Blog
By Haobo Zhang AI SEO Jul 3, 2026 11 min read

A brand-new domain's first two weeks in Google Search Console: every impression, published raw

A domain born around June 20, 2026: 160 impressions, 4 clicks, 27 visible queries, average position 59.1 in two weeks. Full daily GSC data, published raw.

A domain born around June 20, 2026 earned exactly this in its first two weeks of Google Search (window June 19–29): 160 impressions, 4 clicks, 27 visible queries, average position 59.1. Every click was a brand search landing on the homepage. The full daily curve and query breakdown are below — raw, unrounded, n=1.

This is one site, in one niche, measured once. It is not a benchmark — it’s a datapoint where we couldn’t find any.

How long until a new website gets impressions?

Ours got its first impressions on day one of publishing content — 12 of them, the same day its 16 launch pages went live — ramped to a peak of 41 per day on June 25, five days after the domain went live, then collapsed to 4–7 per day at visibly worse positions.

Date (2026)ClicksImpressionsAvg position
06-1900—
06-2000—
06-2131236
06-2212069.6
06-2302468.9
06-2403665.8
06-2504145.8
06-2601049.1
06-270461.2
06-280670.2
06-290783

↓ Download this dataset (CSV) — raw daily GSC figures, CC BY 4.0.

The retreat wasn’t just volume: average position drifted from 45.8 on the peak day to 83 by June 29. Our read is that Google tested the new domain in live results for about five days, watched what happened, and recalibrated downward Medium — an interpretation, but one that tracks closely with how Google’s own John Mueller describes new-site handling, quoted below.

Which queries surfaced first on a brand-new site?

One topic owned the fortnight: of the 138 impressions GSC ties to visible queries, 126 — 91% — came from a single cluster about Google’s AI Mode and AI Overviews. The visible queries also aren’t the whole story — some impressions have no query rows at all: Google omits “anonymized” queries (issued by no more than a few dozen users over two to three months) from the query table while still counting them in chart totals, unless you filter by query (Search Central).

Query clusterImpressionsSample queries
Google AI Mode / AI Overviews126 of 138 query-attributed”google ai overviews” (32) · “ai mode” (25) · “google ai mode” (22) · “ai overviews” (17) · long tail
AEO definitional10”aeo acronym” · “what is aeo” · “aeo definition”
Other2”ai share of voice” · “hz ai”
Hidden (“anonymized”)22 of the 160 totalno rows shown — queries too rare to display

The long tail is the texture: misspellings (“ai mod3”, “a1 mode”) and non-English queries (Danish “ai mode synlighed”, Dutch “wat is google ai mode”) — we unpack the cluster in a separate post. And 27 queries is a floor, not a census: Ahrefs measured an average 46.77% of GSC click traffic hidden as anonymized queries across 887,534 properties (Ahrefs). Whether any of this surfaced inside AI Overviews or AI Mode we can’t see: Google’s Search Generative AI performance reports (June 3, 2026) are rolling out to a subset of sites first, and our property’s searchAppearance API query still returns zero rows.

What does an impression at position 76 actually mean?

It means Google put our link somewhere on a results page a user reached — not that anyone saw it. GSC counts an impression “whether or not the item is scrolled into view” (Google’s definitions), so a link served at position 76 was almost certainly never seen by a human.

Position is the topmost spot any link to the site occupied, averaged across the queries where the site appeared — one value per results-page serving — and Google itself warns the same number “can mean different things in different situations”. On tiny serving counts that average swings violently, which is how our daily position bounced from 36 to 45.8 to 83 inside nine days.

A click, by contrast, only counts when the user lands on a page outside Google. Read our fortnight in those units: Google served us 160 times, deep in the results; humans clicked 4 times.

Why does a new site get impressions but no clicks?

All 4 of our clicks were brand or navigational searches landing on the homepage, on days two and three; every content page earned its impressions at positions 30–90 and got zero clicks.

Two week-old glossary stubs pulled 61 impressions each — /glossary/ai-mode/ at position ~52 and /glossary/google-ai-overviews/ at ~76 — Google sampling a brand-new site into hot-topic results at page 5–8.

Established sites are currently watching impressions rise while clicks fall — Ahrefs named it the Great Decoupling, and the underlying study found AI Overviews correlate with a 34.5% lower CTR for the top-ranking page across 300,000 keywords (Ahrefs). A 2026 domain doesn’t even get to decouple: it’s born decoupled. That’s context, not causation — at position 59, nobody clicks you, AI Overviews or not.

How fast did Google actually index the new domain?

Fast, then not at all: 10 of the 16 launch pages were indexed at a median lag of 3 days — and the other 6 were still out at day 11, one stuck at “Crawled - currently not indexed.” Google defines that status neutrally: “The page was crawled by Google but not indexed. It may or may not be indexed in the future; no need to resubmit this URL for crawling.” (Page indexing report). We read it as a new-site trust ceiling — that part is interpretation, clearly labeled. Both halves matter: the 3-day median beat Google’s stated expectation that “It can take a few weeks for Google to notice a new site, or any changes in your existing site” (Google Search Central), and the stall shows Google did not take the whole sitemap. The daily indexing data lives at our index-lag experiment; the same site’s zero Bing crawls over the same window are their own investigation.

Is the Google sandbox real?

Google’s answer is a flat no — Mueller in August 2019: “There is no sandbox.” (Search Engine Roundtable) — but Google does say new sites run on estimates that can overshoot. In a 2021 Search Central video Mueller rejected both the sandbox and a deliberate honeymoon: “it’s again not the case that we’re explicitly trying to promote new content or demote new content… It’s just, we don’t know and we have to make assumptions” (Search Engine Journal).

In January 2020 he was more mechanical: for brand-new sites “we might not have a lot of signals for those websites so we have to make estimates” — and Google may initially show a site more visibly than its signals later justify, then recalibrate (Search Engine Journal). He has also said new-site rankings can fluctuate for up to about a year before settling (Search Engine Journal, June 2020).

Our ramp-to-41-then-retreat curve looks like a live specimen of that estimate-overshoot-recalibrate loop — one specimen, from one site.

Where does the “3–6 months” advice come from?

From practitioners, not from Google. The cleanest attribution is an Ahrefs poll of 3,680 respondents on LinkedIn and X: “It typically takes three to six months for SEO to show results” (Ahrefs). The range then echoes across the industry, usually with no new-domain data behind it (seo.ai is a representative example).

The data Ahrefs does have is sobering: only 1.74% of newly published pages reach Google’s top 10 within a year — down from 5.7% in 2017 — and 72.9% of top-10 pages are 3+ years old, though of the pages that do make it, 40.82% get there within a month (Ahrefs, May 2025).

Google’s own closest figure is from Maile Ohye’s 2017 video: “in most cases, SEOs need four months to a year to help your business first implement improvements and then see potential benefit” (via Search Engine Roundtable) — a sentence Google’s current “Do you need an SEO?” page, updated June 2026, no longer contains.

Why is there no benchmark for a new domain’s first weeks?

Because the public numbers come from operating companies — and nearly a year of them is silently inflated. Databox’s “All Companies” cohort calls around 122k organic impressions a month a good number (live page, checked July 2, 2026 — Databox); it offers no segmentation by domain age and no new-domain cohort.

Worse, a Google logging error inflated impressions — clicks unaffected — for virtually all sites from May 13, 2025 until the fix took effect around April 27, 2026, and the historical data will not be corrected (Search Engine Land; fix follow-up). Any impression benchmark touching that window is on the wrong scale; our window is entirely post-fix.

We searched for another published raw first-weeks GSC dataset from a new domain and couldn’t find one — the nearest neighbor, Rafael Cavalcanti da Silva’s 12-week case study (metrics in its Part 2, April 2026; also 16 pages, also Astro), shares topline weekly numbers like 0 → 847 weekly impressions, not a day-by-day query-level export. If you know of one, send it — we’ll link it here.

Frequently asked questions

Why does my new website show no impressions in Search Console?
Google says it can take up to a week before a newly added property generates any Search Console data, performance data then runs 2-3 days behind, and noticing a new site can take a few weeks. Our domain's first impressions appeared the same day its first 16 pages were published.
How long until a new website gets impressions in Google?
There's no official number — Google only says it can take 'a few weeks' to notice a new site. Ours logged impressions the day content went live and peaked at 41 impressions/day five days after launch; the only clicks in two weeks were 4 brand searches for the homepage.
Is the Google sandbox real?
John Mueller has said directly 'there is no sandbox' — but also that Google must estimate where a new site belongs, that early visibility can overshoot and then settle, and that new-site rankings can fluctuate for up to a year. Our ramp-then-retreat curve matches that description.
Why do I get impressions but no clicks on a new site?
A Search Console impression only means Google served your link on a results page the user reached — even at position 76, unseen. Our content pages drew all their impressions at positions 30-90 and zero clicks; the 4 clicks we did get were brand searches landing on the homepage.
How accurate is Search Console data for a new site?
Treat it as delayed and partial: performance data lags 2-3 days, and Google hides rare 'anonymized' queries from the query table — an average of 46.77% of click traffic across sites, per Ahrefs. On a new site nearly every query is rare, so the visible query list is a floor.

Sources