New Domain to Page 1 in 21 Days: A Public AEO Experiment (Live GSC Data)

TL;DR: A new domain isn't supposed to rank for months — Ahrefs' research found only ~5.7% of new pages reach Google's top 10 within a year. We launched altyzo.com on September 5, 2026 with zero backlinks, zero audience, zero ad spend, and zero socials — and ran the exact method we sell on it. By day 19, Google was test-serving us at average position ~10 across the board; by day 22, we held page-1 positions on real queries — a comparison page at position 2, our llms.txt guide at ~5, core service pages at ~4. This post is the full receipt: the day-by-day Search Console data, the method step by step, what we predicted vs. what happened, and the honest ceiling. We update it every week, in public, because nobody else in this category will.

Why we did this on purpose

Every vendor in SEO, AEO, and GEO sells a method. Almost none show you the method working on something they can't afford to lose.

We deliberately ran the experiment on altyzo.com — the domain we sell from. Not a sandbox site, not a client's test budget, not a side project where failure costs nothing. The storefront itself. If our own playbook didn't work, the site you're reading this on would be invisible — and we'd have no business selling it to you.

That's the point of the test. "We applied our method to ourselves" is a nice sentence; "we staked the company website on it" is a claim with teeth.

The floor conditions

To make the result mean something, we started from the worst position a domain can have — and we mean zero, not "small":

The only thing we brought was the method itself — and one correction worth making honestly: entity work (schema, claimed profiles) is part of the method, so this was never "no marketing." It was zero distribution advantage — no channel other than search and AI answers could move these numbers.

The method, in the order we ran it

Everything below is documented and replicable — the same sequence we run for clients:

1. Baseline audit — we scored the site against our own 40-point AEO checklist: 34/40. The six misses: no entity corroboration, a company byline instead of a person, a privacy page that was inaccurate (it claimed no tracking cookies while GA4 was running), FAQ markup lagging content, no original data, and no internal-link architecture.
2. The corpus — 24 posts targeting real buyer queries across AEO, GEO, per-engine citation guides, and agency-vs-platform decisions. Answer-first structure, question-based headings, FAQ sections, full Article/FAQPage schema.
3. The comparison cluster — 9 honest compare pages (Ahrefs, Semrush, Surfer, Jasper, Profound, AthenaHQ, Peec, Otterly, Scrunch) — real prices, real weaknesses, including where the competitor wins.
4. Routing pages — /geo-agency, /pricing, /process, /managed — because buyer-intent queries were landing on the blog index with no real answer.
5. The entity layer — Organization + Person schema (Mariya B, founder), sameAs wiring to GitHub, claimed Tracxn and Crunchbase profiles — one canonical description, identical everywhere.
6. Technical baseline — every page pre-rendered, sitemap + RSS + llms.txt, indexing requested on publish day (we learned that one the hard way — see "what we'd do differently").

Nothing above is proprietary magic. It's the published checklist, executed end to end. The question was never whether the checklist is good — it was whether a domain with nothing could be moved by it.

The receipts

Here's the data nobody publishes. Real Search Console numbers, every day since launch — warts included:

The positions that survived the settle

The spike is not the finding — the spike is Google testing. What matters is what stayed standing after:

Read the pattern, not the pixels: every page that ranked early is a page built on the doctrine — honest comparisons, answer-first structure, named author, first-party data. The queries where we're invisible are the ones where the top 10 is entrenched authority — where a 3-week-old domain can't win yet no matter how good the page is. Format wins thin SERPs; authority is still the gate on thick ones.

Predicted vs. observed

Because this was an experiment, we wrote down expectations first. Here's the scorecard:

The honest headline: the experiment's finding is positions, not traffic — and that's the right order. Google has to trust you before it can send you volume.

How to read this without fooling yourself

Three discipline rules we applied to our own data, so you don't have to take our enthusiasm on faith:

A spike is a probe, not a trend. Day 19's 80 impressions looked like a breakthrough. It was Google test-serving a coherent new domain against real SERPs — then recalibrating. We treat the settled positions as the signal, not the peak.

Anonymized queries distort the picture. Some of our impressions (the llms.txt guide's 51, notably) come from long-tail queries GSC won't show us — privacy-filtered or too rare to report. We report what's visible and flag what isn't.

3 weeks proves speed, not durability. The honeymoon might still be flattering us. Whether positions hold through months 2–3 — as earned corroboration compounds — is the real test. That's why the update log below exists.

What we'd do differently

• Run the entity work in parallel, not after. Profile corroboration takes days to propagate into entity resolution — we sequenced it behind the content and paid for it in the quiet weeks.
• Request indexing on publish day, every time. Two posts sat "unknown to Google" for days until manually requested. Thirty seconds, free, unblocks the evaluation clock.
• Build routing pages when the first query appears. "Do i need a geo agency" was hitting our blog index at position 75 for two weeks before /geo-agency existed. The query is Google telling you what page it's looking for — listen earlier.

Frequently asked questions

Isn't this just ranking for low-competition queries?

Yes — and that answer deserves more than a shrug. The claim isn't "we beat incumbents"; it's "the method gets a zero-authority domain evaluated and positioned in 21 days instead of months." Ahrefs' data says ~5.7% of new pages crack the top 10 in their first year. Clearing that in three weeks on the right queries is the finding. The commercial queries are the next phase — and the update log will show whether we get there.

Won't the honeymoon fade?

Possibly — and if positions slide, the update log will show that too. Publishing the receipts only works if you publish all of them. The durability question is answered with data, not with confidence.

Two clicks in three weeks — isn't that a failure?

For traffic goals, yes — we're not pretending otherwise. But traffic was never the hypothesis; positions were. You can't get clicks from queries you don't rank on, and you can't rank on queries you haven't been evaluated for. Positions first, then volume — that's the order operations run in.

How is this different from what you'd do for a client?

Only the starting line. A client domain comes with existing authority, existing rankings, existing brand demand — the playbook runs on top of that foundation instead of proving itself from zero. If the floor produces page-1 positions in three weeks, the interesting question is the ceiling.

Why should I trust a vendor's own experiment?

You shouldn't — trust the data, not the vendor. Every number above is from Google Search Console, the method is the public checklist, and the post updates weekly so the claim stays checkable. Vendors who can't show you their own rankings are asking you to buy a method they've never been caught failing with. We'd rather show the test.

Update log

This post is a living document — the experiment continues in public. New dated entries every week.

• 2026-09-27 — Week 3 baseline published. Full method + day-by-day data as documented above. llms.txt guide holding ~pos 5; compare cluster at pos 2–9; commercial queries deep. /geo-agency, /pricing, /process shipped this week — watch next week for routing changes.
• Next update: week 4 — whether settled positions hold and whether the new routing pages start absorbing the agency-intent queries.

The takeaway

A new domain can get evaluated and positioned in weeks — not the months conventional wisdom assumes — if every page is built to be cited and every claim can be checked. That's the result we're publishing; the durability question gets answered in the log above.

If you want this run on a domain that already has something to lose: that's what the Full-Surface Audit maps, and Altyzo Managed is the fleet plus the people running it for you. If you'd rather run it yourself first, the 40-point checklist is the same one we scored ourselves against — unchanged, all 40 points, free.