A study by Somantra, an Australia-based AI search brand visibility platform, has found that ChatGPT advertisements appeared more frequently alongside insurance queries containing shopping-oriented language, including “best,” “top” and “recommended.” The analysis examined Australian insurance-related ChatGPT responses in August 2026, highlighting how query wording may influence advertising exposure in conversational AI environments.
Somantra analysed 34,279 insurance queries and reported examining 31,820 ChatGPT responses after removing repeated response text. Across the study, advertisements appeared in 76% of queries. That figure rose to 85% for questions containing “best,” “top” or “recommended,” an increase of nine percentage points.
The findings offer an early indication of how advertising placement in conversational AI may relate to consumer intent. For advertisers, agencies and brands exploring AI-powered search, the results raise questions about how query context, wording and response variability affect paid visibility.
Market Landscape
Unlike conventional search advertising, where sponsored placements can appear in clearly designated positions around search results, advertising within conversational AI introduces a different context for brand exposure. Users may ask for product comparisons, recommendations, explanations or help understanding insurance coverage within the same interface.
Somantra’s analysis suggests that shopping-oriented language and location references correlate with higher ad occurrence. Queries containing a state or city were 5.2 percentage points more likely to carry an advertisement, while terms such as “discount,” “deal” or “save” corresponded to a 2.3-point increase.
By contrast, questions beginning with “why” were 6.8 percentage points less likely to carry an ad. Questions starting with “what is” or “what does” showed a six-point decrease, while claims-related questions were 5.1 points less likely to feature advertising.
Even these lower-ad categories retained substantial advertising exposure: questions beginning with “why” still carried ads 72% of the time, according to the study. The results therefore suggest differences in advertising frequency rather than a clear separation between commercial and informational queries.
Somantra reported no clear relationship between the individual terms “quote,” “price” and “buy” and the presence of advertising. This underscores the importance of examining broader query patterns rather than assuming that every conventional purchase-intent keyword predicts paid placement.
One-Word Changes and Advertising Variability
The study also examined whether changing a single word affected advertising outcomes. Across 747 pairs in which the same question was asked twice, the advertising outcome differed 31% of the time. This baseline variation complicates attempts to attribute an individual advertising result to a specific keyword.
Against that background, Somantra reported that adding “best” increased the observed ad rate from 82.5% to 96.9% across 97 question pairs. Advertisements appeared in 17 pairs following the wording change and disappeared in three. Adding “cheap” increased the reported rate from 67% to 91% across 33 pairs.
The analysis also identified differences for individual advertisers. In comparisons involving “cheap” or “cheapest,” Budget Direct advertisements appeared only on the version containing the term in 23 pairs, compared with four pairs where the ad appeared only without it.
These results are suggestive, but the relatively small samples for some word-level comparisons and the reported baseline variability limit the conclusions that can be drawn. The announcement does not provide enough methodological detail to establish whether the observed effects generalize across other insurance categories, locations or periods.
Strategic Outlook
For advertisers, the findings point to a potential role for conversational query monitoring in AI advertising measurement. Brands may need to understand not only which questions consumers ask, but also how advertising exposure varies across recommendation, comparison and informational contexts.
For agencies, the results suggest that conventional keyword research alone may not adequately explain paid visibility in AI interfaces. Repeated testing, query-level analysis and careful separation of correlation from causation will be important when assessing performance.
However, the study does not establish that particular words directly trigger advertisements, nor does it demonstrate improved conversions, brand awareness or sales. Its findings concern ad occurrence, not campaign effectiveness.
The broader implication is that measuring advertising in conversational AI requires methods that account for response variability and context. As this advertising format develops, transparent methodologies and repeatable tests will help marketers distinguish meaningful placement patterns from normal fluctuations.
Top Insights
- Shopping language correlates with ad exposure: Queries containing “best,” “top” or “recommended” carried ads 85% of the time.
- Informational queries showed lower rates: Questions beginning with “why” were less likely to carry ads, but still showed substantial exposure.
- Repeatability is a challenge: The same question produced different advertising outcomes in 31% of tested pairs.
- Some word changes showed larger differences: Adding “best” or “cheap” corresponded to higher observed ad rates in the tested samples.
- Effectiveness remains unproven: The study measured ad occurrence, not conversions or incremental commercial outcomes.
Get in touch with our Adtech experts
