Citability Field Lab
Research corpus

Field notes and answer reviews

The research index arranges Citability Field Lab materials into a few rows of records: notes from individual runs, state-level answer reviews, regional citability gap cards, short methodological notes, and query archives. Materials stay in series built around comparable runs, where a single oddity has begun to form a repeated contour. Taken together, the index feels like a box of labeled field samples rather than a news feed. Each row keeps its region, category, and failure type attached.

A repeated regional citability gap becomes visible when runs under the same documented conditions keep returning the same kind of split between a business name, region, category, and citation trace.

How a Repeat Run Separates Noise from the Gap

A final meta-map of the series: how repeated AI-search runs separate random noise from a regional citability gap.

A missing local company is not automatically invisible. In the lab's run series, entity skip is recorded only with its prompt wording, Regional frame, model, comparison context, and nearby citation traces, because absence can come from regional framing, category widening, or weak source support.

When a Local Company Disappears From Similar Prompts

Why entity skip is recorded as a narrow failure in a specific Field run, not as proof that a business is fully invisible to AI.

An omitted company is a field observation, not proof of total AI invisibility; the lab records the prompt, region, model, comparison, and source conditions before treating entity skip as a finding.

Why an Omission Is Not Full AI Invisibility

The material explains why entity skip should be read as a narrow observation by query and region, not as a verdict on a company.

The material treats commercial roofing category drift as a failure of specialization: AI-search may cite a real local contractor while widening the description toward general contractor language and weakening the source-to-service tie.

When a Commercial Roofer Becomes a General Contractor

How the query commercial roofing north texas widens into general contractor language, and where the citation trace does or does not support that shift.

Directory dependency appears when an AI-search answer treats a broad listing as the practical voice of a local business, letting generic directory language outweigh a thinner or less consistently surfaced company source.

Why a General Directory Sounds Stronger Than the Company Site

A look at directory dependency: AI-search may retell a local business in directory language, even after the company site has been updated.

A citation can sit beside a local-business claim without carrying it. In the lab's model comparisons, the useful question is not whether a source appears, but which part of the answer it supports: the entity, the region, the category, or only a directory-level trace.

Where the citation hides beside the answer

How ChatGPT, Perplexity, and Gemini attach sources to a local business: to a category, an address, or a weak directory.

Changing only the city frame can reshape the local list without changing the service intent; the lab treats this as a regional substitution pattern when the answer imports a neighboring market template.

How One City Changes the Local Business List

A repeatable run with only the city frame changed shows how AI-search can move a local list toward a neighboring market.

The material frames HVAC regional substitution as a map-level failure: the service category can remain stable while the county, service area, or citation trace quietly shifts the company out of the intended local frame.

Why an HVAC company drifts into a neighboring county

A field note on regional substitution in independent HVAC repair queries, where the model holds the service but moves the local map into a neighboring county.

A local citation can confirm that a business exists while failing to support the service description attached to it, making citation-description split one of the quietest and most persuasive regional citability gap patterns.

When a Source Confirms the Address but Not the Service

A look at citation-description split: a source can confirm the business address while failing to support the service AI-search assigns.

An old third-party listing can become a sticky source seam: the model keeps the current local name, while a stale branch, region, or service tag leaks into the description. In Citability Field Lab notes, this belongs to citation-description split and directory dependency, often overlapping with category drift or regional substitution.

How old listings pull models into stale categories

How an old branch listing can pull a former district or service into an AI-search answer and trigger citation-description split.

In the lab notes, model comparison shows that ChatGPT and Perplexity can miss the same local business in different places: one smooths the category while the other exposes sources that do not fully support the description.

What Changes Between ChatGPT and Perplexity

A comparison of one local query shows how models hold companies, categories, and citation trace in different ways.

The material treats bookkeeping category drift as a narrow local failure: the company name may remain visible while the category, client segment, and citation trace loosen enough to produce a misleading AI-search answer.

Where bookkeeping firms turn into tax software

A look at how AI-search keeps a local bookkeeping firm's name while sliding it toward tax preparation, payroll software, or a broad financial-services category.

Rural urgent care queries expose regional substitution when an AI-search answer keeps the medical intent but replaces a thin rural frame with a nearby city, hospital network, or county-level source that only partially supports the local description.

Where Urgent Care Loses Its Rural Frame

How AI-search widens a rural Maine urgent care query into a city frame and weakens the answer's rural grounding.

Each review shows where the AI-search map gave way.

The index helps move from the loose feeling that "the model got it wrong" to a sharper reading: what happened to the entity, category, region, or source.

View the reviews