Creating a site using AI is probably not going to generate h1, h2, h3 tags, probably not going to introduce LSI keywords and probably not going to bold some of the content that would make Google smile.
I would not do that for myself nor for a client in that manner as it is a shortcut that rarely works, and even then, it only works for a short time if you are lucky.
I've been thinking a lot about what OP was talking about. There's not enough details to know exactly what he's thinking in terms of execution, also not sure how much automation he means. Like partial automation, section by section, fully automated site generation? idk.
GPT does create H1, H2, H3 tags, and it will do bold, and keyword focus, and image alt tags etc. It can do pretty much whatever HTML and CSS. But not necessarily in 1 output. I like using it with wordpress shortcodes, I have my autoblogger do basic things it can't mess up like doing formatted stylized tables via shortcode that adds the styling. There's good shortcodes for tabbed sections, JS/css F.A.Q. sections, you can even do content sliders and rich features like that.
I spent an hour one time going back and forth with GPT to make a landing page style product descriptions with rounded rectangle images and text sections with CSS and HTML from scratch, including mobile responsive aspects. ai understands pretty much whatever HTML and CSS, but is limited by output size, requires more processing than plain text.
Anyway, I was looking into doing a similar thing, but so far I've opted instead for wordpress + autoblogger, though of course I do a good bit of manual set up as well with theme and plugins and do the review pages and stuff mostly manually(also homepage header image with photoshop, and logo which I use ai for but I don't just use the first output I describe a design and keep going till I like the result.
You could feed it a static HTML/CSS full site template, and give it some image URLs and stuff, and you could get a decent homepage out of it, but it's not like it can populate with a lot of content a good 5 page site in one output. The 4,000ish token max per output with API can do individual pages ok, but multiple pages stretches it too thin, and on occasion may screw up, if they upped to like 50,000-100,000 tokens, a decent full site with navigation would be doable. Though I've pretty much abandoned/put on the backburner the idea of doing static HTML sites for now, wordpress can just do so much more.
Though, a multi-output workflow can be automated with something like Make.com(or maybe if you know python or something, idk I don't), which can interphase with different programs including GPT API, and adjust and change prompts based on information from a spreadsheet and stuff, you can potentially automate a near full workflow to do a full size small site with ai.
Though really, they won't do any good if we're talking like 5 page very basic sites, would just have 100 sites that nearly never get any real visitors.
While beyond my abilities, it would be possible to do it properly with batch outputs. In which case it would be worthwhile if we're talking like a 50 page site, with ai generating article topics(which, it can do a pretty good job of, but needs configuration).
As for the person who said they made sites with 30,000 posts and made $1 a day, I'm curious what niche. I have 28 sites now, none have anywhere near that many post, though I've had some results, so idk, I think with 30,000 posts I could kill it, like good lord, personally I think 50 decent posts with good content strategy is enough for some sales or leads, though idk if my niches would have 30,000 topics to even talk about. Also that'd cost a fortune in api credits, my 4 part articles even with 4o and 1 or 2 images still are $0.10-$0.20. Also, shit, does google not index most of it? Because otherwise, makes more sense to build a lot more sites with fewer posts if the 30,000 posts have diminishing returns vs a much smaller site with a similar domain with like 50-200 posts(sorta like what I've been doing). If it's just random ass content and no focus, then yeah, wouldn't work, but idk, sounds strange to me you didn't get a better result.
I still think there is good potential in the idea, depends on execution.
If the sites look good enough, are coherent and well organized, domains are well targeted for rankable searches, and there's good topical coverage, then yes it'd work. But that's harder said than done. 100 trash sites will do fuck all. But if you can configure it well, and each site domain and content is specific to a long tail or something yes, and I'd also be interested. I have a good workflow going, but I still need either better performance on existing sites or a lot more of them to get where I want.
The real value I see in this (and what I've been working on of a smaller scale and also, not like fully automated, again not sure what level of automation you want vs how much hands on configuring you mean) is making tons of sites with exact match domains for low competition searches. Not with the goal of tons of visitors, but a trickle of ones very targeted.
Also, if the only reason for static is cheaper hosting, seems like a big project to be worried about the hosting. I have an unlimited site hosting plan from cenchu(saw them advertise on here, it's wayyyy faster than the namecheap turbo hosting I was using before) for $120 a year. If the goal isn't like hundreds of visitors a day on the sites, but more like a google click here and there(depending on the niche and value of the leads, some niches you just need 20 targeted visitors in a month to get a lead and if a high value niche then it's worth it, other niches/business models you need thousands of visitors to make anything). Otherwise, if you were able to do it with wordpress, for one, I'd be interested how if you have a programming way to automate it, and for 2, if he is spending $1,000+ a year on domains alone, is really the hosting cost that much of a problem? If the goal is a small amount of visitors for specific niche searches, I don't see how that'd be a problem, say cost hypothetically as 1 very large site really.
Really depends on a lot of factors. If we're talking like very local specific domains for lower competition searches, I think it could be worthwhile.
