At this level, you’ve heard all of it about AI. AI can assist you write higher. AI can assist you create quicker. However can it assist your social captions carry out higher?
As a busy social media marketer, something that guarantees to make the content material creation course of simpler and quicker is interesting. And with instruments like ChatGPT making AI accessible and user-friendly, it’s value testing for writing captions.
Utilizing AI to write down social posts, particularly on Twitter, additionally begs the query: Do AI-written captions influence engagement and attain? Does the Twitter algorithm penalize tweets written by AI? Can it actually even inform if a Tweet is written by AI? I ran an experiment to seek out out.
My speculation for this experiment is that human-written tweets would get extra engagement and attain than AI-written ones.
Why? Nicely, I’ve been on X (previously Twitter) for greater than ten years, so by this level, I ought to know my very own viewers finest. Plus, people have been writing for hundreds of years and AI has solely been doing this for two — that has to rely for one thing, proper?
However earlier than I carried out the experiment, I wished to get extra perception from somebody who works instantly in social each day. So I requested one in all Hootsuite’s social specialists, Trish Riswick, about her expertise with AI and if she’s observed any distinction in engagement in comparison with human-written captions.
Riswick shares that the Hootsuite social staff has experimented with utilizing AI to write down social captions in a few methods.
“At Hootsuite, we now have two various kinds of tweets that we put out,” says Riswick. “We put out enjoyable, linkless, personable tweets for model consciousness. And we additionally put out stuff that’s purpose-driven for once we need folks to take an motion, like when we now have a product replace or a webinar that we wish folks to join.”
Riswick says that she has observed a distinction in engagement and attain when she makes use of AI for purpose-driven content material.
“We see actually nice engagement and attain once we’re utilizing AI for bottom-of-the-funnel posts as a result of it’s capable of ship the data faster than we’d have the ability to if we had been writing it,” says Riswick. “After we give AI a particular, simple job (that doesn’t depend on model humor or data of present occasions), the outcomes assist us create efficient CTAs or a special method of relaying that message.”
On the flip aspect, she has not observed an enchancment in engagement or attain when utilizing AI to create model consciousness content material.
“These are the issues that AI instruments don’t actually seize in addition to a human does,” says Riswick. “With the ability to converse in your model voice is one thing AI remains to be studying. AI instruments present an incredible place to begin and assist velocity up the method, however human oversight and enhancing remains to be a should.”
To check whether or not human-written Twitter captions generate extra engagement and attain than AI-written ones, I got here up with three units of tweets to share on my personal Twitter account.
I began by sharing my very own tweet in my common tone and magnificence. Then, I prompted ChatGPT to provide you with an analogous tweet so the 2 could possibly be pretty in contrast.
For instance, one in all my private tweets was a tip for fellow freelancers. I then gave ChatGPT a easy immediate to duplicate my tweet:
Whereas there are a ton of creative prompts you can give ChatGPT, I gave the AI instrument some prompts that had been on the easier aspect, just like the one above.
I wished to maintain it fairly bare-bones to provide ChatGPT the liberty to generate an natural outcome somewhat than one thing that was closely refined. (The thought is to see if we are able to get the identical outcomes whereas saving time, in any case.)
For this experiment, I posted 4 tweets that had been textual content solely and two tweets that included a hyperlink. Let’s dive into the outcomes!
I checked the efficiency of every tweet about 24 hours after I posted utilizing Hootsuite Analytics. Listed below are the outcomes:
|Submit with hyperlink
|Textual content put up 1
|Textual content put up 2
At first look, it’s clear that the AI-written tweets had the next engagement price than my tweets two out of the thrice. Nonetheless, two out of the three AI-written tweets additionally had fewer impressions than the human-written ones.
What do these outcomes imply? Let’s break down every of those tweets in additional element.
As a author, I usually flip to Twitter to share my current work. So my first tweet for this experiment was a tweet with a hyperlink selling an article I wrote.
Within the tweet, I summed up the article and supplied a little bit perception into why somebody might wish to learn it by main off with a private truth. This can be a fairly commonplace system I take advantage of for my tweets.
Listed below are the outcomes of this tweet, in accordance with Hootsuite Analytics. The tweet generated an 8.71% engagement price, 68 whole engagements, and 781 impressions.
As I stated, it’s regular for me to share my work on Twitter. I additionally like so as to add some private background to most of these tweets to offer context so persons are extra compelled to learn them.
On high of all that, the model I discussed within the tweet (@Wethos) retweeted it, which undoubtedly helped enhance the variety of impressions.
I requested ChatGPT to create one thing comparable utilizing one other article I wrote for a similar model, so there was as little bias as attainable. Right here’s the tweet ChatGPT got here up with:
And listed below are the outcomes of the AI-generated hyperlink tweet. This tweet acquired an 11.43% engagement price, 4 whole engagements, and 35 impressions.
Whereas this tweet technically generated the next engagement price, the variety of impressions wasn’t even near the quantity the human-written tweet acquired. I’d additionally notice that in my view, it’s much less personal-sounding than the tweet I wrote.
And in contrast to my first tweet, the account I discussed didn’t work together with this put up. To be truthful, I shared the 2 tweets that talked about the model a few week aside. There’s an opportunity that they had been much less prone to reshare once more in such a short while body.
There’s little question that the retweet on the primary hyperlink tweet boosted the variety of impressions, and never getting additional protection on this tweet resulted in very low visibility.
One other matter I usually put up about is freelancing. I recurrently share ideas that I feel different freelancers would discover helpful. The put up beneath is an instance of a tweet I’d write on a standard day.
This tweet generated an engagement price of two.2%, acquired 5 whole engagements, and 227 impressions.
On paper, these outcomes aren’t a lot to write down residence about. However they’re on par with my common efficiency, particularly contemplating how small my viewers is (fewer than 300 followers). This tweet additionally carried out nicely above the common Twitter engagement rate of 0.05%.
I requested ChatGPT to recreate an analogous tweet by prompting it to write down a tweet that shares a tip for freelancers. Right here’s what it got here up with:
If I’m being sincere, I didn’t count on this tweet to carry out in addition to it did. To me, it sounded a little bit tacky and didn’t really feel like one thing I’d say naturally. However the outcomes inform a special story.
In line with Hootsuite Analytics, this tweet generated an engagement price of 8.26% and acquired 20 engagements and 242 impressions.
One of many important causes I feel this tweet carried out higher is as a result of it included a hashtag and a few emojis. I didn’t embody both of these in my preliminary tweet as a result of I normally don’t.
Regardless of my hesitation, although, I suppose there’s a cause the Hootsuite Blog recommends adding hashtags and emojis to your Tweets as a basic finest observe in any case! That engagement spike is tough to argue with.
My remaining tweet for this experiment is about what it’s like working for your self and being your individual boss. Right here’s the tweet I wrote:
After checking Hootsuite Analytics, I can see that this tweet generated a 7.22% engagement price, 7 engagements, and 97 impressions
As I discovered all through this experiment, ChatGPT *loves* including emojis and hashtags whenever you ask it to write down a tweet. So for this immediate, I particularly requested it *not* to incorporate any emojis or hashtags.
Although the hashtags clearly made a distinction final time, I wished to see if the AI might generate content material that might be extra just like the tone and magnificence of my common content material.
Right here’s the immediate:
And right here’s the tweet ChatGPT got here up with:
Right here had been the outcomes of this tweet. This tweet generated a 1.92% engagement price, 1 engagement, and 52 impressions.
This tweet is textual content solely, so it’s similar to the human-written one. However it generated a a lot decrease engagement price and acquired fewer impressions.
Why had been engagement and attain so low this time? Nicely, so far as tone goes, whereas the AI-generated content material is just like my type of tweets, it sounds fairly generic.
My preliminary tweet was a little bit extra playful, however ChatGPT’s model is fairly flat — and the numbers again that up. Just one particular person interacted with this tweet, which fits to indicate how unnoticed it went within the feed.
General, the AI-written posts carried out higher than human-written tweets — no less than by way of engagement.
In case you’re a busy social media supervisor and your major success metrics are attain and engagement, these outcomes show that AI can assist you create content material that follows basic finest practices and can doubtless carry out nicely.
Nonetheless, though a number of the tweets written by ChatGPT had the next engagement price than those I wrote, the interactions weren’t essentially as significant.
For instance, the AI-written hyperlink tweet generated the next engagement price than the human-written one, but it surely had only a few interactions. The one I wrote acquired 16 profile visits, 15 hyperlink clicks, and even just a few bookmarks, in accordance with Twitter’s native analytics. In my expertise, these interactions usually tend to result in work, so I contemplate them extra precious.
Listed below are extra of my insights from this experiment:
If there’s one takeaway I’ve from this experiment, it’s that AI instruments are quick. These instruments can crank out tons of tweets in a matter of seconds.
It could take me some time to write down my tweets, and I solely share a handful on any given week. So I can solely think about how lengthy it takes to create every week or perhaps a month’s value of captions for a model account that’s lively on Twitter.
AI instruments might be useful to make use of when you might want to write a giant batch of captions however don’t have the time to spend writing from scratch.
One other stunning takeaway from my experiment? ChatGPT is well-versed in Twitter finest practices. It understands easy methods to write one thing brief and attention-grabbing, and even is aware of that together with hashtags can assist with attain and visibility.
To be sincere, I had by no means put a lot thought into hashtags or emojis earlier than. As an expert author, I all the time prioritize the tone and development of a Tweet greater than anything, and it’s labored nicely for me (up to now).
However after seeing how nicely ChatGPT’s Tweet with hashtags carried out, I’m open to including a hashtag or two in my Tweets going ahead — particularly after I wish to attain a wider viewers.
Even when the captions AI tools create aren’t completely according to your model’s voice or type, they provide a very good place to begin and can assist you brainstorm concepts. They’re additionally nice at repurposing and summarizing long-form content material.
Whereas ChatGPT requires a little bit of finesse to create social captions, an AI instrument constructed for social media like OwlyWriter AI could make the method quite a bit smoother.
As an example, OwlyWriter’s “Begin from scratch” choice means that you can choose your required tone earlier than producing content material. This positively might’ve been helpful in creating tweets that align with my writing type.
We’ve all seen tweets which are clearly AI-written. They normally sound generic, lack context, or simply don’t make sense.
In case you use a instrument to write down your tweets or captions, they in all probability gained’t be publish-ready immediately. You’ll normally have to evaluation for accuracy and tweak till your posts sound such as you.
Finally, if you happen to’re going to make use of AI to write down captions (and we really think you should!), consider the output as extra of a place to begin. You’ll have to make use of your social experience to refine your outcomes, however you may in all probability get them executed a lot quicker than if you happen to had been ranging from scratch.
Begin constructing your social media presence utilizing Hootsuite. Schedule and publish posts on to X, interact your viewers, measure efficiency, and run all of your different social media profiles — all from one easy dashboard. Attempt it free as we speak.
Save time and develop quicker with OwlyWriter AI, the instrument that immediately generates social media captions and content material concepts.