BUILDING A BETTER WEB
The week of September 21, 2014

Let’s destroy ‘Around the Web’ links

By Nick Douglas

I’ve found a way to block “Around the Web” links—the headlines that often appear below a blog post or news article, sometimes with pictures, leading to other sites. They’re served by content discovery networks like Taboola, Outbrain, and Disqus, and typically bring out the worst in Internet journalism. They distract us with baiting headlines like “10 Trends Men Hate” and “18 Celebrity Hairstyles We Hope to NEVER See Again.”

I have rid myself of them, and I could tell you how to block these ads too. But if content recommendation engines are ever going to anticipate my interests and elegantly whisk me from one fascinating article to another like a self-driving car—the way executives from Outbrain and Taboola hope to—these networks need to evolve. And for that to happen, you’re going to have to start bending these services to your will.

(OK, I lied. Here’s how to remove the links from your own blog’s Disqus section, and how to block Outbrain and Taboola on any site, using Adblock.)

Plenty of us hate these links. The top four Google results for “around the Web links” give instructions for disabling Disqus’s links. Most instructions for link removal are prefaced by complaints. For one blogger, Disqus images “didn’t seem appropriate for an educational audience.” Another blogger was disgusted because Disqus was “recommending garbage.” The links from Outbrain, Taboola, and Zergnet are “the basest form of linkbait. … They are, quite frankly, embarrassing and ugly. I hate them, and I want them dead.” Bloggers who can’t control their platform’s links also complain. One Daily Kos user calls the site’s Taboola links “counterproductive and bizarre.”

Even balanced journalistic analyses, which acknowledge the links’ benefits—they’re a new revenue stream for publishers and outperform traditional ads on mobile—start by pointing out how they dangle embarrassingly off superior articles. The Washington Post’s Paul Farhi calls them “silly, scandalous or salacious links, the cheesy and the sleazy” in a piece titled  “You’ll never believe how recommended stories are generated on otherwise serious news sites.” Tech consultant Rob Millis dismissed them as “Those F%#&!ing Content Links.”

I don’t mind if TMZ or the Huffington Post Entertainment section want to follow their fluffy content with links to other sites’ fluffy content. My problem is that more serious or academic content doesn’t come with more serious links. Below an Atlantic piece, “Remembering the Nazis’ Disabled Victims,” Taboola gave me links to “12 SF Obsessions New York Hasn’t Discovered Yet” (New York magazine), “30 Candid Met Gala Snaps You Haven’t Seen (But Need To!)” (Refinery29), and “How to treat COPD (Video)” (HealthiNation). A Slate photo essay on queer life in Kampala, Uganda, where gay sex is illegal, has Outbrain links to “16 Hilarious Road Signs” from Reader’s Digest and Glamour’s “10 Spots Guys Really Want You to Touch.”

It’s surprising to me that intellectual sites earn enough revenue from these links to cover the damage done to their brands, but I couldn’t find any revenue figures. Salon, Cracked, the Atlantic, and the Guardian did not respond to requests for comment. Representatives from Slate and the Onion declined to comment. (Editor’s note: The Daily Dot discontinued its on-site use of Outbrain in June. It generated an average of roughly $10,000 per month in revenue.)

Why does continuity matter so much in film and music and not text?

What’s interesting about content recommendation engines is that they actually work great within a site’s own network. On the aforementioned Slate piece, Outbrain recommends relevant Slate stories like “Love Is Strange Paints an Accurate Portrait of Anti-LGBTQ Workplace Discrimination” and “Two Beauty Pageant Winners Come Out as Lesbian.” Not surprisingly, it’s easier for Outbrain to find relevant content on the site you’re already reading.

Outbrain and Taboola both offer publishers sophisticated recommendation engines that more closely mimic Netflix and Pandora; these internal links seem like perfect models for the rest of the network. But every article on a well-edited site like Slate is somewhat “related” thanks to a shared sensibility, and the amount of possible links to sift through is far more manageable, even for an algorithm. It’s like training software to recognize spoken numbers vs. the entire English language.In a way, the links you see are your own fault.

I talked to Taboola founder Adam Singolda and Outbrain VP of Global Marketing Lisa LaCour. They both emphasized “serendipity” and downplayed the role of matching links to the content they follow. Taboola, for instance, uses an impressive range of factors to pick any given set of links: the publisher’s choices of allowed source domains and topic areas; user data like demographics and Web history; and situational factors like the time of day and week, how the user arrived at the current article, and whether they’re using a computer, tablet, or phone. These factors all affect user behavior, apparently more than the actual content above the links.

Singolda compares Taboola to Netflix or Pandora (he says he started Taboola after wanting TV recommendations more personalized than TV Guide’s), but both of those base most of their recommendations on content matching. Netflix famously organizes films by microgenre, and if I tell Pandora I like the Blow, it plays me a song by Au Revoir Simone with “similar basic rock song structures, electronica influences, a subtle use of vocal harmony, mild rhythmic syncopation, and extensive vamping.”

Why does continuity matter so much in film and music and not text?

Neither LaCour nor Singolda would specify any tones or topic areas that perform best on their networks. I don’t know whether tone and topic matching is a problem they can’t crack, or if they really don’t think it matters. For all I know, people actually crave slideshows of notable haircuts. A bored office worker’s afternoon browsing session might look the same with or without Outbrain’s assistance.

In a way, the links you see are your own fault.

But there’s an even bigger issue with these networks: They still rely on advertising dressed as “content.”

Outbrain’s content guidelines require linked content to contain “editorial substance beyond mere promotion of a product or service,” LaCour notes, but the service’s definition of “substance” is blurry at best. Outbrain links like a catalog page from the clothing store the Line; lists of credit cards from credit card affiliate sellers NextAdvisor and CreditDonkey; and (most substantially) Ancestry.com’s brief article on last-name origins bookended by invitations to a 14-day free trial.

Worse, every time Outbrain or Taboola ban certain links, they just find homes with less scrupulous networks. A little up and to the right of any Taboola section on Salon is another module from Hexagram, which serves real barrel-bottom-scrapers. I just loaded an article defending George Orwell’s “Politics and the English Language”; Taboola includes some Kardashian news, but Hexagram vaults past it with links to two slideshows, from two different sites, of “child actors who grew up to be ugly.” (One of them features Mara Wilson, who is adorable, so I reject it even on its own terms.)

In a way, the links you see are your own fault.

As on Facebook, if you click on something, you’re more likely to see more of it. (Unlike on Facebook, you won’t see a link again after you’ve clicked it.) Taboola even lets users “X-out” links they don’t want to see again, to more quickly train the service. This also means that theoretically, someone out there is getting Outbrain links to peer-reviewed histories of Henry VIII’s courtiers, and the rest of us schmucks should never have clicked the first clickbait that sent us slouching towards Bethlehem.

But it’s more likely that the algorithms of Taboola, Outbrain, Disqus, Zergnet, Yavli, Hexagram, nRelate, Gravity, and the rest just can’t escape the orbit of the mass-shared tastes—literally the lowest common denominators—long enough to serve the intellectually curious something we’re truly and specifically interested in.

I see no reason why they couldn’t eventually do so. They just need more data, more opportunities, and the willing participation of the publishers of that worthy content. But I’m not interested in spending my time X-ing out celebrity breakup stories and 10 tricks to consolidate my debt, carefully clicking the most intellectual stories to train these networks to bring me my slippers instead of peeing on the Internet’s floor.

I’d rather everyone else do the training for me. I’ll come back in a year or two and see.

 

Photo by Nicolas Raymond/Flickr (CC BY 2.0) | Remix by Max Fleishman