Summary
Organic search at a marketing-talent marketplace produced 1,031 leads and $366,294 in signed revenue a year, and everyone treated it as a channel that worked. A first-party audit found 2,650 pages telling Google a different page was the real one, 87% of clicks from people typing the brand name, and a conversion count that was 98% pageviews. This is what a healthy-looking channel was hiding.
When the numbers are good enough not to look
A channel that produces $366,294 a year doesn’t get audited. It gets a line in the board deck and a pat on the head. That was the condition I found it in.
Nobody had checked whether that $366,294 was earned by search, or merely delivered by it. Those are different things, and the difference is where the growth lives.
Measured, not estimated
I had first-party access, which means nothing in this audit is a third-party guess. I ran my Audit Engine on the full data set: 22,422 queries from Search Console and 1,012 landing pages from GA4 for the 28 days to August 8, 2026, a live crawl of 2,651 of the site’s 2,681 pages, and 60 verbatim answers from AI assistants and Google AI Mode. Then I joined all of it to signed revenue in the data warehouse, so every finding could be priced.
Then I hand-verified the findings the crawler flagged, because a crawler’s flag is a hypothesis, not a fact.
Six findings, each priced
2,650 pages were competing with themselves. Every web page can carry a small line of code that tells Google “the real version of this page is over there.” It’s meant to stop duplicates from splitting credit. On this site, 2,650 pages pointed at a different page than the one they were.
Google was being told, at scale, to ignore most of the site. I checked this by hand; it wasn’t a crawler artifact.
Google showed the site to strangers 881,380 times, and 326 of them clicked. Sitewide, 889,349 impressions produced 2,427 clicks, a 0.27% rate. Split brand from non-brand and the picture sharpens: 87% of all clicks were the brand name or a misspelling of it. Non-brand search, the part that finds new customers, earned 326 clicks.
The $366,294 was real. It was mostly delivered to a site that people already knew the name of.
1,281 queries sat between positions 4 and 20 and earned nothing. That’s the shelf: terms Google already considers relevant, close enough to page one to matter, converting to zero clicks. This is where organic growth comes from, and it was sitting untouched.
The conversion data was measuring nothing. GA4 reported 21,636 conversions. 21,133 of them were a pageview. Decisions about what content “converted” had been made on a series that counted page loads.
Awareness was not the problem. All 48 assistant answers and all 12 Google AI Mode answers named the company. Only 18 of the 48 cited one of its pages. The assistants knew the brand and built their answers from other people’s content.
It’s a pattern, not a one-off. Weeks later I ran the same brand-versus-non-brand split on a direct-to-consumer apparel brand with 17,077 clicks in 50 days. 84% were branded. Its head term, “dress shirts,” ranked at position 5.6 with 30,927 impressions and earned 114 clicks, a 0.37% rate. Ranking on page one and earning nothing from it is far more common than anyone admits.
Turning findings into a backlog
An audit that ends in a PDF is a bill. This one became the work.
The canonical fix went to engineering as a specified ticket with the evidence attached. The conversion event was rebuilt around actions that mean money. The 1,281 shelf queries became the content plan, prioritized by impressions.
The 18-of-48 citation gap became the answer-shaped rewrite program. And the weekly audit in my Optimization Loop now reports brand and non-brand separately, pulled at 25,000 rows, because the default 5,000-row export truncates exactly the non-brand tail that changes the conclusion.
At a glance
- The niche: Marketing-talent marketplace; its own 2,681-page site.
- The problem: Technical (2,650 conflicting canonical tags), on-page (1,281 queries ranking 4 to 20 with zero clicks), measurement (98% of GA4 conversions were pageviews), visibility (87% of clicks branded; cited in 18 of 48 AI answers).
- What was built: My Audit Engine run first-party, joining Search Console, GA4, a live crawl of 2,651 pages and 60 verbatim AI answers to signed revenue.
- Implementation: 22,422 queries and 1,012 landing pages over 28 days; every crawler finding hand-verified; findings turned into engineering tickets, a rebuilt conversion event, a content plan from the shelf queries, and a brand/non-brand split in the Optimization Loop’s weekly audit.
- Results: Six verified findings on a $366,294 channel, each with the pages affected and the fix, forming the backlog for the next two quarters.
- Built with: my Audit Engine and Optimization Loop.
