Send a regional oddity from an AI-search answer
The form is for cases where AI-search describes a local business strangely: skips the company, replaces the category, confuses the region, ties the answer to a weak source, or surfaces a broad directory instead of a local trace. The lab reads these messages as incoming field notes; promotion requests do not belong here. The strongest examples include the prompt itself, the region, and a short note on what felt wrong.
Questions
What does Citability Field Lab study?
The lab studies how AI-search systems cite, describe, and attach local businesses in the United States to service categories. Its main interest is regional discrepancy: why a similar prompt in one state can produce a coherent answer, while another turns into a directory with the wrong label. The materials are built around observations, comparisons, and the regional citability gap typology.
What falls outside the lab's work?
Advertising campaigns, promises to fix model output, and rankings of "best" companies fall outside the lab's work. The team also does not treat one omission as proof that a business is invisible to all AI systems. One run is a sample; it is not a verdict.
How should I read the lab's materials?
Start with the region, service category, and failure type. If a review marks entity skip, category drift, or citation-description split, it is describing model behavior, not judging the company. The lab reads an answer like a map: where the label matched, where the contour slipped, where the source was sewn on sideways.
Can I suggest a review topic?
Yes, if there is a concrete prompt example and a visible oddity in the AI-search answer. The most useful submissions include the region, business category, original prompt wording, and a short explanation of what looked wrong. Broad topics without an observable answer usually stay outside the corpus.
What happens to incoming examples?
The lab reviews messages manually and treats them as incoming field notes. Examples with a region, category, and original prompt are easier to compare with other runs. Without those details, a message usually remains a weak corpus note rather than a full review.
Which messages do not fit?
Requests to raise visibility, promotional pitches, and abstract questions without a model answer do not fit the form. The lab works with a concrete oddity: a skipped company, category mixing, regional substitution, or a weak tie between source and description. If the region and prompt wording are missing, the case is hard to turn into a field sample.
What language are the materials in?
The corpus works in English because the queries and local business contexts belong to the United States market. Prompt wording is often preserved exactly, even when it is awkward, because changing language would change the observation. Public materials keep that query language stable.
Send an example
Send concrete AI-search oddities: region, business category, example prompt, and a short description of where the answer drifted sideways. Useful cases include company omission, category mixing, regional substitution, weak ties between source and description, or dependency on a broad directory. Do not send requests to raise visibility, promotional offers, or abstract questions without an example answer. After a submission, the example enters manual review; without a precise region and prompt wording, it remains a weak field note.
A good example starts with exact prompt wording.
The clearer the region, category, and oddity in the answer, the easier it is for the lab to understand whether the case belongs in its observation corpus.
Send an example