “Creative Science.” Web marketing veteran Bryan Eisenberg told me months ago, “You have to see what these guys are doing.” (When he talks, I tend to listen.)

And what do they do, exactly? They use big data and technology to generate PPC ad creative. That is, their software actually writes the ad copy, based on the mass of information at their disposal. As a writer, I had a big problem with this.

In a recent conversation with DataPop’s Tal Halpern (Marketing) and Dave Schwartz (Client Services), I had the opportunity to pick their brains. Big data has been a fascination of mine for a while now and here are these two, telling me it actually means they don’t even need human writers anymore. I don’t know if I can get onboard with this at all.

The Evolution of Web Content Creation & The Problem with Human Writers

robot-writerMany moons ago, I worked with a lot of affiliates, writing their ad copy and creating e-books. I belonged to quite a few forums and had a pretty great community of writers around me.

We had this concept of machine learning, AI and computer-generated content in the backs of our minds even then, but it was the stuff of science fiction.

“Some day, they won’t need us! Computers will just do the writing for people.” Ha, ha. These ideas existed only peripherally (though they were probably coming to fruition in a few dark labs somewhere in Silicon Valley, even then) and didn’t seem a realistic threat.

Crowdsourcing came first and happened to coincide with an almost insatiable need on the part of businesses coming online for masses of content. The Quality Content mantra was just a twinkle in Matt Cutts’ eye in those days. Article spinners, keyword density counting software, and cheap labor reigned supreme.

In 2004, it became clear writers needed to join forces if we were to compete with overseas firms for contracts. Clients demanded massive quantities of content, as many as 800 articles on a single topic at a time. Transitioning to a project manager, I won these jobs and quickly learned how difficult it is to manage a stable of writers.

I had to hire twice as many as I actually needed, because 50 percent of the writers would flake out. Consistency in style, tone, or voice were pipe dreams. Quality was all over the map. Some even used spinners or blatantly plagiarized (too bad I already knew those tricks) to meet the required word counts and keyword density and still, they just kept pumping out the content.

I won’t even get into the content farms… this time. Let’s just say writing for a paltry share of AdSense was still a lucrative second income for many. It was just what serious writers had to do at the time to survive, between more respectable contracts.

Crowdsourcing eventually went corporate and has now become mainstream in every industry from health records to paid search. Writers have had to strike out on their own – see the influx of bloggers and self-published authors since mid-2000s – or go legit, either in-house, in major media, or toiling away bidding on the scraps not chewed over by agencies.

Meet the DataPop Cyborg

Fast forward to today, because I’d just keep reminiscing about the good old days, if my editors would let me. For those still making a living at writing web content, along comes DataPop and their Creative Science platform to spoil the party. They may not be first on the automated content scene, but they caught my eye and so are the subjects of my indignation and wrath as a writer scorned.

DataPop was founded late in 2008, by former Yahoo/Overture staffers. They’re not another crowdsourcing company; ad creative is platform-generated, which is a fancy way of saying written by a machine. Right now they’re only doing this in English, though non-English copywriters should feel the heat of this animal breathing down their neck shortly.

They built the platform with the intention of solving a particular problem: Most search marketers just aren’t getting the best possible results from their ad creative.

“Basically, we take the best of what advertisers have done previously, in addition to their own data, and turn it into custom messages,” Schwartz explained. Their clients are typically enterprise-level companies seeking scalable content and improved ROI using the massive amounts of data at their disposal.

“So your content is machine-generated?” I asked.

“Well, no,” said Schwartz. “It’s more like a Cyborg…”

Creative Science = Content at Scale

You have to understand, at this point, the marketing geek in me is intrigued, but the writer is terrified.

The Creative Science platform can gauge searcher intent, read geosignals, factor in pricing and offer targeted deals based on this data, coupled with its understanding of the company’s previous successes. Yes, I said understanding! It understands things.

DataPop have had to put a number of quality checks in place, one of which is a human editorial process. Now this, I can get on board with.

As Schwartz, Halpern and I discussed the process by which the Creative Science platform generates ad copy, I started to understand the possibilities, perhaps not immediately, but certainly as the technology develops.

Could content creation automation eventually replace bloggers, or even journalists? Maybe in a few years, it will be able to populate social media feeds. If it ever touches creative fiction, we’re doomed.

On the one hand, I can’t be convinced software will ever completely replace writers. In the marketing realm, however, the Creative Science platform simply has the ability to process data writers can’t begin to touch at scale.

What’s Next?

Content creators are going to be pushed out of their comfort zones and many may have to transition to more editorial roles as the world’s need for information grows. Where we used to write to the machine, painstakingly trying to balance the readers’ needs with algorithmic preferences, the machines are now writing to us.

We’ve become better at telling them what it is we want, and they are quickly gaining ground in being able to deliver it to us.

These automated solutions are still driven by humans, people like those at DataPop, who have identified factors hindering content production at scale and are developing remedies.

After spending some time getting to know them and their platform, I fear the days of creative writers toiling away at the winning concept and copy for ad campaigns are numbered. The thought just doesn’t seem quite as scary, anymore.

Through big data analysis and automated content creation, more searchers are able to find what it is they’re looking for, in a format they’ve indicated through previous interactions as preferable to them. This is now possible at speeds and volumes unheard of only a few years ago.

DataPop are leading the charge in the ad copy arena; expect to see them branch out into other languages and even types of content as they further their technology. Human content creators will need to work with the machines; resistance is futile. It may require a new skill set and as I’ve learned, a fundamental shift in the way we think of writing.

At least in the online world, content is king. As its loyal subjects, writers and marketers must get on board with the technologies that will allow us to feed searchers’ insatiable need for the right answers, in the right places, at the right time.

What do you think of automated content creation? Let us know in the comments!

Related reading

Ten ways to pump out a stream of great content without burning out
Five tips to create an SEO-friendly FAQ page
The fall of ad copy, long live ad copy
Nine types of meta descriptions that win more clicks