David Shapiro / Practical AI search worksheet

The Three-Door Check

One prompt. Twelve questions. Choose where to investigate first.

Choose one real buying question. Run it three times in each consumer product you want to inspect. Keep products separate. Record dates, visible sources, missing AI answers, and failed checks.

Prompt: ______________________________________________________________________________________
Product / model / location setting: __________________________________________ Date: ____________________

01

Memory

Brand, PR, earned reach

  1. In how many runs is your brand named? ___ of ___
  2. Which competitors appear repeatedly?
  3. Does the product show brands before a visible search? Yes / No / Cannot tell
  4. When asked about your brand and category, is its description correct? Yes / Partly / No

Rare mentions or an inaccurate description justify an information review. Check your own facts and relevant independent coverage. They do not prove a training-memory problem.

02

Retrieval

SEO, technical access

  1. Do the visible sources include your site? Yes / No
  2. Check three related Google searches. How many show a relevant page from you? ___ of 3
  3. Can search crawlers reach and index the pages buyers need? Yes / No / Unknown
  4. Does your site state product facts, prices or pricing conditions, and fit in plain text? Yes / Partly / No

Missing sources and weak search coverage justify checking access, indexing, and coherent topic coverage. Visible citations are not a complete record of retrieval.

03

Selection

AEO, content design

  1. Does the relevant page or section answer the question early? Yes / No
  2. Can each important section make sense on its own? Yes / Mostly / No
  3. Are the facts current, supported, and dated where useful? Yes / No
  4. Does a competitor give a more direct answer with better evidence? Yes / No

A relevant page that fails to answer the question justifies a content review. An opened or cited page does not reveal why the product selected an answer.