What evidence should I keep when an AI visibility engagement ends?
Keep the records needed to recompute the report after the engagement ends. HubSpot CRM was named in 113 of 150 answers captured on 22 September 2026, while its own domain was cited in 34 of 150. A final slide showing only one of those numbers would leave you unable to reconstruct what was counted.
What we measured
Measured. The September Absence Index, published on 23 September 2026, asked ten fixed buyer questions in CRM software, three times on ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode. HubSpot CRM's 22 September record retained 150 valid answers. Its naming rate was 75.3 percent and its own-domain citation rate was 22.7 percent. Counts describe the dated question set; they do not establish future presence, a buyer's purchase or a product-quality ranking.
The published Wilson 95 percent intervals are 67.9 to 81.5 percent for naming and 16.7 to 30.0 percent for own-domain citation. Keep those intervals with their event and base in the handover.
What this means for you
Reasoned. Request the handover before the final reporting date. It should identify the retained question bank, engine names, collection dates, repeated observations, exclusions and event definitions. Keep the underlying answer records and their source links beside the calculated totals, in a format you can read without the supplier's dashboard.
For the HubSpot example, you need the distinction between named and own-domain cited, the 150-answer valid base and the date. You also need to know whether a later file uses the same questions and conditions. That is evidence continuity, not proof that two exports with the same column name measure the same event.
Separate measurement records from delivery records. For a page change, retain the agreed factual edit, approval and served URL. For an unresolved issue, keep its state and the next responsible person. Ask who can correct an error in the supplied data and whether your access to the final files continues after the contract ends. Do not rely on a promise to recreate the evidence later.
- If you are the buyer: open the final files and recompute one count before accepting the handover.
- If you are an agency: keep client records separate and preserve their individual definitions.
- If you change suppliers: carry the event definitions with the records, not just the headline score.
Also asked as
What evidence should I keep when an AI visibility engagement ends?
Related
- HubSpot CRM's dated record
- What a valid measurement needs
- Compare old and new metrics
- Agency evidence comparison
Next step
Use the Free 10-question check (one engine). No email is needed; runs are limited per network and per day, and by the site's daily allowance. Save the check's date, engine and limits with its result.
For a page correction left in the handover, Fix one answer, $490.
Every figure comes from the published method, v1.1.
Claiming or correcting a vendor record is free and never depends on a purchase.
Prepared by the Broadcastwell team.
Page owner: Priya Nair, Management Consultant. Corrections: hello@broadcastwell.com