How often does AI visibility change for the same question?
Across 817 comparisons of the same question on different dates, the tracked brand appeared in one check and was absent in the other 82 times (10.0%).
There were 45 gains and 37 losses in brand presence. Link status changed in 79 of the 817 comparisons (9.7%).
The sample covered 36 questions, eight engine views, and three tracking configurations for PodPitch-owned websites.
Results by engine view
| Engine view | Comparisons | Presence changed |
|---|---|---|
| ChatGPT | 107 | 6 (5.6%) |
| ChatGPT web | 107 | 3 (2.8%) |
| Claude | 107 | 14 (13.1%) |
| Gemini | 107 | 9 (8.4%) |
| Gemini web | 106 | 15 (14.2%) |
| Google AI Mode | 95 | 7 (7.4%) |
| Grok | 81 | 21 (25.9%) |
| Perplexity | 107 | 7 (6.5%) |
| All engine views | 817 | 82 (10.0%) |
How we measured it
- We used completed VisibilityOS monitoring checks on our own websites from October 19 through November 17, 2026 (UTC).
- We matched the exact recorded question, tracked brand configuration, engine view, recorded model and measurement mode, country, language, and location. We kept the last eligible check on each UTC date.
- We compared each result with the most recent eligible result on a different date. The gap was one to eight days. A brand was present when the check recorded a mention or a link. We counted link changes separately.
What the sample can tell us
This was a retrospective sample of monitored questions about our own websites. Some questions and results contribute to multiple comparisons, and the engine views had different date coverage. The percentages describe this sample. A broader estimate or a causal explanation would require a representative sample and additional controls.
For these queries, repeating the same check sometimes changed the answer to a simple question: did the brand appear at all? Recording the prompt, engine view, and check date makes those changes easier to spot and interpret.