A friend asked me what a new site looks like in Google Search Console before it has any authority. Here is the raw shape of it, because most write-ups show the success curve and skip the flat part.
The numbers, 90 days in
- 33 indexable URLs, each with a real HTML shell and 600 to 1800 words of prose
- 312 impressions
- 1 click
- Average position 12.3
- 44 distinct queries with at least one impression
- External links, per the Links report: 0
That last line explains the other five. Nothing was pointing at the site, so nothing was pushing any URL upward, and 312 impressions is roughly what you get when Google is passively testing a site it has not decided about.
The part I did not expect: ten queries in the top ten, zero clicks
Being on page one usually means clicks. Not here. The queries looked like this:
- "a food handlers prescription medicine must be stored servsafe"
- "the first step in cooling a large roast"
Those are verbatim question stems. Somebody was stuck on one specific practice question and typed it into Google. A page ranking 3rd for a 40-searches-a-month question gets, realistically, about one visit a month. Eight of the ten top-ten queries were that exact shape.
So the site had two problems wearing one coat:
- No authority. The broad terms, the ones with thousands of searches a month, were not at position 40. They were absent from the first hundred. There is no partial credit for relevance when the domain has no link equity.
- Relevance only where demand is tiny. Question-level long tails convert well and are nearly free to win, but each one is worth about one visitor a month.
What actually moved
Nobody fixes (1) in a week. It is links, and links arrive when other people decide there is something worth citing. That is slow and mostly out of your hands.
(2) is fixable, and cheaply. If your winning queries are question text, then the question text should be in the HTML. Ours was not: the questions lived in a JavaScript file and were rendered into the DOM on load. Google does execute JavaScript, but it does it later, with a smaller budget, and for a domain with no authority the first HTML-only pass is often the only pass that matters.
The fix was not a rewrite to server-side rendering. It was narrower:
- take the questions that already exist in a data file
- render a subset of them into collapsed details/summary blocks
- put them on pages that are already indexed and already match the topic
Fifty questions across six existing URLs, about ninety minutes of work. The interactive quiz stayed exactly as it was.
What I would tell someone else at the 90-day mark
- With zero backlinks, track the number of queries that have any impressions. It is the first number that moves, and it tracks coverage rather than luck.
- Sort your queries by "could I answer this verbatim on a page I already have". Those are the only queries you can win without authority, and they are also the highest-intent ones.
- Do not add pages to chase them. Adding URLs mostly grows the denominator of things Google has not crawled yet. Put the content in a page that ranks.
- Check whether your best content survives being read as plain HTML. Run curl against your own page and grep for the text. If it is not in the response, no amount of on-page optimisation will help.
The two pages this cycle was about are the practice test at https://foodsafeprep.com/servsafe-practice-test and the answer breakdown at https://foodsafeprep.com/servsafe-practice-test-answers. Both are static HTML now, and neither needed a framework change to get there.
Top comments (2)
asrocky010, this is a brilliantly pragmatic breakdown of the "flat part" of the SEO curve.
your
curltest is the ultimate truth-teller. too many devs assume googlebot's js rendering budget is infinite, but for a zero-authority domain, that initial html-only pass is often the only pass that matters.the
<details>/<summary>trick is genius. it satisfies google's need for verbatim text in the raw html without bloating the initial payload or forcing a heavy ssr migration. it’s the perfect middle ground.since you injected this content into already-indexed urls rather than creating new ones, i’m curious about the gsc latency: did you see the impression spike for those specific long-tail queries within the standard 48-hour recrawl window, or did it take closer to the weekly refresh cycle?
also, did you have to manually submit those specific urls via the inspection tool to force the recrawl, or did you just let googlebot find the updated html organically?
would love to hear how the delta looked in the console after the fix! 📈
1 click from 312 impressions is brutal, thanks for posting the raw numbers. The verbatim question stems ranking top ten with zero clicks is the part that got me, each of those is basically one person's homework. Curious if you'll chase the first links manually, communities or cold emails to people you cite, or just keep publishing and wait?