Keep the existing competitor cohort frozen for a reporting series, and evaluate newly discovered businesses in a separate exploratory cohort. Change the core set only with a documented version or a clear rebase.
Why the cohort is part of the metric
A competitive share can move because competitors entered the denominator. That movement is different from a change in the brand's observed answers.
Illustrative example: a brand has ten qualifying mentions and its original competitors have ten, giving it 50% of this mention total. Adding another business with ten mentions changes that share to roughly 33.3%, even if every original observation is unchanged. This is a calculation example, not a market result.
OpenAI's evaluation guidance recommends evaluating against explicit objectives and datasets. Our frozen-core recommendation applies that discipline to competitive reporting.
Separate discovery from comparison
Maintain two named sets. The core set supports like-for-like reporting. The exploratory set helps investigate new names, adjacent categories and businesses customers actually consider.
For a candidate addition, record why it belongs: product overlap, buyer substitution, geography or evidence from the answer panel. Being mentioned by a model is a research lead, not automatic proof that it is a direct commercial competitor.
Review exclusions as well. An acquisition, retirement or category pivot can make a formerly useful cohort less relevant. Preserve the old definition even when you stop using it for new reports.
Rebase deliberately
- Publish the old and new entity lists.
- State the first period using the new set.
- Keep prompt, provider and counting rules explicit.
- Recompute prior periods only when retained evidence supports the new definition consistently.
- Show a bridge explaining the effect of cohort changes separately from observed answer changes.
If historical evidence cannot support a reliable recomputation, start a new series. A gap is more honest than a synthetic historical comparison.
Give the report a useful interpretation
Show brand presence and the competitor-relative measure together. A brand can have stable mention rate but a changing share because other businesses become more prominent.
Avoid “we lost visibility” when the only change is a denominator update. Explain exactly what changed and whether an operator should take a content, product or measurement action.
Use the competitor citation guide for source-level investigation. Next, review the competitor set behind the AI visibility workflow and separate new candidates from the frozen reporting cohort before the next comparison.
Fact-checked: 17 October 2026. Examples are illustrative; no customer results are reported.