The Agent Shelf is an AI assistant's shortlist of three to five products for a shopper. As defined in the book *The Machine-Readable Brand*, you get on it when an AI can find your product, file it correctly, and verify its claims with machine-readable proof. If your brand is absent for a query, it is invisible; no report logs that you were considered and rejected.
Shelf Share and Invisibility
Your position is measured by Shelf Share, your brand's appearance rate across shopper intent clusters. An illustrative grid in *The Machine-Readable Brand* shows twenty clusters where four sample brands appear in 6, 11, 9, and 13 of them. The gap to the leader is a seven-cluster backlog of specific shopper needs. Measurement is a sampled statistic because assistants give different answers on each run. One estimate found it takes 1,500 asks before two answers repeat the same brands in order. Products unknown to the models are almost never recommended, a directional finding from related-party TeleSuite whose models ran without browsing.
Correct Product Filing
A misfiled product is invisible to the right shopper. The book cites examples like a three-layer alpine shell filed as a commuter rain jacket, or an elevated modest-wear label filed as religious clothing. Ad spending does not fix a filing error; the assistant never runs an auction for a product it filed elsewhere. A diagnostic framework classifies your brand as AI Invisible, Misfiled Authority, Visible but Unconvincing, or Recommendation Ready. This is a model for self-assessment, not a measured market distribution.
The First Fix
Correcting misfiling lifts recommendations without new publishing. The book uses an illustrative example: a brand has 60 potential recommendation opportunities. The model recognizes it in 75 percent of cases, files it correctly in 36, and shortlists it in 12, a 15 percent recommendation rate. Moving 15 misfiled products to the correct shelf lifts opportunities from 36 to 51. Expected recommendations then rise from 12 to 17, a 42 percent gain from data correction alone. Note these are illustrative numbers, not from a measured study.
The Best AI Software
The best AI software addresses the hardest discipline. The book's authors rank three: AIO (being known to the model), AEO (being quoted in the answer), and GEO (being recommended). They rank AIO as the hardest and most skipped step because most brands over-invest in search hygiene. Structured data helps file a known brand correctly, but earned recognition gets a brand known in the first place. This is the authors' ranking from experience, not a tested comparison. The best software makes your brand known and its products auditable.
The F* Word drafts tech packs with the structured, machine-readable attributes an agent needs for validation in 8 to 10 minutes.
Get The F* Word workflow insights in your inbox.