I built a free calculator that estimates your hosting bill before you deploy (because of stories like the $104K Netlify bill)

Patrick Lin — Sat, 25 Jul 2026 13:16:29 +0000

A few weeks ago I went down a rabbit hole reading about people who got wrecked by their hosting bill for a project that was supposed to be "free tier, no worries."

Two stories stuck with me:

A developer got hit with a $104,000 bill from Netlify for a simple static site after a traffic spike. He only got it waived after the story went viral on Hacker News.
A Vercel user's staging environment got scraped by bots over a weekend, and a $20/month plan turned into a $23,000 bill.

What got me wasn't that these platforms are evil — it's that "Spend Management" and "usage limit" features exist on most of them, but they're opt-in, buried in settings, and there are real reports of the alerts just... not firing when they should.

So before my next side project, I wanted a boring, honest answer to one question: *if this thing actually got traffic, what would it cost me on Netlify vs. Vercel vs. Railway vs. Render vs. Cloudflare Pages?
*
Reading five pricing pages and doing napkin math every time I evaluate a platform got old fast. So I built DeployRadar — a free calculator that takes a few inputs (expected monthly visitors, average page weight, build minutes, whether you're using serverless/edge functions and how often) and estimates a rough monthly cost across all five platforms side by side, sorted cheapest to most expensive.

A few decisions I made on purpose

No account access, no signup, no OAuth. It doesn't need to read your actual usage to be useful — it's a before you deploy tool, not a monitoring tool. You type in your estimate, it does the math against each platform's publicly published pricing rules, and that's it. Nothing to authorize, nothing to connect, nothing to leak.

It's not trying to be precise, it's trying to be directionally honest. The site is upfront that these are rough estimates based on public pricing pages, not a quote — real bills depend on plan specifics, regional pricing, and provider-specific overage rules that change over time. The goal is "oh, Railway would clearly be 3x cheaper for this shape of traffic" not "your bill will be exactly $47.32."

I shipped the smallest possible version first. No dashboard, no database, no user accounts — just a static page with client-side JS doing the pricing math. The whole thing is one HTML file. If people want more (the obvious next step being: connect your actual account read-only and warn you before you go over budget, instead of just estimating ahead of time), there's a spot on the page to say so. I'd rather build that because people asked for it than guess.

Try it deployradar.netlify.app — takes about 15 seconds to get a comparison.

If you've been burned by a surprise hosting bill (or dodged one), I'd genuinely like to hear the story — and if there's a platform I should add, or a pricing rule I got wrong, tell me. I'm actively updating the pricing rules as providers change their tiers.

Schema for AI: The structured data fields ChatGPT actually needs to recommend your product

Patrick Lin — Sat, 25 Jul 2026 07:05:32 +0000

Search Engine Land just published a piece called "Schema for AI search" about using structured
data to show up in AI-driven results. I read it and immediately ran
AIViz scans on a batch of Shopify stores. One thing
surprised me:

Most Shopify stores get their structured data only half right.

Shopify auto-generates Product Schema — name, description, image, and price are covered. But AI search engines need more than that. When ChatGPT decides whether to recommend your product, it also looks at aggregateRating, brand, gtin, and availability. Without these, the AI doesn't have enough confidence to surface your product over a competitor's.

AI search engines don't read your page like Google does

Google's ranking algorithm weighs hundreds of signals — backlinks, domain authority, page speed, user behavior. You've probably spent years optimizing for them.

ChatGPT and Perplexity ignore all of that.

According to Recomaze's research, ChatGPT pulls product information from two sources: Bing Merchant Center product feeds and Schema.org structured data on your page. It uses GTINs
(barcodes) to cross-reference products against known databases. Perplexity crawls your page in real-time and reads JSON-LD Product markup directly.

Your SEO ranking doesn't matter here. If your JSON-LD is missing key fields, AI search engines won't recognize your product.

What Shopify's auto-generated Schema misses

I scanned 30+ Shopify stores with AIViz. A clear pattern emerged. Shopify's Dawn theme (and most themes) auto-generate Product Schema that includes name, description, image, and offers with price and priceCurrency.

Almost universally missing:

aggregateRating — Rating score and review count. HubSpot's Consumer Trends Report found that 72% of consumers plan to use AI for shopping more frequently. When someone asks ChatGPT "recommend a good bluetooth speaker," products with rating data get prioritized.

brand — The brand name. Sounds basic, but most themes display the brand on the page without including it in the JSON-LD. AI crawlers read the structured data, not your visual layout.

gtin / mpn — Product barcodes or manufacturer numbers. ChatGPT uses these to cross-reference against Bing's product database. Products with GTINs are matched more accurately.

availability — Stock status. You don't want AI recommending an out-of-stock product to a potential customer.

I assumed at least brand would be common. After scanning, only 4 out of 30 stores had it correctly marked in their JSON-LD.

How to fix it

The good news: you don't need to edit your Shopify theme's Liquid templates (though that's the cleanest approach). The fastest path is injecting additional JSON-LD via a Shopify App.

But first, figure out what's missing. Go to AIViz, paste your product page URL, and the report will show you exactly which Schema fields need attention — with copy-paste fix code.

For example, if you're missing aggregateRating and brand, AIViz generates this snippet:

  {
    "@context": "https://schema.org",
    "@type": "Product",
    "brand": {
      "@type": "Brand",
      "name": "Your Brand Name"
    },
    "aggregateRating": {
      "@type": "AggregateRating",
      "ratingValue": "4.5",
      "reviewCount": "128"
    }
  }

Add this to your product page, and AI search engines can read the complete picture.

Beyond Product Schema — two things people forget

After fixing your structured data, there are two often-overlooked items:

robots.txt: Make sure you're not blocking AI crawlers. Add this to your robots.txt:

  User-agent: GPTBot
  Allow: /

  User-agent: PerplexityBot
  Allow: /

Shopify's default robots.txt usually doesn't block AI crawlers, but if you've customized yours, check.

llms.txt: A Markdown file at your site root that helps AI crawlers understand your site structure. AskNeedle's research found that 83% of AI citations come from pages updated within the past 12 months. An llms.txt file signals that your site is active and maintained.

What happens to your score after fixing this?

From what we've seen, filling in missing Schema fields typically moves an AIViz score from 30-40 to 80-100. The biggest score jumps come from Product Schema (20 pts) and aggregateRating (10 pts)
— that's a 30-point swing from two fields.

But the score is just a proxy. The real difference: AI search engines now have enough information to recommend your product instead of skipping over it for a competitor who did the work.

The one thing to do right now

Scan your best-selling product page. Check which Schema fields are red in the report. Fix those.

Free scan: AIViz — AI Visibility Checker
Shopify users: SEO Checkup scans your entire store at once

Sources: Search Engine Land, HubSpot — AEO Trends 2026, Recomaze, AskNeedle.

AIViz is open source — GitHub repo. Star it if this was useful!

DEV Community: Patrick Lin