The first tab I used to open every morning was Google Search Console. Four AI-focused blogs — Claude Lab, Gemini Lab, Antigravity Lab, Rork Lab — plus two reading-focused sites I keep in a separate folder. Six properties in total. Cycling through each one and staring at yesterday-vs-today click numbers had become a morning ritual that quietly ate fifteen to twenty minutes before I touched anything else.
Since the long weekend in early May 2026, I started handing parts of that round to Claude in Chrome. I am Masaki Hirokawa, a contemporary artist who has been shipping iOS wallpaper, healing, and manifestation apps as an indie developer since 2014. Running four AI blogs alongside that work meant that a careful Search Console walk in the morning was reliably stealing my best creative hours. After ten days of this new flow, the texture of my mornings has shifted more than I expected, so I wanted to record what changed.
What the morning Search Console walk was really costing
The minutes themselves were not the problem. The problem was how unstable I felt while I read those numbers. A property showing minus thirty percent in clicks day-over-day would lock my head in place. Was it a Monday dip, a Helpful Content aftershock, or just weather? To answer I would open the query-level table, then the page-level table, then the device-level breakdown, and only then would I land on "this is just a normal Monday." Around five minutes per property. Multiply that by six, and the morning was already half spent.
The quiet days were fine. But after Claude Lab's index dropped from roughly 3,500 to 95 in early May, a small piece of bracing-for-impact never quite left me when I opened the dashboard. Spending twenty of my sharpest minutes that way felt wrong. In 1997, when I was sixteen and first got online, I used to peek at how developers across borders exchanged information in pure text — there was a kind of openness in that air. I wanted my morning numbers to feel closer to that than to a quiet dread, and that is what nudged me toward automation.
What I delegated, and what I kept on my plate
The first rule I gave myself was that Claude in Chrome could summarize, but could not judge. Search Console reports — Discover, Search performance, Indexing — each carry their own quality of signal, and folding all six properties into a single human-readable note seemed both possible and useful.
I delegated three specific tasks:
- Build a single table per property comparing the last 7 days against the last 28 days for clicks, impressions, and average position.
- From "Page indexing", list only the error types whose counts increased versus yesterday.
- From "Search results", surface up to five top queries whose CTR had moved sharply against the prior week.
What I did not delegate is the evaluation step — anything resembling "this might be a Helpful Content effect" or "this is probably noise." Letting AI write those sentences just produces softly hedged speculation that I end up re-checking from scratch. Observation goes to AI; judgment stays with me. That line has held up well over the ten days.
What the ten days actually looked like in numbers
The clearest change was the wall-clock time. The old walk across six properties took fifteen to twenty minutes; with the Claude in Chrome digest opening first, my morning Search Console pass now lands around five minutes on average.
From my own usage log:
- Morning GSC review time: 18 min average → 5.4 min average
- Days I actually opened all six properties in the browser: 3 of 10
- Manual deep dives I initiated: 7 in total across ten days
- Average start time of creative work: 25 minutes earlier
I do not feel that my reading has become sloppier. If anything, the time I no longer spend reacting to small swings has freed up real time for deeper investigation when it is warranted. Mid-month I caught Rork Lab sitting at a 0.26% CTR — a number I might have taken another full session to surface in the old flow — because the digest opened with "three queries down 42% in CTR week-over-week." From signal-to-question took about a minute instead of ten.
What the automation revealed about my own blind spots
Once observation is handed off, what you used to look at — and used to skip — becomes visible.
For me, the realization was that I had been over-watching daily click counts and under-watching average position drift. Once the digest started flagging "queries that dropped from top-3 to position 4 or worse," I understood that the moment a page slides out of the top tier is heavier information than the click dip that follows. Click-through rates at position 3 versus position 4 differ enough that catching the slip early changes the cost of recovery considerably.
The second realization is the classic one: the report I want to look at is not the same as the report I need to look at. The clicks table is comforting (or fakes comfort), which is why I opened it first every morning. What I actually needed was the indexing-error delta and the Core Web Vitals notifications. Asking AI for the full sweep returns both, without my taste filter on top, and slowly straightens out the bias.
Both of my grandfathers were temple carpenters. I grew up hearing that buildings outlast generations not because someone tended the obvious spots, but because someone tended the easy-to-miss spots. The way I read Search Console seems to obey the same rule, and it is easier to follow now that an outside observer is laying every report side-by-side.
What I want to refine from here
Ten days is enough to tell that this is worth keeping. A few open questions remain, and I want to log them so I can come back in a month.
First, a single morning note instead of one long one. The current digest packs all six properties into one message, and on rushed mornings it sometimes goes unread. I want to split it into three urgency tiers and only push the urgent slice as a separate notification.
Second, diff-tracking the indexing report. Watching "Crawled — currently not indexed" move day by day will give me real ground to stand on when the next Helpful Content update lands. I am still deciding whether to dump these into SQLite myself or have Claude in Chrome keep a plain-text delta file each morning.
Third, a little further out: walking Search Console and Crashlytics in the same pass. I wrote recently about a two-week Crashlytics weekly review with Claude in Chrome, and folding both into a single five-minute morning ritual would let me read the full state of development and operations side-by-side. With one person across six sites and four apps, that compression matters.
One last thing. Even with the observation handed off, the anxiety never fully disappears, and probably never will. What changed is that my sharpest morning minutes are no longer being eaten by it. They are going back into the work itself — making things, and tending what I have already made. If you are also running multiple sites on your own, I hope something here is useful. Thank you for reading.