Trigger and output
Run this for a buyer question where your company is missing, misrepresented or supported by weak sources. The output is a source-gap record with captured evidence, a clear diagnosis, an owner and a next action. It is not a citation guarantee.
Keep a fixed monthly question set. New questions go into a separate exploration set so changing the prompts cannot masquerade as better visibility.
Prepare the question set
Use questions from calls, CRM notes, proposals and search research. Cover problem, method, comparison and vendor selection. Write the target buyer and why each question matters commercially. Avoid a set made only of prompts that explicitly include your brand.
Use 20 to 30 core questions. Record the engine and mode, locale, sign-in state and sample frequency. Where model details are unavailable, write “not shown”. The tracker should preserve the raw observation rather than fill unknowns with assumptions.
Run the workflow
- Ask the exact recorded prompt in the selected engine. Save the answer or screenshot with its date and settings.
- Record whether your brand is named. Separately record whether a source you own is cited.
- Open the cited URLs. Identify the claim or comparison each source supports, its date and its source type.
- Compare the answer’s framing to your approved positioning and evidence. Mark errors and unsupported claims.
- Diagnose the gap: missing page, thin evidence, weak discovery, wrong category framing or missing external coverage.
- Create a specific next action. Name the page or source surface, proof needed, owner and review date.
- Publish only after the relevant claim and source checks. Re-run the same question in subsequent samples and preserve earlier results.
A source-gap receipt
- Question
- How should a B2B team connect Google Ads leads to pipeline?
- Observed answer
- Uses implementation documentation and an undated agency article
- Gap
- No practical source explaining definition, handoff and import reconciliation together
- Next action
- Publish a pipeline decision playbook with an import workflow and a dated field note
- Measure
- Future sampled citations, resource use in sales and qualified conversations
No live engine result is claimed in this example. Download the question tracker and the 25-question starting set. Replace the sample questions with your buyers’ language before calling it a baseline.
Exceptions and interpretation
| Observation | What to do |
|---|---|
| Answer contains no citations | Record “no citations”; brand presence is still a separate observation |
| URL is inaccessible or changes | Preserve the original captured answer and mark source access status |
| Engine response differs on a repeat | Keep both samples; report frequency rather than the best screenshot |
| Brand named, page not cited | Record a mention, not an owned-source citation |
| Page cited for the wrong claim | Fix the framing or evidence, then observe future samples |
| Prompt explicitly asks about your brand | Keep it separate from unbranded discovery questions |
| A page is published but not indexed | Inspect technical discovery before assuming an editorial failure |
Validation and cadence
Check a sample manually to confirm the cited page supports the recorded claim. Have the editor review any proposed factual correction. Save the exact prompt and raw answer so a second operator can inspect the diagnosis.
Review core questions monthly and changeable platform references more often. Report sample count, engine and question set with every visibility measure. Use ordinary indexability, readable content and internal links; Google does not require special AI markup for its AI search features. Google Search Central.
Commercial use
Use the source-gap record to explain a decision to a buyer or to brief the next piece of work. Connect it to the resource sent, the conversation and any resulting opportunity. The aim is useful evidence that improves a buyer’s understanding, with sampled visibility as one supporting measure.
Publication and source record
Published on GTMhub: . Last reviewed: .