How letAGENTcook Measures AI Visibility and Later Results

1. Establish the baseline

A Website Baseline Report begins with a Website URL and Brand Name, identifies five starting keywords and about five competitors, and generates up to 20 prompts that the user can edit, delete, or add before execution. The standard maximum is 20 prompts × 6 platforms = 120 executions across ChatGPT, Gemini, Claude, Grok, Google AI Search, and Microsoft Copilot.

An App Baseline Report begins with an App Store or Google Play listing URL, extracts five keywords, identifies about 5–15 relevant top-ranking competitors, and compares names, descriptions, keywords, and listing structure. Both report types are independent snapshots. Continuous history begins only after the relevant paid workflow is activated.

2. Record the execution environment

An observation is interpretable only when its environment is retained. When available, the record should include date and time, platform, model, signed-in state, account plan, region, language, web-search state, prompt, answer status, citation URLs, brand position, sentiment, recommendation condition, incorrect facts, and the reason for a failure or unavailable result.

Unsupported or unavailable fields remain unknown. They are not inferred. A failed or unavailable execution is recorded separately and must not be converted into a zero-visibility result.

3. Keep metrics and observation types distinct

The methodology can calculate or organize Discoverability Rate, Citation Rate, Third-party Citation Rate, Citation Absorption Score, Brand Mention Rate, Recommendation Rate, Top Recommendation Rate, Positive Recommendation Rate, Correct Fit Rate, Citation-to-Brand Alignment, and AI-assisted Conversion Rate when the required observations and authorized analytics data are available.

A link is not automatically a citation. A citation is not automatically a brand mention. A mention is not automatically a recommendation. A recommendation is not automatically a conversion. Owned citations, third-party citations, competitors, sentiment, incorrect facts, recommendation position, unavailable answers, and verified zero results therefore remain separate records.

4. Record the completed work

Each task should retain the original finding, affected prompt, page, listing, citation, keyword, competitor, or metric; the proposed scope; the user-approved action; completion time; and the reviewable output. Website work may produce a development plan and AI IDE prompt, but review, testing, approval, and deployment remain under the user’s control. Supported app-store text metadata may be changed only through an authorized official API workflow.

5. Recheck later observations

Later executions are compared with the baseline, execution environment, and completed-work record. The result should state what changed, what stayed the same, what failed, and what could not be observed. App-store workflows may check after 3, 5, and 7 days and in later cycles. Website workflows use later AI Visibility Tracking observations and optional locally authorized analytics context.

A hypothesis may be retained, adjusted, withdrawn, or marked inconclusive. The next iteration should follow the evidence rather than preserve an earlier recommendation merely because it was already implemented.

6. Limitations and attribution