Does Deep Ocean guarantee your brand gets cited by AI?
No, and neither can anyone else selling this today. Nobody in this category has published a controlled record of one tracked query flipping from uncited to cited next to the dated intervention that caused it. We held our own promise to that bar, it did not clear, so we withdrew it and published why. What we back instead is written below, and it is a warranty we can actually pay out on.
No. We do not sell an outcome guarantee on AI citations. What we back today is an Evidence Integrity Warranty on the Sprint: every covered finding ships with its dated raw record, including the request, user-agent, timestamp, and HTTP response for each access check. A material finding with no valid record behind it, and not curable, refunds the full $490.
We held our own guarantee to the evidence bar and it did not clear (2026-07-09)
We had published a 90-day citation guarantee: cited in the agreed tracked answers by day 90, or we work free. It was a promise we could not yet back with evidence. Other vendors do publish before-and-after visibility and citation numbers; some are substantial. What we have not found anywhere, our own brand included, is a specific tracked query documented flipping from uncited to cited next to the dated intervention that caused it, on a protocol a buyer can reproduce. That is the bar, and our own meter has not cleared it: our brand has been sampled on a published panel since July 2026 (2026-07-08 on panel v2, 2026-07-16 on panel v3, a versioned break, with gaps published rather than backfilled), and has not yet flipped a single tracked query from uncited to cited.
To be exact about the distinction: a vendor case study that reports a brand's overall visibility or citation share rising is real evidence of movement, but it is not the same as a controlled, query-level record that isolates one intervention against one answer, published failures included, on a published protocol with the dated captures kept. That stricter record is what we hold ourselves to before we would stake a refund on the outcome, and it has to hold for us first.
We are not announcing a condition for the outcome guarantee's return. The meter stays public, the protocol stays published, and our tracked citation number is published in a dated ledger, zeros and missed runs included. If the guarantee ever changes, it will arrive with its own dated evidence, not as a promise in advance. This is the same build-in-public judgment we run everywhere else on this site, applied to the promise that mattered most.
Nicolò Brignoni
Founder, Deep Ocean · follow the work
What the Evidence Integrity Warranty covers
| Step | What it is | Price |
|---|---|---|
| Visibility Sprint | The 48-hour diagnostic: a systematic panel of buyer questions, who gets named instead of you and from which sources, a crawler access check per engine, and a prioritized direction. Every covered finding ships with its dated raw record; this is the product the warranty applies to. | $490 |
| Foundation | In design, not yet available for order: gated on a legal entity, jurisdiction, contract, and per-engagement scope of work that do not exist yet. As designed, the one-time build the Sprint scopes: access fixes, a tracked panel with a baseline evidence report, source briefs, and the first answer-shaped content. Not covered by the interim warranty on this page. | Price opens with orders |
| Watch | In design, not yet available for order. The monthly meter on its own: your panel re-run every month, with drift, new entrants, and source changes reported. It sells measurement, not an outcome, so it carries the same evidence discipline as the Sprint and no outcome guarantee. | $290/mo, opens with orders |
| Care | Planned future continuation, not sold today. The fuller recurring layer that would continue what Foundation starts, with ongoing work on top of the meter rather than the meter alone. Offered once a demonstrated flip exists and your meter shows movement worth retaining. | Price opens with orders |
| Prices for the two unopened products came off this page on 2026-07-26, the same reasoning as the partner rate card on 2026-07-24: a price you cannot order from does not help you plan. Each price publishes when its product can actually be ordered. | ||
Agencies can buy the Sprint white label at the retail price today; wholesale pricing publishes together with the wholesale terms at deepoceanstudio.com/partners before any wholesale order is accepted. All prices in USD.
What the warranty promises, and what it does not
The warranty applies to the Sprint and it is about the record, not about repetition. Every covered finding ships with its dated raw record: for an access check, the URL, the exact user-agent, the timestamp, and the HTTP response returned to a simulated request; for an answer finding, the dated capture of what the engine returned at sample time. An editorial error, or a record that exists but was left out of delivery, is cured within 5 business days. A material finding with no valid contemporaneous record behind it, and not curable, refunds the full $490. Claims within 30 days of delivery, in writing to hello@deepoceanstudio.com; the interim terms state the procedure. What is never promised: that a new run returns the same answer or the same HTTP status. AI answers vary between runs, and server responses vary with IP, location, time, CDN, and WAF behavior; that variation is not a defect. A crawler-access check is a user-agent simulation: it records how the server responds to a request carrying that crawler's user-agent, not proof that the real crawler, identified by its own IP, reached your site, which only your server or CDN logs can confirm. It says nothing about whether an AI engine chooses to cite you afterward: the engines decide which brand gets cited, and we do not.
The honest boundary. What we work are the inputs the evidence links to citations: presence in the sources the answers draw from, content at the exact questions asked, and crawler access that actually resolves. Presence in those sources is the evidence-supported lever. A demonstrated flip from that work does not yet exist, ours included, and the placement-to-flip mechanism is under live public test on our own brand first, published at The AI Citation Index.
How to audit the meter yourself
Before any engagement starts, we agree on a fixed panel of buyer-shaped queries, the exact questions your buyers ask an AI engine. Today each query is sampled in ChatGPT, the engine your buyers verify with, via a published browser protocol on an intended weekly cadence, missed runs shown as gaps, and every domain named in the answer is recorded, engine-versioned and dated to the week it was sampled, because retrieval mechanics change. Additional engines join the meter as they are armed, visibly, never silently. You can repeat the protocol on your own phone; the answer itself can vary between runs, which is why each row is dated and kept as a capture. It follows the same protocol as the public record we keep on ourselves: The AI Citation Index.
Common questions
- What happened to the 90-day guarantee?
- We retired it on 2026-07-09. We were not willing to keep selling a citation-outcome stake before any flip existed to back it, ours or anyone's. See the note above; it is the same reasoning, plainly.
- What does the Evidence Integrity Warranty cover?
- Every covered finding in the Sprint diagnostic: the sources named, the crawler-access results, the retrieval-rank check. Each finding ships with its dated raw record; access checks record the URL, user-agent, timestamp, and HTTP response returned to a simulated request, and sampled answers are delivered as dated captures. Editorial errors and delivery omissions are cured within 5 business days; a material finding with no valid record behind it, and not curable, refunds the full $490. Claims within 30 days. A new run is not promised to return the same answer or the same HTTP status.
- Will an outcome guarantee come back?
- Not on any announced trigger. No demonstrated flip meeting our protocol exists yet, ours included. We publish the tracked number weekly either way; if that ever changes, it will arrive with its own dated evidence, not as a promise in advance.
- Does a high readiness or hygiene score mean my brand will get cited?
- No. A high AI readiness score does not predict whether AI answers cite you. We checked it on ourselves first: our own site scores 94 on standard readiness checks and was named in 0 of 11 tracked answers on The AI Citation Index. It holds beyond us too: SE Ranking's 300,000-domain study found llms.txt has no measurable citation correlation, and Google's own guidance says schema and AI-specific markup are not required to be cited. In our own sampling we have also seen a near-perfect-readiness tool cited zero times in a category answer while a lower-scoring rival was cited, on a single dated sample (2026-07-08); counts like that move run to run, which is why we publish the pattern here, not a scoreboard.
- Which engines do you track?
- Today the weekly meter runs on ChatGPT, sampled via a published browser protocol, dated to the week it ran, because that is where your buyers verify. Perplexity, Claude, and Google AI Overviews join the meter as tracked engines when they are armed, visibly, never silently.
- Do I have to start with Foundation?
- No. The Visibility Sprint is the first paid step, and today it is the only one: you see the real AI answers and the source map, then decide whether to continue once Foundation opens for order. The Sprint is the diagnosis; nothing else is committed, or committable, until you have it.