Just 'cose you can it does not mean you have to. And given your current situation you really should not.
@Henn91 I took your advice and downloaded opencode, put $5 on deepseek, skipped over the ICP stuff for now. I re-created a customer reviews page from my site using a single prompt in opencode, along with a csv file. I have to say I’m astounded what it was able to do, and it only cost $0.04 of the $5. I think I’d be a fool not to pursue this.
This is the way. Once you are ready to explore ICP just tell your AI to go to https://skills.internetcomputer.org/llms.txt.
I’ve been looking at the different options, but it seems, especially because of my need for e-commerce shopping cart, that I’m going to have to stay with Wordpress for the time being.
Where do you host your cart?
Woocommerce? And you need a method to migrate to the ic, depreciate your web2 accounts, and be entirely on icp?
Where does your “items and cart” exist? Inside woocommerce or magneto?
@Henry_Suso Yes, it’s a Wordpress site with WooCommerce shopping cart, and Elementor for the page builder
You definitely can leave traditional webhosting behind. Just make sure you actually need your site replicated across the subnet.
@WebTreeSoftwareSolut I’m not entirely sure what you mean by that?
Did you mean @WebTreeSoftwareSolut or @Dionysus ?
Because honestly, I’m not quite sure which one you’re referring to anymore
It’s funny you say that, friends.
I know exactly what you mean.
@WebTreeSoftwareSolut has entered the chat and somehow knows everything 
I wouldn’t say that.
Nobody knows everything.
What the heck are you guys talking about?! I was referring to @WebTreeSoftwareSolut 's comment where they said:
“Just make sure you actually need your site replicated across the subnet.”
What I meant by that is simple: on the Internet Computer, every node on a subnet has to replicate and execute the data you deploy to a canister.
If your site doesn’t actually require decentralization or tamperproof guarantees for every single asset, paying the performance and cycle cost to replicate all of it across an entire subnet might be overkill.
It really comes down to whether the site’s use case justifies full on chain replication.
You definitely can do it.
But should you?
Only you can answer that.
Thank you, I understand now.. I was just trying to think ahead. With all this talk about AI being able to hack into systems better and faster than humans, and Wordpress in general being a big target. I thought perhaps it might be a good idea to think about moving my business off WP, and onto the IC where it might be safer. Not that I think I’m being targeted or anything. But suppose one day we wake up and some AI replicated itself a bazillion times and hacked into and destroyed all the Wordpress sites on the internet; that would suck.
I do wonder why your site is so huge secret that you cant share it. If you make money with it, wouldnt it be useful to actually promote it.
There’s no reason to share it here and I like to keep it private. It’s enough to say what kind of site it is, and the tech stack that it runs on in order to address my question. And you have addressed my question, so many thanks for that.
I run several of these in production, a handful of sites plus a couple dozen canister projects, and one of them takes real money from real customers. I won’t name the domains, and that’s worth explaining: on the IC the canister pays for its own update calls (reverse gas), so a public canister ID plus an anonymous endpoint is a cycle drain target. Every site leaks its canister ID in a response header anyway, but I’d rather not hand out a curated list on a forum this technical. It isn’t a knock on the IC. It’s just the one thing that genuinely changes when you host this way.
On your questions: all of it is doable. My pipeline is Figma straight to canister with Claude Code, no CMS anywhere in between. One rule saved me there: build against the exported frame renders, not the Figma node tree JSON. My first attempt used the JSON alone and came out genuinely bad, wrong colours per section, missing buttons, wrong nav. There are a lot of small tricks like that, and you do pick them up.
What you get: one of my replacement sites is 95 KB against 567 KB for the WordPress original it replaces, and it scores 100 Accessibility, 100 Best Practices and clean SEO with no CDN in front of it at all. No plugins, no database, no PHP, nothing to patch, no admin login to brute force. That is the real answer to your burnout. There is simply nothing left to maintain.
Two things to plan for. Latency: there are no IC boundary nodes in Oceania, so from Australia the raw IC gives me 700 to 1000 ms TTFB. Put a CDN in front with an IPv4 only origin and it drops to 270 to 370 ms. Payments: outbound HTTPS calls from a canister are replicated by default, and I watched a single “create payment session” call turn into ten orphan sessions at the gateway, one per node, because consensus compared ten different responses. Use a non replicated outcall for the ephemeral part and let the webhook carry the money truth into replicated state.
Also, keep every .icp0.io and icp-api.io URL out of your pages. Your front end can call window.location.origin for everything and the boundary node routes by canister ID in the path, so no third party host ever appears in your HTML. That matters if you run Google Ads and want landing page consistency.
Fair warning: it’s a one way street. Once you’ve shipped a site this way you will never open WordPress again.
One more thing I should have put in the post above, on cost.
While your app has few users, nothing beats a canister on mainnet. It is close to free, and there is no monthly bill sitting there waiting for you whether anyone visits or not. As it grows you start paying something more realistic, and how much depends on what you accumulate and how you developed it, because storage is the part you keep paying for forever.
The good part is that it stays in your hands. You can always work the code, or change where the data lives: keep the canister heap for what is actually hot and move the rest out, to stable memory, or to object storage behind a CDN for anything big. On the IC the bill is an architecture decision, not a plan you are locked into.
From your whole post, this was the only part I understood 