Re: AI-generated blog post images are not cool any more

13 July 2026

Update: I refined my point of view. Read more: “On AI-generated images: a reflection with myself”.

I came across this article by Panayotis, for whom “AI-generated blog post images are not cool any more”.

It’s not the first article on that topic I have read¹, and I feel particularly targeted, since I use AI-generated cover images for most of my articles, this one included. So, here’s my view on it.

I’m totally fine with it. I think it’s a usage of AI which is fun and even creative. I also think it’s more original and personal than looking for and using an image I would not have created myself.

For that matter, I generally spend quite a lot of time writing a detailed prompt about what I would like to see in the generated image and how it should look, related to the content of the involved article. This usually requires some back-and-forth between the AI and me, before finally generating an image that is the closest to what I pictured in my mind. It’s also not rare to tinker with the image a bit by myself, to refine some details.

Sometimes, I even give the AI personal assets to help it generate my image. Let’s take an example: for the cover image of my article (in French) “La ludothèque idéale de la NES”, I gave the AI a piece of an old picture of my childhood bedroom, where the wallpaper was well visible, in order to reuse the pattern on the generated image. That wallpaper pattern is visible on another picture of that bedroom, on this article: “My teenage years in 12 albums”.

In my prompt, I described my bedroom at that time, the view from the window, what I looked like as well as my friend with whom I spent hours playing NES back then (and who is still one of my best friends today), a particular game to see on the CRT TV, and so on. Again, I had to go back and forth several times with the AI before being satisfied with the final image, which is still not perfect (e.g. the NES is not connected to the TV, I realised it after I published my article), but it’s quite accurate to what I imagined and wished for.

I think I made use of creativity and thinking, and that I didn’t just quickly ask the AI for a random image just to illustrate my article.

I don’t consider that the images I generate with AI are AI slop or crap. Not at all. I spend time on it, I think about what I would like (style, content, sometimes even composition), and it’s part of my publishing process. It’s even a pleasure, once an article is written and ready to be published, to finally reach this image-generation stage of my process, and to see how I will manage to generate an image that fits my article.

I’m not pretending to create an original piece of art that belongs to me, on which I spent hours to put into practice years and years of drawing and photography learning. It’s just a tool, a way to decorate another creative work, for which I’m not relying on AI: writing my article.

As such, I read in the comments of Panayotis’ article (and elsewhere) that the simple use of an AI-generated image might suggest that the article itself was also generated by an AI.

I think that’s a bit of an easy leap to affirm that. I don’t see how looking for and using an image created by someone else (on an image bank, for example), as opposed to using an AI-generated image, would make the text of an article more legit, less suspicious. Writing and creating images are two different things, two different skills.

One solution would be to create the cover images myself from my own photos, pictures, or drawings. Well, it happened (e.g. my “Framing Japan” series) and will happen again. Though, I need to have the right image that would fit my article, and most of the time, I don’t. I could always spend some time trying to take a photo, but I think it would be difficult to take one that would be related to my article. I could draw something, but I don’t have the required talent (though that could be a fun thing to do, to not give a damn about my drawing skills). I could make some collages from several pictures (some mine, some not), gathered here and there (I have the skills to tinker with some graphics software). But I also use AI-generated images, because it fits my needs and I find it fun (being able to generate an image, more or less accurate, based on what I’m thinking about and writing).

Oh! I could also commission an artist for each of my writings, to create a cover image, just for myself, based on the content of my articles. I would then pay the artist what their art is worth, hoping that I would be able to quickly receive my image (because, you know, I usually don’t plan my articles weeks in advance) and that I wouldn’t have to make too many back-and-forth with the artist, to limit costs. Nah. Sorry, but I have no time, no energy and no money to invest in that, just for the sake of my blog. I prioritise other ways to support artists I like.

Another radical solution would be to not use cover images at all. Easy. Why not. But no, I don’t want to impose that on myself, just because a number of people think that “AI sucks”. I like to illustrate my articles with a cover image, and too bad if, in the eyes of some, it’s lame and would question the authenticity of my writings. Plus, it indeed looks nice when the URL is shared here and there.

So, yeah, it’s easy to use an AI whose generated images are, it’s true, the result of the collection and analysis of millions of images created by real human artists, without their permission. It’s not great, I admit it. But it is what it is, I use the tools and resources I have and AI is part of them. I think the way I use AI to generate images doesn’t harm the artists whose original creations were used to feed and train the AI. But maybe I’m wrong.

Anyway, I don’t really know how to conclude this article, but maybe we should moderate our judgements and stop imposing opinions in such a peremptory way. Use stock photos, use your own photos, use your drawings, use AI-generated images, use no image at all… Do whatever you want. It’s your blog, your space on the web, and no one should dictate to anyone how to behave.

That being said, after reading several points of view on this topic, and writing this article, I told myself I should, sometimes, try using something other than AI-generated images, think twice about it and see if another solution would work. But I’m not giving up on AI-generated images.

Update: As AJ wrote in his blogpost about this topic, “clarity around how one uses AI is key”, and that’s why I’m also considering publishing a /ai page soon.

PS: By the way, I have the same opinion for AI-generated code.

ai, blogging, meta, opinion