It’s no secret that writing blog posts has become easier with the help of LLMs like ChatGPT and Gemini. You might have even noticed we did a little experiment and uploaded a fully AI-generated post earlier. We wanted to see what kind of reactions it would get but also use it to make a point in this post.
It was brought up during an Office Hours livestream. One of our attendees was concerned that the blogs on a site they manage were deindexed due to AI-generated content. They wanted advice on how to use AI more effectively for blog writing.
Now, if you read our AI-generated post, you might have gleaned some decent information. Read on here, though, for in-depth, applicable advice on how to write using AI.
The initial concern
The whole reason this blog exists is that the content our viewers mentioned was deindexed because it’s AI-generated. Here’s the question itself:
What are you advising clients on how to create blog posts with ChatGPT? Having a potential client ask why the content is deindexed, and the content is obviously done by AI, and has been confirmed.
So what’s the problem with using AI for content? This site is undoubtedly not the first one to do it. Let’s explain.
The problem with using raw AI content
Content ranking poorly or being completely deindexed because it’s purely AI-generated is nothing new. Google’s AI guidelines focus on EEAT, which stands for Experience, Expertise, Authoritativeness, and Trustworthiness. In short, they prioritize original, helpful, people-first content created to help readers, not to manipulate ranking algorithms.
Unreviewed AI blogs like the one before this (or any content, really) are typically low-value, generic, and mass-produced. They also can:
Contain incorrect information.
Miss out on important context.
Sound nothing like you, the author.
Repeat the same thing but with different words over and over.
Use many words to say very little. Bloated.
So, mystery solved, we know why it happened. Entirely AI-generated content isn’t great for your site. But does that mean you should never use AI to help you? No, not at all. You might have noticed we said “Unreviewed AI blogs” in the previous paragraph; it’s an important distinction.
AI is your assistant, not your writer
Nathan, our Office Hours host, explained the purpose of AI in blog writing perfectly with one sentence: AI is your writing assistant, not your writer. He then further elaborated that AI works best when directed.
Instead of asking AI to do all the thinking for you, use it to organize your own thoughts and workflow. It can help with many things.
Helps you start: Those first few sentences in a blog post are the hardest. Leverage AI to give you a starting point or ideas that you can then expand into your own intro.
Speeds up drafting: An LLM can spit out a blog draft in seconds. Use that to spark your own creativity and mold it into something that reflects your personality and writing style.
Tighten your grammar and punctuation: AI can spot such errors in an instant. Missed commas, spelling errors, incorrect tenses, or anything else you can think of, it can point them out to you.
Summarizing: If you are struggling to explain a complex idea or concept in simple terms, a chatbot can help you immensely. It will boil it down to a few sentences you can then write in your own words.
All of the things we mentioned are productivity boosts. AI can speed up your workflow and help you be more efficient with your time. However, there are a few things you should be wary of that AI isn’t great at.
Producing original content: That’s simply something it struggles with or can’t do at all at the moment. It knows how to use already existing content to piece together its own replies.
No expertise: An AI bot has no subject-matter expertise. It cannot replicate the personal experience of someone in a given field. It can make up anecdotes if you ask it to, but it’s incapable of replicating a person’s life experience.
Correctness: We would never advise anyone to explicitly trust any information provided by AI, unless it’s an absolute certainty; the boiling temperature of water, for example. We encourage you to research any information you need beforehand from reputable, proven sources.
Replace your tone of voice: Sure, it can mimic your writing style and tone, but it can’t fully replicate it. As writers, we all have our own quirks and stylistic choices that readers can often notice.
Knowing what AI is good and bad at can help it help you be a better writer. It’s not cheating or disgraceful to use the tools at our disposal to improve ourselves. With that in mind, let’s round out this post with a few valuable tips on how to do that.
Guidance and direction
An LLM like ChatGPT can significantly help you write if you are there to work alongside it. It should never be the driver of your writing process. AI content is best when treated as a spark for your own creativity.
Instead of asking the bot to write everything for you, use it to brainstorm and get a feel for the blog post you want to write. Never let it do the driving. If you are uncertain how exactly to interact with a chatbot, then our livestream on mastering prompt architecture is the perfect starting spot for you. With that in mind, here are a few bits of advice on how to best leverage AI when writing your blog.
Give it direction
We mentioned earlier that AI often misses important context when writing. That’s because it’s working with limited information. “Write me a blog post on <insert topic here>” is a terrible prompt that will yield terrible results. It will spew out something generic.
However, if you provide additional information about your audience, the article's goal, the tone you want, and key points, it will produce a better response. Instead of generic, it will be more targeted and useful.
Start with an outline
Instead of asking AI to write you a complete draft, try talking to it first about the topic of your post. Try these few things.
Brainstorm key points with the bot about the topic.
Discuss which points would matter most for your target audience.
Build a rough outline with active input from you.
Keep refining the structure until it feels right to you.
Once you have the outline and flow nailed down, you’ll see how easy it is to begin writing yourself. You probably won’t even need the LLM to generate sections of the draft for you.
This process of introducing the bot to your topic and talking it through adds to the context it needs. It helps narrow down its otherwise generic answers.
Editing is non-negotiable
If you do decide to have the LLM draft a blog post for you, you should never publish it raw. It’s during this phase of the process that you can ensure the text fits your personal preferences.
This is where your experience, opinions, and perspective will matter the most. AI can’t replace those things, nor should you ask it to. Close your browser tab, take a break, then come back to your piece with fresh new eyes.
Edit it until it feels right to you, as if you had written it yourself. And once you click publish, you can rest assured that it won’t get flagged as useless content by Google.
AI isn’t cheating
Using AI to help you write better content isn’t inherently bad. In fact, it can free you up to focus on what actually matters most in the process: ideas, style, tone, and value for your readers.
The issue arises when unedited AI content gets published. It’s heartless, soulless, and is all about quantity over quality. SEO is no longer about that; quality is the bar now. AI isn’t a shortcut to attracting traffic (or thinking). It’s a part of a thoughtful content process that, when given direction and context, can help you improve as a writer.
And if you have faced similar issues with AI content, have a question regarding running an agency, or anything else related to web hosting and websites, register for Office Hours. Our host, Nathan Ingram, and our incredible community will be happy to help you.
FAQ
If I edit AI-generated content thoroughly before publishing, is there a risk Google will still detect and penalize it?
The risk isn't in using AI; it's in publishing content that is generic, repetitive, thin on substance, or clearly written for search engines rather than readers.
Google has been clear that its guidance targets low-quality, unhelpful content, not content produced with AI assistance that has been meaningfully edited to add genuine value. A thoroughly revised draft that incorporates your own expertise, real-world experience, accurate information, and distinctive voice is qualitatively different from raw AI output. Google's systems evaluate quality signals rather than simply trying to detect whether any AI was involved in the production process.
Does the type of blog post matter when deciding how much to rely on AI assistance?
Yes, and this is a useful lens for deciding where AI earns its keep. Informational or explanatory posts ("What is X?", "How does Y work?") are areas where AI can produce a reasonable structural draft, since this type of content is well-represented in its training data.
However, posts that require genuine first-hand knowledge like case studies, opinion pieces, experience-based tutorials, or anything where a specific viewpoint or lived expertise is the whole point, are areas where AI assistance is most limited and where your own input carries the most weight.
How should I handle it if existing AI-assisted blog posts on my site have already been deindexed?
Deindexation for content quality reasons is recoverable, but it requires genuine improvement rather than surface-level changes. Simply running the posts through AI again to rephrase them won't address the underlying issue. Google's systems are evaluating whether the content is genuinely helpful, not just whether it appears original.
The practical steps are to audit the affected posts, identify which ones are worth salvaging (based on whether they cover a topic with real search demand and audience value), and substantially rewrite them with real expertise, accurate information, and a distinct perspective.
Are there specific AI tools that work better than others for blog writing assistance?
The choice matters less than how you use any given tool, but there are meaningful differences worth knowing. General-purpose large language models (like ChatGPT, Claude, or Gemini) are strong for brainstorming, structural drafting, and editing assistance, and all have broadly similar strengths and weaknesses for this use case.
The more important variable is the quality of your prompting and the depth of your editorial involvement: a mediocre tool used well, with rich context and thorough editing, will consistently outperform a sophisticated tool used as a one-click content generator.
