AI can do a variety of unimaginable issues. However of the numerous hundreds of makes use of on the market, certainly one of my favorites is as a writing editor.
As soon as upon a time, these chatbots might barely deal with a few paragraphs of textual content to look by, however now, they’re greater than able to diving deep into pages and pages of textual content, critiquing it and analyzing each professional and con.
The task
For each of the chatbots, I gave them a piece of text that was around 600-700 words. It was from a recent article I’d written covering the news of ChatGPT going into a ‘code red.’
Every chatbot was given the immediate: “You’re a sub-editor for a tech journal. Learn by this textual content and pick any grammatical errors. Spotlight any ways in which it may very well be rewritten to enhance the textual content.”
I additionally made positive to make use of a ‘Considering’ model for every of those, which is an upgraded model of the chatbot that spends longer contemplating a job.
ChatGPT
ChatGPT was critical. It described sections of my writing as ‘clunky’, ‘wordy’ and ‘clumsy’. However, with each of its put-downs, it did quickly follow up with a way to fix the problem.
In total, it had 24 different suggestions for me. None of them were essential, and the text still made sense without them. In some cases, I even preferred the original, with the suggestions removing emphasis where it was needed.
However, for the most part, ChatGPT nailed the task. It took my writing and analysed it deeply.
For each suggestion, it included the original version, the changed alternative and the reason why the change should be made.
It followed all of this up with a smaller list of ‘general notes’ including repeated phrases, changes in tone and lack of clarity in some places.
Gemini
Gemini opened by telling me the text “suffers from wordiness, repetitive phrasing and passive sentence structures”… thanks, Gemini.
It then goes on to list the changes that need to be made, along with the original and new versions.
Most of these are the same six or seven suggestions made by all of the other chatbots. However, it goes a step further by grouping them into categories of grammatical mistakes and listing out every time they appear.
It then finishes up by offering a rewritten version, implementing all of these changes in one go.
What I appreciated about Gemini’s edit over its competitors is that it offered advice instead of just giving the changes.
For each type of grammatical error, Gemini warned of how to avoid it in the future and gave tips on how to improve the types of errors I was making.
Claude
Claude surprisingly had the fewest suggestions to offer up, especially since it is known as one of the better writing chatbots and has a focus on the academic side of things. It felt quite lacking here.
It offered up suggestions in the same way as the others, listing the problem area, the suggested change and why. But it found far fewer of these compared to its competitors.
Most of these changes were minor and were often more personal preferences than necessary alterations.
It also lacked the added list at the end that its competitors had, listing out more general concerns on tone, style or angles.
Copilot
Microsoft’s Copilot uses ChatGPT-5 in its most up-to-date model, which, in principle, ought to then generate an analogous response to ChatGPT.
Whereas Copilot’s response was much like ChatGPT’s, it did it with far fewer phrases. It gave examples of what it will change with an accurate model, in addition to a really transient description of why it thinks it must be altered.
Relying in your preferences, this can both be nice or fully missing. It strips out a variety of the reasoning, specializing in the tip outcome.
It additionally adopted up by providing a totally rewritten model. Whereas this may very well be useful for some, I prefer to make modifications myself, figuring out what I’m altering and why, solely making alterations that I agree with.
Grok
Grok had rather a lot to say on this job. It gave me an extended reply, consisting first of some grammatical points with strategies on the right way to repair them, adopted by smaller notes on minor grammatical points.
Whereas a number of the strategies had been helpful, others didn’t actually make sense. In a single case, it supplied up an error, adopted by a correction. Besides, the correction was precisely the identical as the unique model.
It additionally appeared to battle with adapting to latest information, claiming that ‘Claude Opus 4.5 didn’t exist.’ This mannequin got here out final week.
Nevertheless, there have been loads of helpful strategies. Grok gave me a listing of locations the place the copy may very well be decreased if a bit was overly wordy, in addition to phrases that had been used too often.
Winner
While all of the chatbots gave suitable responses, ChatGPT and Gemini felt like the strongest options.
Both Chatbots picked out the key errors, but went about the answer in different ways. ChatGPT felt very much like my hand was being held, pointing out each problem, along with why it needs changing, and what to change it to.
It was one of the more detailed responses, and followed the brief as it was asked.
Gemini, on the other hand, focused its attention on guiding me to the answer. While it offered up a corrected version if I wanted to simply use that, it also gave me each individual issue, along with the original, the new version, and tips on how to avoid that problem going forward.
Comply with Tom’s Guide on Google News and add us as a preferred source to get our up-to-date information, evaluation, and evaluations in your feeds.
Extra from Tom’s Information
Again to Laptops
