AI ANSWER EVIDENCE AND CORRECTION WORKSHEET
Branding Tatva
Version: 12 November 2026
Guide: https://brandingtatva.com/insights/generative-engine-optimisation-guide

Purpose
Keep a reviewable record of what an answer actually said, which page it cited
and whether a factual correction appeared in a later observation.
This worksheet makes no claim about rankings or future recommendations.
Save your completed copy privately. Remove personal information and obtain
permission before sharing client material or publishing conversation captures.

1. DEFINE THE SAMPLE BEFORE RUNNING IT
Business and canonical website:
Business facts reference page:
Service and buying decision:
Target market: Canada / Australia / UAE / UK / USA / India / other
Exact planned questions:
Number of planned observations per question:
Platforms/products and search settings:
Selection rules and planned recheck interval:

Keep branded fact checks separate from questions that do not name the business.
Use the full planned sample, including absence and unsuccessful observations.
A target market in a prompt is not evidence of the reader's actual country.

2. ONE RECORD PER OBSERVATION
Observation ID:
Exact question:
Did the prompt name the business or provide its URL?:
Date, time and timezone:
Platform/product and displayed model, if available:
Search setting, language and relevant prior conversation:
Requested market:
Actual audience country, only if independently known:
Status: completed / unavailable / failed
Reason if unavailable or failed:
Full response capture or export reference:
Exact URLs linked as citations in the answer:
URLs present only in a Sources panel:
Unlinked business mention, with exact wording:
Explicit recommendation, with wording and supporting context:
What claim did each cited page support?:
Business absent from this completed response?:

A Sources panel listing alone is not a citation in the answer.
A citation supporting a definition is not a recommendation to hire its publisher.
Leave unknown conditions labelled unknown. Never turn an unavailable run into zero.

3. CHECK AN ERROR AGAINST A SOURCE
Error ID and linked observation ID:
Exact factual statement needing correction:
Reliable source URL, relevant passage and date checked:
Current verified fact:
Is the incorrect fact present on a page you control?:
Affected canonical URL and relevant internal links:
Correction made, date and release/change reference:
External source correction needed, if any:
Platform feedback reference and date, if actually submitted:
Next recheck date:
Status: identified / source corrected / feedback submitted / recheck pending /
        correction observed / error persists / outcome unavailable

Correction observed requires a saved later response, not just a corrected web page.
Keep the original and later captures. A missing mention alone is not a factual error.
Do not publish fabricated reviews, rankings, client outcomes or hidden instructions.

4. RECHECK UNDER COMPARABLE CONDITIONS
Original observation ID:
New observation ID:
Question and settings matched? List differences:
Current exact answer and citation URLs:
Was the specific factual error corrected in this response?:
Evidence reference:
Other site or platform changes during the interval:
What this sample supports:
What remains unknown:

5. REPORT DISTINCT OUTCOMES
Completed observations:
Unavailable or failed observations and reasons:
Responses with a verified citation in the answer:
Responses with Sources panel links only:
Responses with an unlinked business mention:
Responses with an explicit recommendation:
These categories may overlap; do not add them into one visibility total.
Keep search impressions, platform citation reports, referral visits and qualified
enquiries in separate reporting rows with their source, dates and filters.
Report citation observations against the complete completed sample. This is a
sample of answers, not a share of every possible answer or proof of causation.

Useful references
OpenAI crawler controls: https://developers.openai.com/api/docs/bots
Google AI Search guidance: https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
Branding Tatva editorial policy: https://brandingtatva.com/editorial-policy