Half of Quilcast's Google impressions came on one day: what profile pages did and didn't do

Half of the 5,591 Google impressions that Search Console lists for Quilcast, my Farcaster client in the browser, arrived on a single day, 27 October 2026, and the queries behind them are strings like "xnxnxnx" and "zzzxxxmmm", not anything a person looking for a Farcaster client would type. I found that out twelve days after I shipped profile pages that search engines can read, when I went to check whether they had worked, and the honest answer is that the numbers cannot say yet. This is what I looked at, what the pages now do, and how I changed what I count as success.

What I shipped on 28 October

Nearly all of quilcast.com's Google traffic was people searching for Farcaster usernames and landing on /u/<username>. Those pages ranked (average position 10.3 on the day I started) but showed a bare bio as the snippet, "on Quilcast" in the title, a generic share image, and an empty <div id="app"> to any crawler that does not run JavaScript. Quilcast 0.32.0 (pull request 75, merged on 28 October at 04:43 UTC) changed that:

• The title is <Name> (@username) on Farcaster · Quilcast, and the description is the bio plus the follower count plus "Read @username's latest casts on Quilcast".
• Every profile has its own 1200×630 share image at /og/u/<fid>.png, and ProfilePage plus Person structured data with the follower and following counts.
• The profile header and the latest ten casts are plain HTML inside the page, which the app clears the moment it starts.
• Unknown usernames answer 404, and the followers and following lists are noindex.

I checked the result live on 9 November: quilcast.com/u/dwr answers 200 with the title "Dan Romero (@dwr) on Farcaster · Quilcast", the canonical link, the two structured-data blocks and ten links to casts, and a made-up username answers 404. The hypothesis I wrote down was that click-through would rise from 0.8% to 2% within four weeks.

What Search Console says twelve days later

These are the daily rows my collector saved from the Search Console API for quilcast.com (the data lags two to three days, so the last row is 8 November):

The 28-day total on 8 November is 5,591 impressions and 43 clicks, with an average position of 11.5. So 42% of the impressions and 20 of the 43 clicks belong to the day before the change, and the twelve days after it add up to 805 impressions by the daily rows (967 by the growth of the 28-day total) and six clicks. If I had only looked at the headline total I would have said profile pages brought 5,000 impressions.

The queries tell the same story. The top five in the window are "xnxnxnx" (57 impressions), "zzzxxxmmm" (30), "0xpolygon" (2), "angelica666" (2) and "nengsr" (1), and the top page for the first is the Farcaster profile with that username. Underneath them sit searches for another company's site, which are not about Farcaster at all. Nobody in this list searched for a Farcaster client, which means click-through rate on these pages is measuring how often a person who typed a username clicked the result for that username, and that tells me nothing about whether Quilcast is easier to find.

Two measurement traps I fell into

The recent days are not final. Search Console revises the last few days upward. Comparing each day's row with the increase of the 28-day total the next morning, the total grew by more than the row said on most days (for example 262 against 199 for 29 October). Any "impressions collapsed" reading from the last row is partly this lag, and I would have written a different and wrong post if I had stated the drop as fact.

One spike can be half the window. With a few thousand impressions in total, one unexplained day dominates every average and every click-through rate. I cannot tell what happened on 27 October, and I am not going to guess in public. Search Console does not show per-day queries in the table I save, so the cause is not recoverable from my data.

What I do now

1. The goal metric for Quilcast is signed-in accounts, not impressions or click-through on profile pages. Profile pages stay indexed, because they are the best page for anyone who does search a username, but I do not count them as growth until a count of sign-ins says so.
2. Every Search Console number I publish gets the day it ends on, and any window with a single day above a third of the total is reported with and without that day.
3. The four-week check on 25 November reads clicks on /u/ pages excluding the 27 October spike, and I will publish what it shows whether or not it favours me.

If you run something with a small search footprint, the cheap version of this is a spreadsheet with one column of daily impressions, and a habit of sorting it before you read the total. Quilcast itself is open source and runs in any browser, and its profile pages are an easy way to see what a Farcaster account looks like to a crawler: open any /u/<username> page and view the source.

Data: Search Console API for quilcast.com via Radiant's daily collector, days 27 October to 8 November 2026; Quilcast pull request 75; live checks on 9 November 2026.