We don’t generally use automation in our news social accounts. We haven’t really done A/B testing with this, but our philosophy is that the ideal wording for a headline or photo or video caption isn’t necessarily the same as the ideal wording for a tweet or other social post. For one thing, the space considerations are different — as odd as it sounds, even a tweet gives us more space than a headline to craft something conversational and compelling.

Also, if you look at Twitter in particular, the Twitter Card design — where a tweet opens up to reveal more information about the link — will show the same headline twice if you use it as the text of your tweet.

This doesn’t mean we’d never experiment with automation for our social output — it might work for certain types of feeds, and we’ve done a bit of it on business-side accounts to efficiently deliver information on content to our institutional customers. But for most accounts aimed at sharing news content and encouraging conversation around it, we prefer the human touch.

Every post that goes on @CNN and CNN’s Facebook page is handcrafted for that platform. Most of the weekday posts come from a core three-person team, but we’re also lucky to have a large group of social advocates across CNN helping us keep the accounts going 24/7. If we’re ever in a time bind, we’ll tweet a headline as is, but that’s rare. You’ll never see just a headline on Facebook.

We have done extensive testing and experiments. We can say with confidence that tweeting headlines will indeed generate some engagement. And a great headline can increase the amount of retweets, favorites, and replies. If you add a fact or detail to a tweet, engagement increases even more. But the highest engagement is a direct result of tweets with voice. That doesn’t mean you editorialize stories, but you take the radical approach of tweeting like a human.

The core social team respects social as its own platform and produces content for each social network. Every post is carefully crafted for our audience. How would you tell this story to a friend? Would you share this if you came across it? Those are a few of the questions we think about when crafting posts.

Producers, writers and editors manually post our stories to the main social media accounts across the NBCUniversal News Group — including NBC News, msnbc and CNBC. While scaling a strategy 24/7 has taken time, we’ve found that engagement is greater when the accounts are manually curated.

We do experiment with auto-generated feeds on smaller vertical accounts, but our data shows that users are more likely to engage with a person than a bot. A person also creates an added layer of editorial care when a story moves from the web to being promoted on social media.

We use a mix of automation and handwritten tweets. For the @nytimes account, a headline is automatically tweeted when it’s placed on the NYTimes.com homepage or when we send out a breaking news alert. Beyond that, a human on the social media desk sifts through our 300-some URLs per day for ones that could use some handwritten love, or to show the breadth of our coverage beyond the homepage. It could be a story that won’t hit the homepage but we know it’ll resonate with our Twitter audience; maybe it’s an incremental update on a developing story; maybe it has a great photo or graphic; maybe we can surface a resonant stat or quote; maybe we can just create a smoother on-ramp to a story that’s already been autotweeted.

By most measures — including clicks, retweets, favorites, and responses — handwritten tweets outperform autotweets. But there are some not-insignificant areas where autotweets win: speed, reliability, and lesser time invested by staff. We aim to have a balance of the two that gives us the benefits of both; it allows us to be both timely and engaging, while still being able to spend time on additional newsroom priorities.

The USA Today main Facebook page is all manual. The only automation we’ve had while I’ve been the social media editor at USA Today (I started in fall 2011) was through our proprietary breaking news tool, which posts alerts to various platforms (email, push alerts, Twitter, etc.). We were publishing breaking news alerts with links through that tool to Facebook, until we looked at how they appeared in the News Feed and how they performed, saw that both were poor experiences, and stopped. This wasn’t long after I started in my role. Now we assess whether a breaking news alert belongs on Facebook — because not all of them do — and post it manually if it does. I’m constantly looking at how posts perform overall and tweaking our Facebook strategy to ensure success.

I like to say that the bar is higher for Facebook posts than it is on other social networks, particularly Twitter, because we post less on Facebook but see bigger impact from each post. A post can show up in the News Feed for days after it goes out, which adds to virality and half-life.

The main Twitter account is a more complicated situation, and it’s a blend of automation and manual tweets. The automation is a custom RSS feed that’s meant to tweet out the top four stories on the USA Today homepage. All breaking news alerts are also tweeted through our breaking news tool. We add in manual tweets throughout the day, making sure to post engaging stories, with strong wording and images.

I have a love/hate relationship with the feed. While I know it’s not the best experience for Twitter, it also covers us in getting important news out 24/7, and that’s important for a national news organization that also has an international readership, to be active across time zones. As far as finding a balance, I don’t worry too much about what the feed is doing. It runs in the background, and we post manually as much as possible throughout the day. It does sometimes mean that the same story is tweeted out a few minutes apart, but even then, I don’t see much negative reaction to that — I think we’re often hitting different users even at those close-together times, and people are used to seeing the same story more than once. Of course, the manually written tweet almost always performs better than the automated headline, when we do compare, but there is still engagement on the feed headlines.

The WSJ social media team writes every tweet that goes on the main @WSJ account. We run a 24/7 global social operation with social media editors in New York, London, and Hong Kong; all of the members of our team are journalists who are expected to be excellent writers and exercise great news judgement. We see the @WSJ account as a “front page” of The Wall Street Journal on social media: offering our followers the latest breaking news, as well as a thoughtful mix of in-depth analysis, features, and, importantly, direct audience engagement. We value the intimacy and immediacy that our @WSJ Twitter account offers to our followers and we know from our data and from Twitter’s own analytics that tweets that are written manually, rather than being automated lead to greater engagement, which is ultimately our goal.

In addition, our followers have doubled every year for the past two years:

In July 2012: 1.5 million
June 30, 2013: 3 million
May 2014: 4.4 million

We also think our manual feed heightens accuracy and prevents errors, since writing tweets ourselves keeps us accountable for them in a way that an auto-feed would not.

While our major accounts (@WSJEurope, @WSJAsia, @WSJD for example) are run by editors, the Journal does have several sub-branded accounts that are automated. We have encouraged our reporters and editors around the globe to take ownership of those accounts and, at the least, create a mix of automated headlines and manually written tweets. Because we know images help drive engagement, we have encouraged editors to include images in their tweets, which has had the effect of reminding editors of the importance of publishing great visual content to accompany their written stories. In addition, we created a “social headline” field in our backend publishing system that provides a social-friendly headline that can be shared by readers who come to our article pages and that feeds some of our automated accounts, so that the automated headlines, when used, are as engaging, direct and conversational as possible.

The accounts are managed by hand, because we believe that you can’t replace the intuition, analysis, and timing of our editors, reporters, and producers — our people.

We use TweetDeck to manually schedule handwritten posts, and we will schedule handwritten Facebook posts in the platform itself, but for the bulk of the day and evening, there is a human mind and human hand driving the social presence of ABC News.

The people ingesting content are humans and there should be a human providing the content to them; we’re not an automated bank or cable phone line — we’re people, and the people we reach are people, and they can sense the way posts are produced. They know the voice of individual editors and producers, and when it comes out artificial, they can feel it.

I schedule tweets with specific times for specific stories, to hit target audiences. If you go and look at Topsy data, you will see that we slow down when people make their trips home for the evening, and then have a long spike in the evenings into the mornings when some of our manually scheduled content fires. We’re a global news outlet and there is always a live audience there reading and interacting with content.

There may come a time when we have computers smart enough to be more intuitive than just picking up on a spike in traffic for a certain post. But right now, they are not smart enough to replace the analysis and judgment of the people of ABC News, with an expert understanding of the material and platforms they’re pushing.