Architect correction tickets and held verification into the concept.

Amend the black-box solution: audit findings become per-platform
correction tickets; held multi-platform verification is the moat.
Frame as work to design, not built.

docs/SELF_SERVE_AI_DISCOVERABILITY_CONCEPT.md

@@ -63,11 +63,29 @@ Think of it as three frames:
 
 **Frame two — middle.** The claim: one clean NAP the owner stands behind. That is the only version we treat as authoritative for *this* business. Fixing GBP still matters — it is often the best place to start the claim — but the point is the **declared truth**, not loyalty to any one logo.
 
-**Frame three — right.** Corrections flowing out. The system’s job is to make the other copies **match the claim**, including the ugly middle layer most owners never see: **syndication**. Aggregators that feed directories and assistants must be visible on the scoreboard and fixable in the same story. If you only polish the claim surface and ignore syndication, the black box regenerates overnight.
+**Frame three — right.** Corrections flowing out. The system’s job is to make the other copies **match the claim**, including the ugly middle layer most owners never see: **syndication**. Aggregators and platforms that feed directories and assistants — Yelp, Bing, Foursquare, and the rest — routinely **stomp on** a small business’s source of truth with stale hours, old phone numbers, and leftover addresses. If you only polish the claim surface and ignore those copies, the black box regenerates overnight.
 
 Matching NAP is the floor — **presence** with consistent facts. **Recommendability** (would an assistant pick you for a real customer question in your genre) still needs categories, relevance, and the rest of this product story. Opening the black box is how we stop the floor from moving under the owner’s feet.
 
-Assistants answer from whatever they can find and trust. When five copies disagree, the assistant is guessing in the fog. When one claimed truth is pushed outward and the syndicators are brought into line, the fog thins. Strategic bet in plain words: **stop competing with your own stale twins. Claim once. Correct outward.**
+Assistants answer from whatever they can find and trust. When five copies disagree, the assistant is guessing in the fog. When one claimed truth is pushed outward and the other copies are brought into line, the fog thins. Strategic bet in plain words: **stop competing with your own stale twins. Claim once. Correct outward.**
+
+### What we have to architect (not built yet)
+
+The direction above is the bet. Below is the **solution we still have to design and build** — not work already done.
+
+#### 1. Every audit finding becomes a correction ticket
+
+Today an owner who sees wrong hours on Yelp, a dead phone on Bing, and an old suite on Foursquare has to decode three different “suggest an edit” paths. That is the chore we remove.
+
+For each platform that still shows stale or wrong data, the product turns the audit finding into a **correction ticket** with the same shape every time: which field is wrong, what value is showing, what the right value is, and a link to **Google Business Profile** (or the owner’s claimed truth) as proof. Then the owner files that platform’s own correction channel — Yelp’s suggest-an-edit page, Foursquare’s suggest-edit API (`POST /places/{id}/suggest/edit`), Bing Places feedback, and the same pattern for others as we map them. Claim it is wrong. Show proof. Wait.
+
+Our side of the architecture: **mint that ticket automatically from the audit** — pre-filled, one click to file — instead of making the owner hunt which form goes where. The platforms still own their review queues. We own the translation from “your Presence score is broken here” into “here is the exact correction this platform will accept.”
+
+#### 2. Held verification across platforms is the moat
+
+Business verification is the tax. Every platform re-verifies **independently**. Yelp wants its documents. Bing wants its own proof of ownership. Foursquare has its own checks. There is **no shared credential** and no way to say “I already proved this to Google.” Each review takes **days to weeks**, and without standing as a verified actor, corrections get slow-walked or quietly reverted when the next stale feed arrives.
+
+A white-label correction service that **holds verified status per platform** — and files on the owner’s behalf without re-uploading a lease five times — is what makes corrections **stick**. That held, multi-platform verified status is the **moat**: the audit finds the drift; the ticket files the fix; the held verification is why the fix does not get stomped again tomorrow.
 
 Longer term, the Northstar — the destination we aim at but do not build first — is an **industry graph**: a shared, structured map of local businesses that many products and assistants could use. That is not version one. Version one earns the right to that graph by helping owners see how AI sees them today and fix what is broken, with their consent.