November 7, 2024

It's Giving Chatbot: Make Writing Sound Less "AI"

Type Content Team

Experts in AI and writing

@typedotai

There's a whole lot of hand wringing anxiety about AI right now, some for valid reasons and some not.

In the world of writing—especially for copywriters responsible for blogs, social, and other marketing content—there's a worry that our jobs will soon be given to a water-cooled AI server farm churning out hot takes for the clicks. But this seems unlikely—good writing makes you think and feel things, even ad copy. Maybe a CPU can put a sentence together that makes you pause and contemplate it for a second, but there's no way a machine is going to tap your nervous system and evoke an emotional response.

That's why, for talented copywriters the world over, generative AI feels like a gift from the heavens. The internet is awash with content, and we've been dutifully typing this stuff up for 15-20 years now. Generative AI is a tool for us, which lets us automate the mechanical and really focus on the human aspect of our work. Having AI write for you is really more a failure of the writer than a triumph of technology.

Having AI write with you on the other hand, is a human triumph that gets you to the end of your workday a lot quicker—and with great results. Just like any tool, you've got to learn how to use it properly to get the most out of it.

Table of contents

Common Mistakes of AI Writing

For all of AI's impressiveness, it still lacks in two very distinct ways, and trusting its output blindly is an all too common mistake. Writers should take care to thoroughly read every word that an LLM spits out with the skepticism of a newspaper editor. 

Learn the Signs of AI Prose

Let's be real—AI writing has some pretty obvious tells. If you've spent any time reading AI content, you'll start noticing these patterns everywhere. You know, the little words and phrases that suggest your assistant Claude never went to the 8th grade English class that's supposed to strip writers of these habits:

  • The "In conclusion" syndrome: AI loves wrapping things up with "In conclusion," "To sum up," or "Finally." Humans don't need to be told that we're about to conclude our thoughts—we do pretty well with context.
  • Robotic transitions: Watch for "Moreover," "Furthermore," and "Additionally" showing up way too often. Look out for overuse of corporate jargon, too. LLMs were trained on trillions of bytes of data from the internet—a commercial network. It definitely thinks we all regularly use words like "synergy" and "deliverables."
  • The personality vacuum: Generic, safe language that could've been written by anyone, anywhere, about anything.
  • Overcomplicated explanations: AI has a tendency toward long-windedness, using 50 words when 10 would do just fine because the point was made at the beginning of the sentence and there was no need to continue on, but the AI wants to be thorough and make sure you didn't miss anything so it writes needlessly long sentences just like this one to pad out the length and seem smarter.
  • The "As an AI language model" confession: LLMs so frequently add the disclaimer that they're an AI that if you left this phrase in a piece of writing you submit, you can't even blame the AI.
  • Template-style writing: Every paragraph following the same rigid structure, with boldface headers and paragraphs that look almost identically long.
  • Overuse of "one": One should always keep in mind that no one talks like this, and only the driest of writers would include this kind of phrasing. 
  • Too-perfect grammar: Real people break grammar rules sometimes—AI usually plays it safe. 

The trick isn't just spotting these patterns—it's knowing how to fix them. That's where the human touch comes in.

Big Little Lies

Another of AI's drawbacks is what's commonly called the "hallucination." If a human made up facts the way AI does, we'd call them a "liar" or a "politician." But artificial intelligence has no intent to deceive, so we call them hallucinations. Whatever you want to label them, they're bad news for your writing. 

The most important thing to understand about the way LLMs come up with facts is that, by default, they aren't accessing the internet and researching for you in real time. They're trained on data that already exists, and every LLM has a "knowledge cutoff date." ChatGPT's cutoff is October, 2023; Claude's is April, 2024. With a paid ChatGPT account you can instruct it to search the web for real time answers, but this is not a guarantee that you'll get accurate information. You can even get ChatGPT to provide you with a citation of where the data came from; follow the link and you might be surprised to find that it still made stuff up.

There's an old phrase that serves as good advice when partnering up with someone: Trust, but verify. This doesn't exactly apply to an LLM, which requires an update to the saying: Never trust, always verify. Every fact, metric, and statistic needs to be double checked, no exceptions.

Best Practices for Writing With AI

No matter how advanced they are, tools are just tools and require their users to learn how to best work with them. There's a right way to hold a knife in the kitchen and a proper way to wield a hammer on a construction site. Likewise, there's a right way to work with an AI writing tool. 

Crafting Prompts That Deliver

Getting the best writing out of AI requires writing effective prompts—specific, detailed prompts. Instead of saying "write about the best dog breeds," ask yourself what you mean by "best" and then include that in your prompt: "Write a guide about choosing the right dog breed for apartment living, focusing on temperament and exercise needs." This gives your AI assistant the direction it needs to give you the work you're looking for. 

You can also describe the tone of voice or writing style that you want it to use. Need it to be professional, but not stuffy or overloaded with jargon? You can ask for just that. Some AI assistants—like Type—allow you to upload reference material for style-cribbing. Attach samples of your own writing so you can get something that already sounds pretty close to your natural voice. 

Pro-tip: An AI chatbot interface is both software and the help files. You can ask it for guidance in getting the best response. "Can you give me tips on writing prompts for getting the best responses out of you?" is exactly the sort of question that AI can answer very well.

Oh, the Humanity!

The best way for your writing to sound like you wrote it is to actually write it. An LLM can get you from blank page to first draft in about a minute, which frees your brain up for the fun part of making it all sound good. You don't have to rewrite every word, but you should consistently inject yourself into the narrative.

Add personal anecdotes and specific examples that only a human would know. Throw in some slang. Start sentences with "And" or "But" or "So." Use the phrase "on accident" instead of "by accident." Write sentence fragments. Tell the Oxford comma to bugger off. Break the rules sometimes.

Real people don't always write in perfect paragraphs, and neither should AI content. This is actually good writing advice, whether or not you're using AI: grammar isn't a rule book to be enforced, it's a map to the most efficient ways to be understood. When you're being creative you can absolutely take the scenic route. You don't have to follow the directions, as long as you take your reader where they expected to go.

The goal is to avoid sounding too robotic—an LLM like Claude does better than ChatGPT or Gemini at sounding human, but none of those options are a person. There are some handy tools out there that can help—GPTZero and OpenAI's Text Classifier are pretty good at spotting AI content, though they're not perfect. Not to mention that there’s something counterintuitive about having one machine check another machine’s work to see if it sounds human. It’s weird, actually. If you need a machine to confirm your humanity, that's called a Captcha and they do not work. 

Your best bet: after finishing, read your work out loud. If you stumble over words or phrases, that's probably where the AI is showing through. Look for spots where you'd never actually say it that way in real life. And use tools like Hemingway Editor, which functions to spot overly formal or complicated language, or Type, which lets you highlight a passage and choose to "Simplify" the language—you’re still going to want to put your human eyes on the output, but something like is helpful to steer you in the right direction.

Remember, the trick isn't just detecting AI writing—it's fixing it. Keep a running list of words and phrases you actually use. If you catch yourself writing "moreover" or "in regards to," swap them out for how you'd really say it. Like "plus" or "about." Simple tweaks like that make a huge difference.

Plagiarism Checks Are Your Friend

Even after you've edited AI-generated content to sound more like you, it's smart to run it through a plagiarism checker. AI models are trained on existing content, so there's always a chance they'll spit out something too close to the original source material. Plus, you don't want to accidentally copy someone else's work, even if it's buried in AI-generated text that you've heavily edited.

There are plenty of good tools out there to check your work. Copyscape is probably the most well-known, but you've also got options like Grammarly's plagiarism checker, Quetext, and PlagiarismDetector.net. Most of these have free versions, though you might want to spring for a paid plan if you're doing this regularly. Just remember that no plagiarism checker is perfect—it's a good idea to use multiple checkers on the same piece. Some may find what others do not, and you cover your bases by using more than one.

In Conclusion (see what we did?)

Despite whatever backlash exists toward AI, it isn't going anywhere. But neither is your unique perspective and voice. The key is finding that sweet spot where AI helps you work faster without making you sound like a robot. Think of AI as your writing sidekick—it can handle the heavy lifting while you focus on adding those personal touches that make your writing sound less "AI" and genuinely connect with readers. 

Avoid the common mistakes of AI writing. Play with the AI's responses. Take what works, ditch what doesn't, and reshape the rest. Mix AI-generated content with your own insights. The goal isn't to hide that AI helped—it's to use AI as a tool while keeping your authentic voice. Nobody will second guess your house because the builders used power tools. If you're using AI to help with your writing correctly, no one will second guess your content. It's about working smarter, not harder, while staying true to who you are as a writer.

Keep reading

All posts →

Start writing with Type

Type is the AI-first document editor that helps anyone write high-impact content.

Start writing for free