Prompt Coverage Map

What does a coverage map look like?

A typical map has one row per question and one column per engine, with a status in each cell. The table shows an illustrative layout with made-up results.

Question	Stage	Engine A	Engine B	Engine C	Priority
What is [category]?	Learn	Absent	Mentioned	Absent	Medium
Best [category] tools for small teams	Compare	Listed 3rd	Absent	Listed 2nd	High
[Brand] vs [Competitor]	Decide	Mentioned, wrong price	Mentioned	Mentioned	High
Is [Brand] reliable?	Decide	Negative claim from old article	Neutral	Neutral	High

Common statuses are absent, mentioned, listed with a position, cited, described inaccurately and described negatively.

Where do the questions come from?

Use real buyer language and not only the team's vocabulary. Useful sources include:

• Sales calls, demos and customer interviews.
• Customer support questions and chat transcripts.
• Search Console queries and site search.
• "People also ask" questions and forum threads in the category.
• Competitor comparison pages and review sites.
• Questions raised in earlier audits. See prompt audit.

Group them by buying stage, such as learning, comparing and deciding, and by topic. See query intent mapping, comparison prompts and decision prompts.

How do you build one?

1. Collect candidate questions and remove duplicates, keeping a mix of broad, comparison and brand-specific ones.
2. Write each in natural language, and test a few variations of wording.
3. Choose the engines your audience uses.
4. Run each question several times per engine, and save the full answers and cited sources.
5. Code each result with a fixed set of statuses, and note competitors named.
6. Add business weights, such as how important the question is to revenue.
7. Record the date, engines and method on the map.

Illustrative coding rules (example only)

• Mentioned: brand named in the answer text
• Cited: brand's own page appears as a source
• Inaccurate: a stated fact differs from the approved fact sheet
• Negative: framing is unfavorable, whether or not accurate

How do you use it?

The map shows where to look closer and what to do first. Rows where the brand is absent but the question is valuable become candidates for deeper diagnosis. See visibility gap analysis. Rows with inaccurate or negative answers point to source problems. See source contamination. Totals by engine can feed measures such as answer inclusion rate and citation share. The map is also a starting input for the GEO roadmap.

How do you keep it current?

Re-run it on a schedule, such as monthly or quarterly, and after launches, news or engine changes. Keep the core questions fixed so trends are comparable, and add new ones in a separate section. Record changes to the method. Retire questions that no longer reflect how buyers ask.

What are the limits?

The map reflects sampled answers. Answers vary between runs, accounts, places and dates, and wording changes results, so one cell is weak evidence. Engines do not publish their selection rules. A question set can miss how real buyers phrase things, and being present is not the same as being recommended. Avoid reading small changes as proof. No outcome is guaranteed.