Self-experiment: from new domain to 76 / 100 — and category-leading entity disambiguation — on fernandoangulo.com

Why publish a self-experiment

Most case studies in the GEO category are theatre: anonymized “B2B SaaS client,” cherry-picked metrics, no way to verify the baseline or the methodology. fernandoangulo.com is different on purpose. It is the personal site of one of Citable’s advisors and associates. The domain was registered in November 2025 with zero existing SEO history, zero backlinks, zero schema, and zero AI search presence. We’ve been running the same checklist on it that ships to clients — without budget constraints, without timeline pressure, and with one rule: anything we publish in the Citable playbook has to work on this domain first.

You can verify every number in this study yourself. Run the Citable AI Visibility Checker on fernandoangulo.com right now. The score, the passing signals, the failing signals, and the fix list are public.

About the site

fernandoangulo.com is a one-person personal brand site belonging to a Citable advisor and associate. It runs on Astro (same stack as citable.agency), hosted on Vercel, with bilingual EN + ES content. The content is editorial: long-form writing, project archive, personal bio. There is no commercial layer, no e-commerce, no ads. The site exists for two reasons: as a public credential anchor for the brand, and as the structural-fix sandbox for every methodology change Citable ships.

The starting position

The site started in the worst possible position any GEO playbook can address:

• Domain age: brand-new (no historical authority signal).
• Backlinks: zero (the owner had not done outreach).
• Brand mentions in the open web: roughly twenty, mostly in Spanish, mostly stale.
• Organic traffic from Google: under fifty sessions per month.
• AI search citations: zero across all four major surfaces (ChatGPT, Perplexity, Gemini, AI Overviews).
• Schema markup: none.
• robots.txt: default, blocking AI crawlers.
• Wikipedia / Wikidata entity: did not exist.

The interesting part is what was not going to fix this: no amount of writing more blog posts, no amount of “more backlinks,” no amount of social promotion. The site was structurally invisible. AI models could not extract anything from it because nothing was structured to be extracted, and they did not know what the entity “Fernando Angulo” referred to because there were thousands of other Fernando Angulos in the index with stronger entity signals.

This is the same starting position roughly 60% of small-business and founder clients arrive with. The diagnosis is identical. The fix sequence is identical.

What we did, in order

Month 1 — Crawler access + Organization schema

The cheapest, highest-leverage first move. Updated robots.txt with explicit allows for ChatGPT-User, PerplexityBot, ClaudeBot, and Google-Extended. Shipped Organization + Person schema on the homepage with full sameAs chain pointing to LinkedIn, X, GitHub, and the few professional profiles that existed. Validated through Schema.org and Google Rich Results.

Effect at month 1: Citable Checker score went from “untested” to 42 / 100. AI bot fetch logs started showing PerplexityBot and ChatGPT-User reading the site weekly.

Month 2 — Article + BreadcrumbList sitewide, content extractability rewrites

Added Article schema to every blog post. Added BreadcrumbList sitewide. Rewrote the opening 200 words of the homepage and the top six posts to lead with the direct answer in declarative sentences. Reformatted H2s into question-style headings matching the prompt patterns AI models actually retrieve against.

Effect at month 2: Score climbed to 51 / 100. First citations appeared on Perplexity for the owner’s name + niche queries.

Month 3 — Wikidata entity creation

This is the slow lever and the most important one for entity disambiguation. We created a Wikidata entry for the site owner, sourced it carefully (independent media coverage, professional credentials, brand registration), and added the complete sameAs chain referencing the site, LinkedIn, GitHub, X, and Crunchbase.

Effect at month 3: Score moved to 58 / 100. Gemini and AI Overviews began citing the site occasionally for niche queries. ChatGPT remained the slowest to update — it took until month 5 for the first ChatGPT citation.

Month 4 — llms.txt + FAQPage on cornerstone pages

Shipped /llms.txt at the root as an editorial map. Added FAQPage schema to the four cornerstone pages, with question text matching real AI prompts the owner’s audience types.

Effect at month 4: Score moved to 62 / 100. ChatGPT citations finally appeared (slowest surface, as predicted).

Months 5–6 — Maintenance + the authority draft

By month 4 the on-site structural levers are largely exhausted. Months 5 and 6 were about maintenance (re-running the 50-prompt audit monthly) and starting the hardest lever: off-site authority — the notability and reputation signals that structural fixes alone can’t manufacture. At day 180 the Checker read 64 / 100 (Patchy presence), with off-site authority the clear ceiling.

Months 6–7.6 — the authority lever moves, disambiguation resolves (day 180 → 230)

The day-180 update predicted that if the off-site authority work landed, the Checker would cross into the 70s. It did. Between day 180 and day 230 we ran the slow lever deliberately: dedicated content published on and off the domain, and a strengthened presence across high-reputation surfaces — Wikipedia, WikiHow, LinkedIn, and other authoritative profiles — all wired back to the entity through the existing sameAs chain.

Two things changed as a result:

• The score crossed a band. 64 → 76 / 100, moving from Patchy presence (60–75) into Solid baseline (75–90). This is the first time on-site structure and off-site authority compounded past the ceiling the earlier update named.
• The entity is now fully disambiguated — the headline result. Ask ChatGPT, Gemini, or Google “who is Fernando Angulo” or “how do I contact Fernando Angulo” and you get the right person, not one of the thousands who share the name. And for commercial intent, the category query — “who’s a specialist in GEO” — now surfaces him too. Disambiguation went from the site’s weakest signal to its strongest.

What the Checker says at day 230

The live result on fernandoangulo.com today:

• Score: 76 / 100 · Solid baseline. Up from 64 at day 180.
• Signals passing: Organization schema, AI crawler access, Wikidata entity — plus the off-site authority signals that were previously the ceiling.
• Strongest signal — entity disambiguation: brand queries and the “GEO specialist” category query both resolve to the correct person.
• Still open: deeper llms.txt curation and the long tail of sustained independent media coverage — the levers that separate Solid baseline (75–90) from Citation-ready (90+).

This is the score profile of a well-executed single-operator personal brand that has pushed past the ceiling structural fixes alone can reach — by doing the slow authority work most operators skip. The remaining climb into Citation-ready is a compounding, multi-quarter arc, not a sprint.

What this proves — and what it doesn’t

What it proves: the Citable methodology is not theatre. The same structural levers it audits in client engagements are visible, measurable, and improvable on an advisor’s own site — and the slow authority lever that caps most small operators (Wikipedia and off-site reputation) does move when you work it deliberately. Entity disambiguation, the single hardest thing for a common name, is fully resolved here. That is the result we point clients to.

What it does not prove: that GEO is easy. Seven-plus months of consistent structural work followed by a deliberate authority push — executed by an advisor and associate of a GEO agency on their own site — gets to 76 / 100, not 95 / 100. The remaining points are the slowest, most expensive levers: sustained independent media coverage and deep llms.txt curation. Those compound over years.

What it means for client work: if you are a founder, consultant, or small operator wondering whether GEO is worth investing in: yes, the structural lift is real and accessible. The methodology gets you from invisible (Citable score 30 / 100 or lower) into Patchy presence (60–70 range) inside 90–180 days on almost any domain — and, with the authority arc, into Solid baseline (75+) after that. From there the curve flattens and authority work takes over. The Citable retainer is built around exactly this shape: foundations sprint first, authority arc second.

What we’re measuring next

The day-180 update predicted that landing the off-site authority work would push the Checker into the 70s. At day 230 it did — 76 / 100. Two open experiments continue on fernandoangulo.com through the rest of 2026:

1. The climb into Citation-ready (90+) — sustained independent media coverage and deeper llms.txt curation, the levers that separate a Solid baseline from a Citation-ready one. This is the slowest arc, and the honest test of whether a single operator can reach the top band at all.
2. Quarterly re-audit cadence — re-running the full 50-prompt Share of Answer set quarterly to track decay vs growth on each surface. This is the same cadence we recommend to retainer clients.

Both will be published here as updates to this case study. The methodology is the report.