Last spring I built 28 small tool sites. Generators, converters, pickers, reference pages. Each one answers a question people actually type into Google, each one is a static Next.js build, and together they are my own SEO laboratory: every experiment I would not risk on a client runs there first.
In early July, Google found them. The graphs did exactly what you dream about: one site went from nothing to 1,500 impressions a day, another to 2,500, a third to 3,300. For about a week I thought the lab had graduated early.
The week Google changed its mind
Then, mid July, almost all of it disappeared. Not a slow slide down the rankings but a hard cut: a site peaking at 1,540 impressions on July 14 did 32 the next day and zero the day after. Sitemaps with hundreds of URLs kept three or four pages in the index. The rest was simply gone.
My first instinct was to look for a technical cause. There was none. Every page returned 200, nothing carried a noindex, robots.txt was open, sitemaps were reachable. Around the same time the domains got hit by a spam link blast from throwaway SEO domains, which looked suspicious, except that the one site that kept its rankings got the exact same blast. Correlation, not cause.
What actually happened, as far as the evidence goes: a site-level trust verdict. Twenty-eight fresh domains, zero earned links, launched around the same time, all built from templates. Google indexes something like that, tests it against real users for a week, and pulls it back out. The index is an invitation, not a residence permit.
The lesson that hurt more
The de-indexing was not even the worst discovery. That came from reading the click data of the good week. One query did 8,430 impressions at position 9 and produced two clicks. A position 9 result normally gets a few percent of clicks, not 0.02 percent. The difference is Google answering the question directly at the top of the page.
For coin flips, unit conversions, sunrise times and number lookups, page one is increasingly an answer, not a list of links. You can win the ranking and still get nothing, because the searcher never needs to leave. Impressions are vanity. Clicks are the only proof that a page earns its place.
What I am doing about it
Not building more pages. That was my honest first reflex, and it is exactly wrong: more templated volume on unproven domains makes the trust problem worse. The plan runs the other way. Fewer, deeper pages. Real links earned instead of generated. Tools that do something an instant answer cannot, so a click actually pays off for the person clicking.
And patience, which is the least satisfying strategy ever written down. These domains are months old. Trust verdicts get revisited as domains age and accumulate signals. The lab is doing what a lab is for: failing in instructive ways before any client budget touches the same mistake.
The numbers on my homepage come straight from Search Console, refreshed automatically, the bad weeks included. That is deliberate. If I only showed the graphs that go up and to the right, this whole site would be a sales pitch, and I promised you it would not be.