GEO real case study: a brand-new site into Google’s AI answers in 1 month

Is it possible to promote a completely new website in Google AI answers? Yes, I managed to do GEO for a domain that was less than a month old, and below I share a step-by-step case study that even a schoolchild could repeat.
On 1 August 2025 I was asked to get a brand-new, local service site—bookbinding in London—into Google’s AI answers (AI Overviews / generative results). Classic SEO mattered, but the brief was explicitly GEO (generative en): structure and signal the content so Google’s generative layer would lift our page into its response.

Week 1: foundations that make GEO possible

Platform & domain (1–5 August).
We chose WordPress for speed of execution and control over on-page structure (headings, FAQs, tables, schema). For the domain, we registered pimlico-bookbinding.co.uk. Including the core service term (“bookbinding”) in the name is not a magic bullet, but it helps with topical clarity and entity matching—useful for both classic SEO and the LLM that composes AI answers.

MVP page and blogging (5 August).
We shipped a one-page MVP for bookbinder Maria Ruzaikina, then immediately began a blog cadence. The MVP carried the essentials for GEO:

• Answer-first sections (“What is fine binding?”, “How much does restoration cost?”) with 2–3 sentence definitions before the detail.
• Short, quotable blocks—lists, mini-tables, and FAQs that a model can lift verbatim.
• E-E-A-T scaffolding: real workshop address, phone, opening hours, and a concise bio.

Age & authority via redirect.
New domains lack history. Part of my playbook is to attach a relevant aged domain with a 301 redirect. Maria had a six-month domain in the same craft. Unfortunately it had no proper bookbinding meta and little topical clarity, so we located a better-aligned spare domain and pointed that one instead. The aim wasn’t “juice”, but entity continuity and faster trust.

Week 2: assets, signals, and human oversight

Images—originals first, generated only as filler.
We photographed the real workshop (press, finishing tools, leathers, gold foils) and used those images throughout. Where we lacked an angle, we used temporary generated visuals, clearly secondary to originals. GEO rewards verifiable reality; unique images send that signal.

AI-assisted copy with expert review.
We drafted articles in AI, but every piece was grounded in Maria’s real practice and signed off by her. We added expert sentences and short quotations (“Gold tooling is heat-applied at ~120–140°C…”), because generative systems favour specifics and attributions over generic prose.

Entity consolidation.
We linked to Maria’s Instagram and YouTube—years of documented work. These outposts act as entity corroboration: same name, same craft, same city, rich evidence.

Google Business Profile (ex-GMB).
The profile already existed. We cleaned the categories, added service areas, uploaded workshop photos, and posted a short update. GBP is not just for map packs; it reinforces the local entity Google’s LLM relies on.

Week 3: earned mention and internal depth

A lucky PR touch.
While scanning competitors we discovered an article that mentioned Maria as a specialist but didn’t link to her. We wrote to the site owner; they kindly added a link for free and asked for a fuller interview. This gave us a contextual, topical citation—the sort that both SEO and GEO value.

Service pages begin (from 5 September).
We started building proper service pages, first Fine Book Binding, then restoration, boxes/slipcases, and rebinding. Each service page followed a GEO-friendly pattern:

• Top paragraph with a one-sentence definition and a use-case list.
• Process steps (consultation → design → materials → binding → finishing) in a numbered list.
• FAQ schema with 5–7 concise Q&As.
• A short expert quote and one statistic where appropriate (e.g., archiving papers are acid-free, pH-neutral).
• Internal “See also” links to keep the topical cluster tight.

What we optimised for GEO, not just SEO

1. Answer-first writing. We opened sections with a 25–40-word definition that stands on its own.
2. Citable micro-blocks. Bullet lists, comparisons, and FAQs the model can paste as a ready paragraph.
3. Quotation & stats. We wove brief expert quotes and concrete numbers into the copy—these consistently test well for inclusion in generative answers.
4. Open access. No aggressive pop-ups, no partial paywalls. If the model can’t read it, it won’t cite it.
5. Freshness. We updated dates and added small new data points weekly.
6. Schema. We used FAQPage and LocalBusiness schema to help the system understand structure and locality.
7. Consistent entity signals. Same name, NAP details, imagery, and social profiles everywhere.
8. Unique visuals. Genuine workshop photos; generated art only as a stopgap and never near key explanatory blocks.
9. Internal linking. We connected service pages and blog posts with clear anchor text (“Fine binding techniques”, “Archival boxes”).
10. Polite PR. One relevant editorial link beat ten generic directories.

GEO – 1 month age website in top-2

On 18 September, our Fine Book Binding page surfaced Top-2 inside Google’s generative answer for the query “fine bookbinding in london”—precisely the goal. Classic rankings improved too, but the key win was visibility within the AI block, where users often stop reading.

How we validated: multiple clean sessions (signed-out, neutral location), query variants, and repeated checks over several days to rule out one-off personalisation. As always with generative results, visibility can fluctuate, but the page has remained a consistent candidate since the first appearance.

What moved the needle most

• Quotation + statistics. In my testing, adding a credible quote and one precise stat to the service page materially increased the chance of being lifted into the AI answer.
• Entity clarity everywhere. Domain name, GBP, image EXIF, and social consistency collectively helped Google “know” who Maria is.
• Service-page structure. The clear definition, the numbered process, and the FAQ block gave the model clean, copy-ready segments.
• That earned link. A sympathetic editorial mention with a contextual link provided third-party validation the model could reference.

Pitfalls we avoided

• Keyword stuffing. It harms readability and rarely helps GEO.
• Thin AI copy. Every article had workshop-specific detail (tools, materials, techniques) and expert oversight.
• Stock-only imagery. We prioritised real photos; stock was used sparingly and never as the hero visual.
• Redirecting a weakly related domain. We tested the aged domain’s topicality first; when it lacked bookbinding signals, we found a better match before applying the 301.

A simple checklist you can reuse

• Pick a platform you can edit fast (WP is fine).
• Ship an MVP page with answer-first sections and FAQ schema.
• Secure at least one relevant editorial link.
• Add expert quotes and one statistic to each service page.
• Use original photos and keep the content fully open.
• Align Google Business Profile to your copy (categories, photos, posts).
• Build service pages with definitions, steps, FAQs, and internal links.
• Refresh something weekly (a line, a stat, an FAQ).
• Track generative visibility with clean sessions and variant queries.

GEO isn’t a replacement for SEO; it’s an overlay that rewards clarity, credibility and structure. For pimlico-bookbinding.co.uk, the combination of expert-anchored copy, crisp service-page structure, real images and one timely editorial link was enough to push a young site into Google’s AI answers for a competitive local query—within seven weeks of launch.