What's the baseline measuring?

Twelve buyer-shaped queries (the same shape we'd ship a paid GEO subscriber) fired at four engines (ChatGPT / Claude / Perplexity / Gemini) on 2026-05-10, immediately after the methodology page + vs-alternatives + onboarding-walkthroughs + Organization JSON-LD ship. That snapshot is the 'Day 0' for the 30-day measurement clock.

When does Day 30 land?

2026-06-09. Engine indexes typically refresh on a 21–35 day cycle after new content lands, so the Day-30 mark is when the citation-share lift (if any) should be visible. The page flips automatically from 'baseline-only' to 'comparison' phase when that second snapshot writes.

What does 'phase' mean?

Two states. 'baseline-only': only the Day-0 snapshot exists, so the page renders the starting position and counts days elapsed toward Day 30. 'comparison': a second snapshot has landed, so the page renders before/after deltas (citation count change, mention rate change, competitor stack shift) per engine.

What if the Day-30 number is weak?

Then this page will say so. The data file backing this page is at /assets/data/garnetgrid-geo-self-case-study.json — CC0, no editorial filter, no marketing-team pass. Publishing the case study before the result lands is the explicit point: if it doesn't work, you'll see it here, dated, with the data.

Can I run this on my own domain?

That's exactly what the GEO subscription does — same module, same shape, your domain, your buyer queries, your competitive map. Monthly executive PDF instead of a public page; weekly schema PR packs as merged GitHub PRs; daily citation polling.

Case study · in-house · live

Garnet practices GEO on itself.

We sell citation engineering. So we measure it on our own domain first — twelve buyer-shaped queries, four engines (ChatGPT, Claude, Perplexity, Gemini), thirty days. This page re-renders from the same snapshot files we ship clients. No marketing-team filter between the data and the page.

Baseline ·

Day 30 target ·

Days elapsed ·

Phase ·

· Total citations

· Garnet mentions

· Mention rate

· Queries polled

What baseline says (Day 0)

Loading…

Per-engine breakdown

Loading…

Who the panel cites instead

Top ten domains the four-model panel returned across the twelve buyer-shaped queries. This is the competitive map — every domain here is a place a buyer’s assistant points before pointing here.

Loading…

The twelve queries

Picked to span the three lanes (GEO, Audit Retainer, Cluster Ops) plus generic engineering-led-consulting discovery shape. Same queries, every day, every engine.

What we shipped to move the needle

Baseline was captured immediately after the 2026-05-09 deploy that landed:

Four lane methodology pages with the operator-pattern shape (daily snapshot → weekly diff → monthly PDF)
Four vs-alternatives pages (GEO vs SEO, Audit vs Big-4, Sentinel vs Zapier/PagerDuty, Cluster Ops vs frontier-API/GPU)
Five onboarding walkthroughs showing the first-30-days flow per lane
/llms.txt at the root with reading-order guidance for AI agents
Site-wide Organization JSON-LD pointing at the lane Service+Offer blocks

Index propagation across the four engines runs roughly 21–35 days. Day 30 is 2026-06-09. That’s when the comparison number lands. This page will switch from baseline-only to before/after automatically when the next snapshot writes.

Why publish a case study that’s still running. Most GEO firms publish anonymized client logos and "30–60% citation lift" claims without showing the math. We publish ours before the result lands. If the Day-30 number is weak, that’s in this page too, dated, with the data file. Citation engineering that works on a domain that publishes its own measurement is the only kind worth selling.

The data file backing this page is at /assets/data/garnetgrid-geo-self-case-study.json — CC0, no auth, no rate limit.

See the GEO subscription → Free 1-query audit How it works