Engineering · 2026.10.10 · 6 min
Content SEO for a New Domain: Check Search Volume First, Then Write Guides Around the Data
A new domain whose brand name is taken can only win long-tail traffic with guides. I ranked topics by OpenSEO search volume, checked each claim against primary sources, and loaded GA4 only on public pages.
Reading
I picked the first batch of topics for EZLO's guides section on gut feeling. Only after checking search volume with a self-hosted OpenSEO did I learn that "Bitfinex lending bot", the phrase I'd been targeting, had no volume. The volume was in "bitfinex lending" and a cluster of tax-related terms. (Unless noted otherwise, the keywords in this post are Traditional Chinese queries, translated here.) After that, topics were ordered by search volume, every article was checked against primary sources first, and anything I couldn't verify didn't get written. The whole round of keyword research cost about US$3.5.
The background: the brand name "EZLO" is completely taken by a smart-home brand of the same name, so the only way for a new domain to get traffic is category terms, and the top results for those searches are mostly affiliate posts pushing one competitor or another. I opened a guides section at ezlo.me/learn/, prerendered with @next/mdx and carrying Article and breadcrumb structured data, and wrote 7 articles for the first batch.
Checking search volume with self-hosted OpenSEO
Two weeks after the guides section launched, all I had was Search Console. I couldn't see search volume, competitors' ranking keywords or backlinks. A Semrush or Ahrefs subscription doesn't make sense at this scale.
I found the open-source OpenSEO, which pulls its data from DataForSEO's pay-as-you-go API. The minimum top-up is US$50, and the balance doesn't expire. I first deployed it on the Cloudflare Workers free plan, where parallel queries, and even a single rank-tracking run, hit Error 1102: Worker exceeded resource limits. The same day I moved it to my own server under Docker. In Docker mode, the cron that the Worker used to handle doesn't fire on its own, so I added a systemd timer that hits the scheduling endpoint.
OpenSEO has an MCP server. Once it's connected, Claude Code can query search volume and rankings directly.

The data overturned my original topics
"Bitfinex lending bot" had no volume. What did have volume was "bitfinex lending" (480 searches a month), the brand names of a few competitors (each from a little over two hundred to nearly four hundred), and the tax cluster (about 570 combined).

I rewrote the homepage title to match: "Bitfinex Lending Bot | Flat Annual Fee, No Commission | EZLO". It keeps "lending bot" but makes "Bitfinex lending" appear as one contiguous phrase. Later topics were ordered by search volume, and I added nine more over time, on things like lending tips, whether Bitfinex is safe, lending taxes in Taiwan and withdrawing funds back to Taiwan. The guides section on PROD now has 16 articles.
Rank tracking runs at the end of every month for 29 Traditional Chinese terms and 10 English terms, about US$2.8 a year. I got that estimate wrong at first. Weekly tracking on two device types would have cost around US$18, and switching to monthly brought it back down. The October 1 baseline: only 4 Traditional Chinese terms in the top 100, the best being bitfinex frr at #4. New domains usually take 3 to 6 months to show ranking movement, so from here on it's up to the monthly automatic tracking.
Every claim checked against a primary source
Before writing each article, I sent a Claude Code subagent to check primary sources: official terms, documentation, regulations, press releases. Claims without a primary source didn't get written. This step caught a lot of claims that are common online, some of which I'd gotten wrong myself:
- FRR is the amount-weighted average of active fixed-rate funding, not "a smoothing of recent trades" as many blogs say.
- The annual cumulative foreign exchange settlement limit for individuals in Taiwan is no longer US$5 million. Taiwan's central bank raised it to US$10 million on 2024-10-31, though the text of the regulation still literally says five million.
- A common claim online is that offers placed at FRR are charged an 18% fee. I couldn't find an official source for it. The official docs only say hidden offers are charged 18%.
- People often say a US court has already ordered 94,000 bitcoin returned to Bitfinex. In fact, the restitution ordered in April 2025 was 0, and ownership is being handled in a separate proceeding.
Most of the "lending bot comparison" results in search are written by the competitors themselves, contain errors, and can't be cited.
Two sensitive kinds of articles got their own rules. Tax articles don't draw conclusions like "lending interest counts as overseas income". The authorities have only stated a position on buying and selling virtual assets, so the articles only describe rules that can actually be looked up. Competitors can be named in the guides section but not on the homepage or in meta tags. Comparison articles only list facts verifiable on each company's official site and terms, note the date they were checked, and get rechecked every quarter.
GA4 only on public pages
Traffic analytics uses GA4, on the condition that app page URLs never get sent. Password reset links carry a token, and admin user-search URLs carry an email address. None of that belongs in GA.
So GA4 only loads on PROD public pages, and once a document has visited an app page, measurement stops until the next full page load. The reason is that gtag sends the previous URL as page_referrer, and gtag('set') can't override it. If measurement resumed when the user returned to a public page, the app page URL would go out as the referrer.
Visitors from the European Economic Area, the UK and Switzerland don't load GA, with the country read from Cloudflare's /cdn-cgi/trace. My users are in Taiwan and traffic from those regions is tiny, so I simply don't measure it, and the site doesn't need a cookie consent banner.

Wrap-up
- For a new domain whose brand name is taken, check search volume for category terms first. Don't pick topics on gut feeling.
- Self-hosted OpenSEO plus pay-as-you-go DataForSEO costs a few dollars per round of research, which is plenty for a small project. The Workers free plan can't handle rank tracking, so run it on your own machine.
- Check every claim in a guide against a primary source. If you can't find one, don't write it.
- Before deciding which pages get analytics, think through what their URLs contain.

Investing
From 1,000 to 10,000 USDT in Paper Trading: A Trend Strategy from Backtest to Shutdown
A trend strategy that backtested at 240x lost 10.7% in 2.5 days live; a hindsight-picked coin list had inflated it, and with point-in-time rankings, 10x in a year only happened from a bull-market start.
Letters
Loading comments…