The AI Slop Tax Is Now a Conversion Problem for Cross-Border Sellers
Every cross-border operator I know has quietly become a web developer. Not by choice — by necessity. You spin up a Shopify landing page for a US test, a Shopee storefront for Southeast Asia, a TikTok Shop link-in-bio page, a Temu-adjacent brand site to escape marketplace fee compression. And increasingly, you’re generating those pages with Claude, Cursor, Lovable, or Bolt because hiring a designer for a two-week ad test is economically insane.
The problem is that everything you ship looks the same. Same gradient hero. Same three-column feature grid. Same “modern and clean” nothing. That sameness is not just an aesthetic complaint — it’s a measurable drag on conversion, on paid social CTR, and on the premium pricing you’re trying to command outside Amazon. So when a tool launches explicitly to fix the input to AI site generation rather than the model, cross-border sellers should pay attention. That’s what Thanor AI is pitching.
What Thanor AI Actually Solves (And Why It’s Not a Website Builder)
The maker’s framing is unusually honest for a Product Hunt launch. He built a website builder first, watched everything it produced look like everything else a website builder produces, and deleted it. The lesson he landed on: the model was never the bottleneck. Hand Claude or Cursor the same brief twice and you get two very different pages — the variable is how much art direction went in with the brief.
That’s the entire thesis. Thanor is not a builder, not a no-code platform, not a Figma plugin. It’s a library of full art directions — palette, type scale, motion, and section-by-section structure — written to be pasted whole into Claude, ChatGPT, Cursor, Lovable, or Bolt rather than trimmed down. The pitch line on the launch page is blunt: “Everything built with AI looks built with AI. Thanor fixes the input.”
For a cross-border seller, that reframing matters more than it first appears. You are not buying a page. You are buying a reusable prompt asset that changes the output quality of every AI tool you already pay for. That’s a fundamentally different unit of value than another subscription landing page builder.
The MCP server is the sleeper feature
Buried in the launch copy is the detail I’d actually watch: a new MCP server that pulls any prompt directly into your editor. If you’re running a Model Context Protocol setup inside Cursor or Claude Desktop, this collapses the copy-paste loop entirely. For teams shipping localized landing pages across five markets per week, that loop is where the hours die. Not disclosed is whether the MCP server is included in all plans or gated — worth asking before you commit a workflow to it.
How It Compares to the Incumbents You’re Already Using
Let me be precise about the competitive set, because “AI website tool” is a meaningless category in 2026.
Against v0 by Vercel: v0 is a full-stack vibe coding platform. It generates and deploys. Thanor doesn’t deploy anything — it feeds v0 better inputs. These are complements, not competitors, and the smart play is stacking them.
Against Dora AI and Relume: Dora generates sites from AI; Relume generates wireframes and components for Webflow and Figma. Both output artifacts. Thanor outputs direction. If you’re a Shopify merchant who needs a hero section rebuilt by Thursday, Relume is closer to your job. If you’re briefing an AI agent to build a whole DTC brand site, Thanor is closer.
Against Sivi AI: Sivi generates instant designs — closer to a graphic design tool than a direction library. Different job entirely.
Against the default: pasting “make it modern and clean, premium feel, dark mode, SaaS aesthetic” into Lovable and hoping. That’s your real incumbent. Thanor’s bet is that sellers will pay to stop rolling those dice.
Why Amazon sellers should care more than Shopify ones
This is counterintuitive, so let me defend it. Pure Amazon FBA operators rarely own their storefront — Amazon Seller Central gives you A+ Content modules and a Brand Store, and both are heavily templated. You don’t get to art-direct much.
But the moment an Amazon brand owner starts hedging — building a DTC site to escape referral fees, running TikTok Shop funnels, testing Temu or SHEIN as alternate channels — they need brand-grade web assets fast and cheap. They have zero design infrastructure. They have a folder of product photos and a supplier. Thanor’s paste-a-brief model is built for exactly that operator: someone with taste-by-instinct and no design team.
Shopify-native brands, by contrast, usually already have a theme, a designer on retainer, or a Figma file. Their marginal gain from a prompt library is smaller.
Where the math breaks
Two honest problems.
First, art direction is not the same as brand strategy. A palette and a type scale pasted into Claude will make your page look less generic. It will not make it yours. If you’re selling a commodity across six marketplaces, “less generic” is a real win. If you’re trying to build a defensible DTC brand, you still need a human who understands your customer. Thanor accelerates execution; it doesn’t replace positioning.
Second, prompt libraries decay. Models update. Claude’s rendering of a given prompt in six months will not match today’s. A static library of directions needs active maintenance to stay useful — and the launch page doesn’t say how versioning against model changes will work.
What Cross-Border Sellers Can Actually Borrow From This
Even if you never buy Thanor, the underlying insight is portable and worth stealing this week.
Treat your brief as a product. Most sellers write briefs like texts to a freelancer. The operators getting good output from AI tools are writing briefs like design specs — motion, spacing, section order, tone of voice, reference sites. That discipline alone moves output quality more than switching tools.
Standardize your art direction per market. A US DTC page and a Southeast Asian TikTok Shop landing page should not share a palette. If you’re running localized funnels, build a reusable direction doc per region and paste it into every generation. This is the exact use case Thanor is productizing.
Separate direction from execution. Direction is what you reuse. Execution is what you regenerate. Most sellers conflate them and end up rewriting the same brand brief fifty times.
Watch the cloning tool. The launch page promises a tool that turns a site you admire into a design of your own. For sellers benchmarking against competitors, that’s the feature with the sharpest edge — and the most legal ambiguity. Not disclosed is how “of your own” is defined.
The community and course angle is underrated
The maker is also building a course on Skool and YouTube about using the library to build sites that look like they cost five figures, plus a contributor program paying designers to build directions. For a cross-border seller, the contributor angle is interesting: if you’ve built a converting landing page for a specific vertical — pet supplements, phone cases, kitchen gadgets — you could theoretically monetize that direction as a template. That’s a side revenue stream most operators never consider.
What I’d Watch / Test Next
Three concrete things to do this week, regardless of whether you touch Thanor.
One: Pick your single worst-performing landing page — the one with the highest ad spend and lowest conversion — and rewrite the brief from scratch as a full art direction. Palette, type, motion, section order. Paste it into your AI tool of choice. Compare against the current version. This costs you two hours and tells you whether the direction thesis holds for your niche.
Two: If you’re already running Cursor or Claude Desktop with MCP, test the Thanor MCP server on a throwaway page. The launch page says 50% off all plans, but pricing tiers and the MCP gate aren’t disclosed — so verify before you build a workflow around it.
Three: Before committing, pressure-test the model-decay problem. Ask the maker directly how directions will be versioned as Claude, GPT, and Lovable models update. If there’s no answer, treat this as a six-month tool, not infrastructure.
My read: the diagnosis is right, the wedge is smart, and the execution risk is real. For cross-border sellers burning ad budget on pages that look like everyone else’s, the upside of fixing the input is worth the experiment.
