Are failed AI answers the same as a brand absence?
Five observations were excluded from the September Index's 2,100 scheduled observations captured 22 and 23 September 2026, leaving 2,095 valid answers. An excluded observation is not the same as a valid answer that omits a brand. Counting the exclusions as brand absences would change the denominator without supplying another valid answer.
What we measured
Measured. The September Index exclusion method used 14 categories, ten buyer questions per category, five engines and three runs per question. The engines were ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode. Release 2026-09 was published 23 September 2026.
The 2,100 scheduled observations and 2,095 valid answers describe different bases. Exclusion records preserve the reason an observation was removed. Within a retained answer, the naming field can show that a checked vendor was absent; an excluded observation supplies no such scored answer.
The five exclusions do not establish an overall future failure probability or explain why a vendor was absent elsewhere.
Every figure comes from the published method, v1.1.
What this means for you
Reasoned. Keep attempted, valid and excluded counts in separate fields. Report the reason for an exclusion and show the valid base used for a rate. This lets a reader distinguish a missing observation from an observed omission.
If a later attempt is collected, retain its date and status rather than silently rewriting the original record. A replacement may be useful, but a report needs to show which observations it includes and why they are comparable.
- If you prepare analysis: calculate a naming rate from its declared valid base, with exclusions reported beside it.
- If you buy a report: ask where an unanswered request goes. It should neither become a brand miss nor disappear without an explanation.
- If you sell software: inspect the actual answer behind a recorded absence before treating it as evidence that the engine considered the category without you.
Unknown and zero answer different questions. Zero naming in a valid answer is an observed event under the matching rule. An unusable observation leaves the answer's content unknown. Keeping that distinction intact is part of making the eventual score interpretable.
Also asked as
Are failed AI answers the same as a brand absence?
Related
- September data and exclusion records
- Scoring specification
- What a good visibility score means
- What a client report should contain
Next step
Keep any limitations shown by Free 10-question check (one engine) beside the result; a missing observation should not become an invented brand absence. No email is needed; runs are limited per network and per day, and by the site's daily allowance.
If a retained answer supports one factual page correction, Fix one answer, $490. An excluded or unread answer alone cannot define that change.
Prepared by the Broadcastwell team.
Page owner: Priya Nair, Management Consultant. Corrections: hello@broadcastwell.com