Gemini can give different local-business recommendations when you repeat the same question. A 2026 vendor-reported study found the top recommendation repeated far less often in Gemini than in Google’s Local Pack—but that result describes a particular test, not a permanent ranking or a universal rate. For a useful view of local AI visibility, track fixed prompts across the engines that matter and keep each result’s location, date, and method attached.
What the volatility figures actually measure
Scalevise reports that Steady Demand’s August 6, 2026 study analyzed 1,487 local-intent prompts across 50 U.S. metropolitan areas and 10 service categories, producing 14,472 citations from Gemini and ChatGPT. In a separate literal-repeat test, the methodology review reports that Gemini returned the same top local-business recommendation in 7.9% of cases, compared with 90.2% for Google’s Local Pack. Scalevise summarizes the contrast as about 7% versus roughly 90%. These are results reported for the study’s test, not a probability that applies to every prompt, market, Gemini interface, or future model. Scalevise’s study summary · Digital Applied’s methodology review
As an Amazon Associate I earn from qualifying purchases.
“Same top recommendation” is one outcome, not a complete measure of visibility. A business can be named without its website being cited; a source can be cited without its business leading the answer. Track those outcomes separately rather than treating them as interchangeable.
Why 40% overlap is not the repeatability rate
The often-cited 40% figure describes agreement across three differently worded local-intent queries. It does not mean that Gemini repeats the same answer 40% of the time when the exact same words are entered again. The methodology review distinguishes this cross-phrasing measure from a smaller literal-repeat experiment: six fixed New York queries, four timed rounds, and 72 calls, with reported cited-domain overlap ranging from 26.5% to 46.3%. The measures answer different questions: one tests how wording affects results; the other examines variation across repeated calls. Digital Applied’s explanation of the measures
Why compare engines separately
Generative answers do not necessarily draw from the same source pool as traditional search. In a SIGIR 2026 benchmark, top-1,000 Tranco domains appeared among the top sources for 52.7% of traditional Google Search queries, 40.0% of AI Overview queries, and 32.6% of Gemini queries. Those figures describe source selection in that paper’s benchmark; they are not local-business visibility rates and do not validate the local repeatability study. They do, however, illustrate why results from different surfaces should not be merged into a single unlabeled “AI visibility” score. Grossman et al., SIGIR 2026
For local tracking, treat Gemini, ChatGPT, Google AI Overviews, and Google’s Local Pack as distinct surfaces. Differences between them may reflect different systems and source selection, while changes within one surface over time may reflect other factors. A changed answer by itself does not establish its cause.
How to measure local AI visibility consistently
- Build a fixed prompt panel. Use real customer questions and retain exact wording for the repeatability test. Examples might include “Who’s a reliable locksmith in Canterbury?” or “best X in city.” Keep differently worded versions in a separate prompt set.
- Define the market and category. Record the intended location and service category for every prompt. Do not assume a result for one city or query transfers to another.
- Run each relevant surface separately. Log the engine or product surface, collection date and time, and the prompt used. Keep the process and cadence consistent enough that runs can be compared.
- Record distinct outcomes. Note whether the business was named, whether its own site or another source was cited, and which business led the recommendation. Do not collapse mentions, citations, and top position into one result.
- Report the denominator. Show the number of runs and the share in which each outcome appeared. Segment results by engine, market, category, and prompt type so a change in one segment is not hidden by an aggregate.
- Separate observation from action. Use a changed result to decide what to investigate, not as proof that a particular SEO change caused the answer to change.
Keep a record with at least these fields:
- Exact prompt wording and prompt type: fixed repeat or wording variant
- Engine or search surface
- Location and service category
- Run date and time, plus sample count
- Business named, cited source, and top recommendation
What observational audits can—and cannot—tell you
Findable’s UK audit reports 941 businesses and 61,848 citations and says it was last updated September 24, 2026. Findable distinguishes a business being named in an answer from appearing in a cited source, and states that its measurements do not establish that any one factor causes an AI answer. Its sample is a UK observational audit, so it should not be treated as a direct estimate for U.S. markets or as proof of a ranking tactic. Findable’s audit and stated limitations
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →The same caution applies to volatility: a monitoring log can show how often a business appeared under defined conditions, but it cannot by itself explain why the answer changed. Treat it as measurement of outcomes, not a causal diagnosis.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
