These days, AI is all the rage, with tech startups using it to disrupt every industry, from marketing and software development to healthcare and even agriculture.
So, it should come as no surprise to see AI machines capable of writing text. And there are many of them!
But can these machines write well? Which one is the best? Read on to find out.
Top AI Writing Tools
Conversion.ai is a new tech startup backed by the famous Silicon Valley accelerator YCombinator. Their AI engine is called Jarvis, and they say it’s written all the content on their site.
Jarvis can write copy based on many templates, including blog post introductions, YouTube video descriptions, and Google Ads descriptions.
They also have a long-form content generator for blog posts. You have to describe the post you want to write as well as the target audience, the tone, and the introduction.
The more specific you can be, the better results you’ll get. Jarvis produces one or two paragraphs of text at a time, and you’ll have to tweak the output and give it a hint of how you want the rest of the text to go by starting the next sentence.
The tool generally works well for descriptions and short copy, but blog posts require much more effort, and sometimes you’ll be better off doing everything on your own.
- A large number of templates.
- Fluent and error-free output.
- Unlimited monthly plan.
- Regular improvements and updates.
- Support for Spanish.
- Requires thorough fact-checking.
- Inconsistent spelling.
- Relatively expensive.
This is a relatively new platform similar to Conversion.ai and offers almost the same functionality. It’s built on the GTP-3 language model and produces acceptable outputs.
The unlimited plan is $35 per month, compared to Coversion.ai’s $99 per month.
They currently support an impressive list of eleven languages, including English, French, Japanese, and Russian. Plus, you can contact the team to request support for other languages.
- Affordable pricing and financial aid.
- Seven-day no-credit-card free trial.
- Support for multiple languages.
- Google Chrome extension.
- Requires thorough fact-checking.
- Inconsistent spelling.
This tool also comes with different templates for ads, social media posts, and blog content. However, you don’t get as much variety as you do with Copy.ai and Conversion.ai.
There’s also a step-by-step wizard for long-form content, which makes your job easier. You start by providing a rough target topic and receive ten article titles to choose from.
Next, you need to write a brief description of what you expect the article to look like before Writesonic generates a few outline suggestions and lets you choose one.
You can work on the individual headings from there and have the tool add content piece by piece.
- Generates high-quality copy.
- Large clients can ask for custom features.
- Confusing pricing system.
- Struggles with generating long content.
It comes with the standard templates we’ve come to expect from AI writing assistants, but the output isn’t highly impressive and may contain meaningless phrases.
They also offer a Google Chrome extension and a Shopify plugin, but they don’t have a large user base, so it’s hard to tell whether the integrations are any good.
The pricing tiers are a bit confusing because you pay for credit points rather than words. And each template costs a different number of credits. So, you can’t tell how much you’ll be charged per word.
- Plagiarism checker.
- Works better with product descriptions and short-form content.
- The output contains some incomprehensible sentences.
- Struggles with long-form content.
If you google AI writing tools, you’ll find loads of websites claiming to give you hassle-free marketing copy and blog content in seconds.
But, in reality, most of these so-called writing assistants produce absolute garbage—sort of like the content spinning tools that used to be trendy in the past.
Here’s a (non-comprehensive) list of tools you should avoid:
Can AI Write Articles?
AI is already writing articles, and sometimes it’s impossible to tell the difference. In 2020, The Guardian published an article written by a “thinking” robot.
In the article, the intelligent machine explains how its “brain” can make rational decisions about what to include in the text and how to form meaningful sentences.
It also claims that human beings shouldn’t be scared of a machine with such cognitive abilities—but that’s for us to decide!
The subfield in AI responsible for creating comprehensible language, including writing articles, is called Natural Language Generation (NLG). The objective is to replicate human’s ability to communicate in languages (e.g., English or French).
More specifically, NLG converts data, like tomorrow’s weather forecast, into language that humans can read and understand. So, instead of seeing some numbers and icons on your screen, you’ll get something like this: “it’ll be sunny until Friday with showers on the weekend.”
How’s that possible?
Using machine learning. The whole process is extremely complicated, but at its most basic level, it involves feeding the machine a large number of real-life examples and letting it decide how to generate texts based on those examples.
This approach works because human language is sequential and has rules. So, the machine can discover and learn those rules by looking at examples.
If you’re interested in the technical aspects, you should learn about deep learning and Recurrent Neural Networks.
The most successful language generation engine to date is the GTP-3. It gives programmers the tools to use NLG in their applications and get creative with different use cases without reinventing the wheel.
What Are AI Writing Tools Good For?
Essentially, AI writing tools are good at taking structured data as input, identifying patterns, and summarizing those patterns using language models. So, every time you have a set of data and need to create roughly similar texts from them, AI writing assistants can be a tremendous help as they boost efficiency.
Here are a few interesting use cases for AI writing tools:
Social Media Content
You’ve probably seen how LinkedIn automatically gives you a short summary that matches your CV.
To develop this feature, the engineers at the company gathered over a million summaries from user profiles.
They then fed the data to their AI machine to generate summary templates, which were reviewed by writing experts.
Now, every time a user wants to write a summary, the platform suggests one of those templates based on their profile and activities.
AI writing tools also do a decent job when writing short, straightforward texts like social media posts.
Platforms like Oktopost and GhostWriter.ai can take your existing content and extract multiple social media posts with unique descriptions.
Many news agencies have started outsourcing their mundane writing tasks to bots. For example, the Associated Press has increased the number of its quarterly earnings reports by tenfold.
And the LA Times has their Quakebot writing earthquake news, while The Washington Post and Yahoo Sports have used automated writing tools to generate sports-related content.
All of these tasks have one thing in common: they need to turn some predefined information into well-composed writing without much creativity. And as we’ve seen, AI is good at that.
However, journalists still have the advantage because they have superior storytelling skills, and they can explain events and numbers better.
What Are AI Writing Tools Not Good For?
Although AI tools can be incredibly effective at creating fluent text, they don’t actually “understand” their output. So, at least for now, they may churn out completely false information. As a result, writing tools aren’t great for long, complex writing like blog posts or reports, especially if the piece requires aggregating information or presenting analyses.
For example, this YouTuber used conversion.ai to generate a blog article about Beethoven’s life and found that it contained dozens of inaccurate claims.
Speaking from personal experience, not only will these tools not increase your speed when creating blog posts, but you’ll have to spend nearly twice the amount of time to fact-check everything.
Then, once you’ve removed all the false information, you’re left with a half-written article that’s pretty hard to salvage.
What’s more, these tools are limited to their initial training data. Meaning that if something isn’t in their database, they can’t write about it—unlike humans who can do research on the spot.
So, if you want to write about an underserved topic, which is how you generate earnings from blogging, you’re almost definitely out of luck.
So, overall, AI writing assistants are still not a viable tool for content writers who want to rank on Google and offer value to their readers.
But if you’re trying to write on a subjective topic without hard facts (e.g., why do people like broccoli?) or an essay for school, you’ll be fine.
AI copywriters and writing assistants are becoming more prevalent these days, and they have bold claims. However, after working with conversion.ai, I found that the hype isn’t real.
These tools can’t generate the sort of content that lets you capture and hold the top SERPs.
That said, you can still use an AI assistant to generate headlines, outlines, introductions, product descriptions. And that might be worth something to some people.