June 6 is Sweden’s National Day. It’s also the day United Robots is forecasted to deliver our 1 millionth automatically written article since we launched in 2016. While reaching that milestone is definitely something to celebrate in itself, it also tells a crucial story — beyond the actual number — about robot journalism as we see it.

At United Robots, we now send well over 2,000 articles a day to our publisher partners’ content management systems. The texts cover all manner and divisions of sports, property sales, company registrations and bankruptcies, traffic, weather, and stock market movements. And key to the business case is that, no matter how many automatically generated articles a client publishes, the cost is the same.

We can do this because our technical infrastructure is based on a serverless and completely cloud-based architecture, allowing us unlimited parallelism and a very cost-efficient output. A typical text takes around one second to generate. If we generate 100 texts at the same time, they are all completed after that same second. That’s the tech that we’re responsible for.

For our publishing partners, it’s about leveraging that volume of editorial from a business perspective. And in our experience, a lot of the value is specifically tied to the volume itself. Let me describe two publisher use cases to illustrate how:

1. Personalisation: In Scandinavia (a majority of our clients are Swedish media houses), content personalisation is becoming an increasingly important feature as publishers grow their digital subscription businesses. By leveraging the data they have about their readers (demographic, geographic, and behavioural), they can push relevant content to the right person at the right time.

But — and this is key — publishers can only achieve the necessary granularity of content needed for meaningful personalisation if they have access to enough articles to begin with. (NiemanLabs’s recent look into what “news” features in Facebook’s local news initiative, Today In, is instructive: One in three users in the United States live in towns where there is not enough local news published to even launch the feature.)

In other words, publishers may not have the newsroom resources to produce enough content to give every reader personally relevant articles every day. The text robots, on the other hand, produce several different stories on every single football match played in every division and every house sale completed in every village.

As Li L’Estrade, head of content development at Mittmedia puts it: “A really good robot text can have a bigger impact and be more read than a really good news article — but only if it’s a topic readers really care about. On the whole, it’s the volume of articles that’s the robot’s big advantage. Each article reaches a smaller group of readers on average, but in total we get an exchange on par with anything written by our most read reporters.”

2. Launching verticals: We have a number of publisher partners who have built verticals, particularly about specific sports/regions. They identified an opportunity to attract new audiences and the same is true with generating new ad revenue. With the articles written by a sports robot, they have the regular flow and volume of new content needed to keep readers interested and advertisers on board. The same logic could be used to set up specific sites around a local property market or to publish regional business news.

If United Robots’ growth rate matches what we achieved in 2018, we’ll send out more than 750,000 articles this year alone. But more importantly, that volume will allow our partners to improve their bottom lines.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .