Entity reclamation is not a website project

The internet did not forget The Tinbasher. It just lost the forwarding address.

Paulwoodhouse.com and The Tinbasher are now a live GEO experiment. This site is not simply reporting the case study. It is one of the interventions being tested.

The website disappeared. The domain went away. Years later, search and AI systems could still describe the business blog, connect it to me and find third-party sources that remembered it.

That is the interesting part of the story. Not that an old blog ranked. It did not have a website left to rank.

The website disappeared, but enough independent evidence remained to connect the work to the right person.

A website is evidence, not the entity

Marketers naturally treat the website as the centre of the digital universe. That is understandable because it is the thing we control. We can rewrite it, measure it and put a conversion form on it.

AI systems are not limited to what a company says about itself. They encounter businesses through trade coverage, books, directories, profiles, archives, regulatory records, interviews and the relationships between all of those sources.

A company can therefore be recognised without an active website. Recognition can also be fragmented, attached to an old name or connected to sources that no longer represent the current business.

The job is not always to create an entity. Sometimes the entity already exists and needs to be reclaimed.

Reclamation means proving continuity

Entity reclamation begins with a harder question than “What schema should we add?”

It asks whether the current organisation can prove that it is the legitimate continuation of the history, names, people, work and third-party recognition that already exist.

That requires evidence. Historical aliases have to refer to the same business. Third-party sources need to be classified correctly. Identity profiles are not the same thing as articles that merely mention the company. Dates and ownership changes matter. Name collisions matter.

Only after that reconciliation work does the website become useful as the new canonical home for the entity.

Structured data clarifies, but it does not prove

Structured data can state the identity clearly. It can connect a person to an organisation, a current name to a former name and an official website to credible identity profiles.

It cannot turn an unsupported claim into a fact. It cannot make an unrelated press article become an identity profile. It cannot guarantee that an AI system will cite a page.

Schema helps systems resolve and exchange information about an entity. It works best when the underlying identity and visible evidence are already coherent.

That is why good GEO can improve SEO more readily than conventional SEO improves GEO. Entity reconciliation, clear ownership, useful content and stronger source relationships also help search engines understand the site. A title-tag rewrite does not automatically repair a fragmented identity across the wider web.

The measurement problem

Treating GEO like a keyword-ranking project produces bad decisions.

Ranking and traffic still matter when the commercial journey involves search. They simply do not describe the whole system.

Entity work needs separate measures for recognition, canonical identity, retrieval, citation, category consideration and commercial response. A model recognising a company by name is not the same as recommending it for a category. A citation is not a lead. Adding structured data is not proof of impact.

Those states change on different timescales and require different evidence. Collapsing them into one “AI visibility score” makes the dashboard tidier and the analysis worse.

What The Tinbasher can and cannot prove

The Tinbasher provides a useful entity-reclamation experiment because its historical footprint predates the current site and remained discoverable after the original website disappeared.

The experiment can document what sources persisted, how systems described the entity before any intervention, what identity signals were added and whether later answers or citations changed.

There is a second part to the experiment. Paulwoodhouse.com was a new website with none of The Tinbasher's domain history. It was nevertheless able to connect my current identity to that older entity, become part of the evidence AI systems could use and precede an update to the relevant AI Overview.

That makes this more than a story about dormant authority surviving. It is also a test of whether a new, well-connected source can help resolve an old entity more accurately. The before-and-after record matters here: the query, dates, answer wording, citations and site changes all need to be preserved before this is presented as a causal result.

It cannot prove that every established company has reclaimable authority. It cannot prove that one technical change caused every later response. It cannot promise another business the same result.

That boundary is not a weakness. It is what makes the case study credible.

Start with the entity, not the content calendar

An established B2B company considering GEO should not begin by commissioning a pile of AI articles.

It should begin by establishing what already exists:

• how the company is recognised.
• which historical names and relationships belong to it.
• which sources support that identity.
• where current and historical signals conflict.
• whether the company appears in relevant category decisions.
• and what a buyer or an AI system cannot currently verify.

Only then does it make sense to decide what the website, structured data, content and external validation need to do.

The point is not to make the company look busier online. It is to give its real authority a current, verifiable home.