Written 14 October 2026.
Not always. Before acting on an AI shortlist, ask again in a new chat with the same wording. What appears both times is the steadier part of the answer.
What I measured
On 9 October 2026 every accounting software prompt ran twice on ChatGPT and twice on Gemini, each in a fresh temporary chat. Sage Accounting was named in one generic ChatGPT run and not in the other. What held: ChatGPT picked FreeAgent for the self-employed plumber in both runs. Study and CSV.
How to do it
1. Open a new chat.
2. Paste the same wording.
3. Keep what appears both times. Treat the rest as unsettled.
If you track AI visibility, the same goes for your numbers: one run shows what’s possible, not what’s typical (Five labels every AI visibility number needs).
What I will not claim
- How often answers change. Two runs per prompt is enough to see answers move, not to measure it.
- That what appears twice is right. Repeated isn’t the same as correct.
- Anything beyond two engines, three prompts and one evening.
This is hack 8 on the search hacks page.