TL;DR: PC Deal Check reads a retailer’s PC listing with AI, then grades the asking price against its own component price estimates. Individual checks are free in the US right now, need no account, and produce a dated report with a lettered grade. It is a price reference rather than a benchmark, and every estimate carries a 10 percent band.
What PC Deal Check is
PC Deal Check is a small web service at pcdealcheck.com that answers one question: is this prebuilt PC a good deal? You paste a retailer listing link or upload a screenshot of the price and specs, the service works out which parts sit inside the machine, and it compares the asking price against market estimates for those parts.
The split is worth noting. AI handles the reading step, pulling the GPU, CPU, RAM and storage out of whatever mess of a listing you hand it. The grading comes from a hand-built price scanning engine that the creator says draws on market data including eBay listings with quoted shipping. It is a self-financed prototype, and both the engine and the scoring bands are the parts most likely to change.
The origin story sits on the About page. Adam, a former IT professional who now builds digital twins of cell towers, once found a good deal but spent too long checking component prices, and by the time he got around to buying, the deal was gone. The service exists to compress that gap.
Running your first check
The check page at app.pcdealcheck.com/check accepts three kinds of input: a pasted listing URL, a screenshot, or pasted text. A retailer link beats a screenshot because you skip the OCR step, and screenshot cropping runs in your browser, so only the region you select is uploaded.
Supported listing sites are shown right on the form:
- Amazon
- Best Buy
- Walmart
- Newegg
- Micro Center
- eBay
- Woot
- Sam’s Club
- Costco
The creator said in the HN thread that coverage runs to roughly the 20 biggest retailers, with Canada planned after the US.
Retailer links often carry tracking parameters and bounce through a couple of redirects, so if a paste fails to resolve, copy the canonical product URL out of the browser address bar rather than the share link. A quick command line check tells you where a URL actually lands:
curl -s -o /dev/null -w '%{http_code} %{url_effective}\n' \
-L -A 'Mozilla/5.0' 'https://www.microcenter.com/product/...'
The check itself is free and needs no account. Usage limits apply to keep the service available, so a check may ask you to wait or use a different input method if you push it hard.
What the pricing engine produces
The published sample report is a PowerSpec G531 Gaming PC from Micro Center, listed at $1,399.99 on 1 November 2026, and it doubles as the best documentation of how the engine works. It identified four major components and priced each one separately.
| Component | Estimate | Pricing basis |
|---|---|---|
| GeForce RTX 5070 | $800 +/- 10% | eBay listings, checked 1 Nov 2026 |
| Ryzen 5 7500X3D | $290 +/- 10% | reviewed by the team 27 Oct 2026 |
| 16GB DDR5 | $230 +/- 10% | new-part price, checked 1 Nov 2026 |
| 1TB SSD | $100 +/- 10% | eBay listings, checked 1 Nov 2026 |
| Case, cooling, motherboard, PSU | $150 +/- 10% | one combined allowance, not priced individually |
Every estimate carries a 10 percent band, a date, and a stated basis, and every component gets price-check links out to eBay and Google. Those links follow a fixed pattern, so you can reproduce a check by hand whenever you want to see the same market yourself:
https://www.ebay.com/sch/i.html?_nkw=GeForce%20RTX%205070&LH_BIN=1&_sop=12
https://www.ebay.com/sch/i.html?_nkw=Ryzen%205%207500X3D&LH_BIN=1&_sop=12
https://www.google.com/search?q=1TB%20SSD%20price
The eBay parameters are worth decoding: LH_BIN=1 restricts results to Buy It Now listings, and _sop=12 sorts them by price from low to high, which is how you find the cheapest current copy rather than the most popular one.
Two rows in that table deserve a second read. The RAM estimate is based on a new-part price, so a prebuilt carrying 16GB is valued against what new 16GB costs, not what used 16GB sells for. And the last row is a single $150 allowance covering case, cooling, motherboard and power supply, which means a build with a $250 motherboard gets no extra credit for it. The report does call out the motherboard when it can identify one, noting a Gigabyte B850 Gaming WiFi6 and flagging that its features might be worth additional value.
The grade itself is a comparison between the asking price and the estimated parts value. The sample came back an A, described as great value at the checked price, because the $1,399.99 asking price sat 10.8 percent below the $1,569 parts estimate, a gap of $168.63. A and B are the two grades the deal alerts plan to notify you about.
What to watch for
This is a price reference, not a performance test, and the report says so plainly at the bottom of the page. Nothing in it tells you whether the machine hits 60 frames in a game or whether the storage drive is about to fail.
The used-PC gap is the biggest one. Estimates can use new or secondhand market prices, yet used and refurbished machines are graded with the same numbers, so condition, warranty and drive wear are not factored in at all. The report states plainly that drive wear and remaining life are not verified. A grade on a refurbished machine means less than the letter suggests.
Coverage is limited too. Checks are US only for now, prices are in US dollars, and shipping and sales tax are excluded from the comparison. A Canadian listing is rejected with an error, which is what confused one commenter in the HN thread before the creator clarified that the Canadian market is next in line. Retailer coverage is limited to the biggest names, so a listing from a smaller shop will not resolve.
Reports are temporary too. A report link expires after 72 hours, and anyone holding the link can view it until then, so treat a shared report as public. The privacy policy is straightforward about data: only the image crop you select gets sent for reading, uploaded images are not saved in the app, extracted text and specs may be kept briefly for repeat reads, and activity records use coded identifiers instead of your email or IP address. There are no advertising trackers.
The business model is still unsettled. Individual checks stay free, deal alerts are free during the beta and are expected to become a paid feature later with pricing announced before the beta ends, and the creator takes optional PayPal tips to cover AI costs. There is no paid plan to sign up for yet, and email alerts are not actually sending. The service is closed source with no public repository, and the creator asked explicitly for feedback on scoring, interface and usefulness, so the grading bands will likely move as more people use it.
Source and Accuracy Notes
- Project page: https://pcdealcheck.com/ : verified via direct visit
- Sample report: https://pcdealcheck.com/sample-report : figures read from the published example
- Terms: https://app.pcdealcheck.com/terms : updated 3 November 2026
- Privacy: https://app.pcdealcheck.com/privacy : updated 3 November 2026
- Source repository: none published : the service is closed source
- HN launch thread: https://news.ycombinator.com/item?id=50366929 : 32 points, 12 comments, 4 November 2026
Related Posts
ai-setup
Prism – AI Video Workspace and API for Creators (YC X25)
Prism is a YC X25 AI video platform combining generation, editing, and an API for workflow automation. Generate assets, edit on a timeline, and integrate via.
5/28/2026
review
Mosaic – Agentic Video Editing With a Node-Based Canvas
Mosaic lets you build and run multimodal AI video editing agents on a node-based canvas. Upload raw footage, design a workflow, and let AI handle the heavy.
5/28/2026
review
Kita – VLM-Powered Credit Review Automation
Kita uses vision-language models to automate document-based credit review for lenders in emerging markets, parsing 50+ document types from PDFs to photos.
5/29/2026