AI visibility: what gets measured.
Deep Ocean focuses on brand mentions, cited sources and other brands in dated AI answers. The free AI Access Scan checks crawler rules and simulated request responses separately. Brand-review questions, platforms and scope are agreed individually. No public fixed price or strategy-call package applies. Introductory calls can be scheduled through Cal.com and email inquiries are open; paid orders, payments and confidential client document intake remain closed.
Two separate checks
The free access scan tests robots.txt permissions and responses to requests using crawler user-agent strings. It does not query AI answers or measure brand citations.
The brand review records actual AI answers to agreed customer questions, with mentions, links, other brands and collection conditions. It is an assisted review, not an automatic result from entering a website URL.
What a review includes
The proposed one-time review brings together the agreed questions, answer records and source links, a written summary of findings and priorities to investigate, and a discussion with Nicolò. Suggested next steps explain their evidence and may include no change or a need for more information. Your team decides and implements; implementation and ongoing monitoring are not included.
For example, a phone repair business could ask how answers describe its local availability and services. The homepage sample illustrates three questions and three platforms with fictional content. It is a format example, not a delivered client engagement or proof of usefulness.
Before an engagement, the written scope confirms the brand, market, language, questions, repeats, platforms, deliverables, fee and delivery date. There is no standard delivery promise or subscription. Any later comparison needs a separate scope. Introductory scheduling does not open paid orders or confidential intake.
From an answer to a next step
A factual discrepancy is checked against current information confirmed by the business. A cited source is inspected before a correction is proposed. Brand absence alone does not establish its cause, a page defect or a reason to seek placement in a directory. A later answer change does not prove that a correction caused it.
The value being evaluated is the usefulness of the combined record and interpretation for the team’s question. A few independent checks may be sufficient for a simple question; neither a larger report nor the demo establishes incremental value.
Reading the visibility map
Rows are questions and columns are AI platforms. Each cell links to the answer record. Use “Mentioned”, “Cited”, “Not observed”, or “Not captured”, with text labels as well as colour. Mention and citation fields remain separate; both may apply to one answer.
Report counts per platform and show valid answers and attempted captures separately. For example, a count must state its denominator and question set. Do not turn the sample into market share, recommendation probability or a universal rank.
Questions and collection conditions
Agree the market, language, question wording and platforms before capture. Record repeats, dates, surface and model where observable, account context, missing answers and failures. Separate category discovery questions from brand-named questions.
Answers vary across time and users. Small samples are exploratory. A later comparison requires comparable questions and recorded conditions; it does not establish why an answer changed.
Current availability
The free scan works now. Contact Nicolò to discuss your brand using public information. No payment, account connection or confidential upload is requested here.