I swear to God if I read “In the era of” one more time…
Now I’m no AI doomsayer and admit that it’s helpful with text if it’s summaries, suggestions, or even copywriting to an extent. But to directly use its outputs for blog writing not only nets poor results, but is eroding the internet. The internet, on which the next generation of AI models will train, leading to a snake-eat-snake cycle.
Blogs have been one of the major sources of information since forever so I’ll mainly be talking about AI’s use case in blog writing specifically.
Where it’s been helpful
Before I start pointing out the bad, I appreciate where it’s adding value, and changing how things used to be before it. Let’s take meta descriptions as an example.
In early 2022, a part of my job involved editing and readying the articles for publication in WordPress. One of the things to add was meta descriptions that give a brief view of what the article is about to both GoogleBot and readers.
Now we didn’t have ChatGPT yet, so you had to actually use your brain for some copywriting at the time. You could either use:
- Problem -> Solution framework where the solution is in the article.
- What the reader will learn, including keywords from the title.
- A catchy but strong statement from the body of the article
and so on…
Now? Just paste the article text in any chatbot, and it will spit out various versions in mere seconds. I’m not complaining, but grateful for this since there was no guarantee it was going to be used anyway.
GoogleBot decides if it's going to use your meta text or spin its own via machine learning and algorithms. So giving it an output created by a similar process not only makes sense but saves time too.
I do remember early ChatGPT being terrible at this, but the quality has improved so much now that using AI for summaries, meta descriptions, and alt text IS the best way to do these things.
AI in blog writing
First, a little reminder about blogs' purpose that they aim to serve information that adds value to readers. Every blog has the opportunity to add value to a topic even if it’s been talked about before. You could add updated information, more examples, tips, custom images, better reference links, or anything to an existing topic, and now your article is uniquely adding value.
Enter AI in blog writing, all LLMs scoured the same internet and mostly have the same training data. If an LLM has unique training data like ChatGPT and Gemini have exclusive access to Reddit… well, use these models as a learning model, and now you have indirect access to this data!
My point is that all these LLMs, no matter what company, have access to the same set of training data directly or indirectly. When you prompt AI to output paragraphs to use in your new article, it’s just summarizing its training data in an orderly manner.
Now, the first person to use this output could say that their article is unique, but the hundreds and thousands doing it after will get similar outputs, no matter what prompts and temperature is used since the training data remains the same.
In the end, you’re not adding any value, and your readers are probably better off getting information from AI directly without any ads, popups, or distractions on your blog.
So why is AI still used?
Okay so AI sucks at blog writing but I can’t close my eyes to realities that make people use it anyway.
Ridiculously cheap
Even if the headline makes it obvious, I want to additionally highlight the continued death of the blogging industry.
Blogs mainly use ads for revenue generation so they’re highly dependent on traffic. But ads are distracting so it’s not a surprise that people are moving to ChatGPT, Claude for information. So blog traffic is and will continue to go down.
But less traffic means ever-decreasing budget to create more blog content which means your blog is now dying without updates. Now between no content and content semi-generated by AI, choosing the other option doesn’t seem too bad does it? I hate it but it’s just how things are now.
Less confidence
Some people are too afraid to make mistakes and don’t realize that English without perfect grammar is still understandable, even better sometimes. As long as it makes sense in your head, it’s probably fine as is.
This mostly affects non-native English speakers and explains why some writing services proudly display their writers English-speaking heritage. But really, you don’t need to put every sentence in ChatGPT or Grammarly to fix it. Making mistakes is fine, it doesn’t show your lack of care but proof of your humanity.
We also forget that the time we spent prompting to get the perfect one-shot could’ve been spent writing instead. So many times these LLMs won’t produce the output you’re envisioning so you keep trying and wasting time rather than believing in yourself.
Just be done with it
Being able to express your thoughts in writing is one of the most valuable things to learn, but it does take time and effort. When you can’t or don’t want to spend it, having AI spit out an article in minutes starts to look like a good idea. This is why we see AI’s use in writing everywhere we look; social media, emails, blog posts, you can’t escape it.
At this point, just like we eventually learn to ignore ads, we have been getting better at spotting and ignoring AI content. I deeply resonate with Alex’s post when he says, “It's rude to show AI output to people”. It’s just insulting that you’re showing AI-generated text and expect a human response, like at the very least, please edit the output.
It doesn’t please me to say that most pitches we receive for guest posts follow this pattern. But I’m sure this is the norm with every site that allows a submission; whether it’s looking for resumes, pitches, outreach, or anything.
AI writing is an unescapable reality
Despite my feelings or those who agree with me, AI writing is a reality, and we’ll never go back to how things used to be (...when spam used to take some effort?)
I also find it hilarious that Google used to have a policy against AI generated content in 2022, but soon the hard question hit them, “Wait, how are we going to detect it?”.
To this day, there’s no reliable method to check AI-generated content without false positives. Eventually, they reversed this policy in 2023 with a promise to reward quality content irrespective of how it’s made.
AI content detectors are borderline useless too with their false positives and unreliable accuracy and a breakthrough seems doubtful. Unlike AI-generated images that have metadata, text is easily manipulated, so that just leaves spotting patterns which is unreliable and easily fooled if need be.
It's unfortunate but we now have to live in a world where words carry less weight than before, where you aren't sure if these words actually carry a thought or were spit out by a machine.
Disclaimer
It would’ve been funny if this post were generated with AI but no, I didn’t use it at all. I should also mention that these opinions are of mine alone and don’t necessarily reflect SitePoint’s editorial policy or its published posts.
If you enjoyed reading this, you might want to consider joining our newsletter. I do use AI's help for summaries in there but it's still a good way to keep up with trending articles and tools in the dev space.