Teardown No. 107 · Programmatic SEO

105 Clicks vs 68: How Rob Boirun found one fuel calculator outearned the 53 city pages he generated with Claude Code

Rob Boirun builds and runs bestrvrentalcompanies.com, an RV rental review and comparison site, with Claude Code: 361 of the repository's 387 commits carry a Co-Authored-By Claude line. Between June 24 and July 21, 2026, one fuel cost calculator earned 105 of the site's 390 Google clicks. In the same 28 days, the 53 city pages he had generated with Claude Code earned 68 clicks between them, and 30 of the 53 earned nothing. His method is to sort every URL by page type, compare click-through rates at similar positions, and only then decide what to generate next.

Rob BoirunVerified Nov 12, 2026Markdown copy

Try this step

Sort every URL into a page type. Export the last 28 days of Search Console Performance data and total clicks, impressions and zero-click URLs for each type.

The problem

The site is built in Astro and hosted on Vercel. The fuel cost estimator shipped on November 14, 2025, ten days after the first commit, alongside an RV size recommender quiz, a mileage allowance calculator and a packing checklist. Up to the end of the export window, six commits touched the calculator file and 81 touched the city pages. The city pages came from a template: the generation doc “Loads Alexander AR as template” and runs a “Smart find-and-replace for city-specific data.” That left residue. A commit on April 19, 2026 cleaned “Arkansas template residue” off nine city pages, and on July 23 the Apache Junction, Arizona page still carried Arkansas data. In the clean-up that followed, he and Claude deduplicated template sections, stripped generic list sections from 24 city pages and purged “fabricated experience” from all of them. One commit message reads “agents hit session cap mid-run.”

The blueprint (as he describes it)

  1. Sort every URL into a page type. Export the last 28 days of Search Console Performance data and total clicks, impressions and zero-click URLs for each type.
  2. Compare click-through at similar positions. The calculator averaged position 7.11 and converted at 1.24%. Four review URLs near it, at positions 7.56 to 8.32, had 25,501 impressions and 74 clicks, a CTR of 0.29%. Build more of the type that converts, not the type with the most impressions.
  3. Read your best page’s queries through the page filter. The site-wide export’s 1,000 query rows account for 49 of the 390 clicks, and at least 91% of the calculator’s impressions came from searches that are not in the export.
  4. No queries, no tool. Find a handful of distinct queries asking for a tool’s answer before building it. The visible query rows include 35 distinct ways of asking about RV fuel cost, and he found no query for a packing checklist or a mileage allowance.
  5. Lint generated location pages before they ship. Check each one for the source city’s name, state, ZIP and airport code, and ship only the cities where you hold real local data.
  6. Put next week into the page type that earns, then repeat the table in 28 days.

Stated results

  • Fuel cost calculator: 105 clicks, 26.9% of 390, from 8,498 impressions at a 1.24% CTR.
  • City pages: 68 clicks from 26,178 impressions at a 0.26% CTR. Of the 53 city pages live in the window, nine had no impressions and 22 had impressions but no clicks, about 1.3 clicks per city page in 28 days.
  • Across the site, 269 URLs had impressions and 194 of them, 72%, got zero clicks. Half of all clicks went to six URLs. The glossary had 78 URLs and 4 clicks.
  • The three other tools that shipped the same day as the calculator had no rows at all in the export.

Caveats

  • It is one 28-day window from one export. He has no earlier or later export for the site, so there is no trend and no before-and-after for the July clean-up.
  • His own arithmetic does not quite close: nine pages with no impressions plus 22 with impressions and no clicks is 31, while the page says 30 of 53 earned nothing.
  • It shows clicks, not money. There is no revenue or affiliate data, and he makes no claim that the calculator earned more than any other page.
  • Average position is an average across many queries, so the matched-position comparison is “a fair check, not a controlled test.” The page types are his own classification of URL paths, and page-level rows sum to 391 clicks against 390 in the chart.
  • The calculator is simple: its MPG ranges and $3.50 default gas price are typed into the page by hand, and its subtitle still says “real MPG data from 50+ rentals,” a figure the site lowered to 25+ elsewhere on July 23, 2026.
  • He notes that a city page cloned from Arkansas, with an Arkansas ZIP on an Arizona page, sits close to Google’s descriptions of scaled content abuse and doorway pages, “whatever my intent was.”
  • Handles: name only. An X account named Rob Boirun (@Robboirun) exists with no posts, bio or link, so it is not verified. The PopNet Media brand account is not tagged.

FAQ

Did the generated city pages work for Rob Boirun?

Not as well as one tool. In 28 days the 53 city pages earned 68 clicks between them and 30 of them earned nothing, while a single fuel cost calculator earned 105.

Did the calculator win only because it ranked higher?

Partly, he says, but position does not explain most of it. At a similar average position the calculator converted at 1.24% and four nearby review URLs at 0.29%.

Should every site build calculators?

He says no. Three other tools shipped the same day and had no rows in the export, so his rule is to find queries asking for a tool’s answer before building it.

How were the city pages made?

From a template: a generation doc loads one Arkansas city page and runs a find-and-replace for city-specific data. Later commits cleaned leftover Arkansas data off several pages.

Sources

Want agents like these working your distribution? Get It Built. Get new teardowns weekly: subscribe (double opt-in).

Get one teardown a week.

Same format, a new operator every issue. Double opt-in.