A new analysis of Australian insurance brands shows why the next phase of AI search measurement will not be won by simply counting mentions. In an article published by NetContentSEO on August 31, Stefano Galloni examined Somantra’s latest audit of 20 insurance brands and argued that visibility inside AI-generated answers is now splitting from recommendation-like positioning. The difference matters because consumers are no longer only searching for links. Increasingly, they are asking AI systems to compare options, explain trade-offs and narrow a decision.
The central finding is straightforward but strategically uncomfortable: a brand can appear often in AI answers without being the brand the answer frames most strongly for a user’s specific need. According to the NetContentSEO report, Somantra’s August audit put Budget Direct first in overall observed AI-search visibility, with 9,960 mentions across ChatGPT and Google AI Overviews. Yet Shannons, which ranked tenth by raw visibility, ranked first on Somantra’s Brand Consideration measure within the specialist vehicle segment. ING also showed a gap, ranking eleventh in visibility but third in consideration.
That makes the insurance sector a useful early case study for a broader AI-search problem. Traditional SEO separated impressions, clicks and conversions because each step measured a different commercial reality. AI search now needs a similar separation. A mention tells a marketer that a model surfaced the brand. It does not reveal whether the system treated that brand as a generic example, a weak alternative, a source of background information or a strong choice for the user’s circumstances.
The new funnel is not just visibility
Most early AI visibility tools began with a simple question: does the brand appear in the answer? That remains useful. A company that is absent from relevant AI-generated responses has a discoverability problem, especially in categories where shoppers ask assistants to summarize complex options. But the Australian insurance data suggests that mention frequency is only the first layer of a more complicated funnel.
In the NetContentSEO analysis, visibility is separated from understanding and consideration. Visibility asks whether the AI answer names the brand at all. Understanding asks whether the system appears to correctly reconstruct what the brand offers, who it serves and how it differs from competitors. Consideration asks whether, when the user’s situation matches a real decision, the answer positions that brand as a likely recommendation rather than simply one more name on a list.
This distinction is especially important in insurance because the “best” provider depends heavily on context. A budget-conscious driver, a classic-car owner, a frequent traveller and a homeowner comparing policy exclusions may all need different recommendations. A universal leaderboard compresses those conditional decisions into a single score. That score can be useful for market monitoring, but it can also hide the situations in which a specialist brand wins.
Why Shannons and ING are more revealing than the leader
Budget Direct’s position at the top of the visibility table shows the value of broad AI presence. But the more revealing lesson comes from brands with weaker overall visibility and stronger contextual positioning. Shannons’ gap between tenth in visibility and first in specialist vehicle consideration suggests that narrow relevance can matter more than mass exposure when AI is acting as an advisory interface. The model does not need to recommend a specialist insurer in every generic insurance conversation. It needs to surface that insurer when the user’s circumstances match the category the brand is known for.
ING’s visibility-to-consideration gap points in the same direction from another angle. If a brand looks modest on a raw mention chart but performs strongly once it enters the shortlist, the strategic response may not be to chase mentions everywhere. It may be to reinforce the evidence around the use cases where the AI already sees the brand as relevant. Conversely, a brand with high visibility but weak recommendation framing may be present in the conversation without being persuasive inside it.
That is a difficult message for marketing dashboards built around share of voice. Ten thousand mentions that place a brand late in an answer, or frame it as one option among many, may be less valuable than fewer mentions concentrated around high-intent decisions. The commercial question becomes not only “How often are we named?” but “When the model is helping someone choose, what role do we play?”
Google and ChatGPT are not the same market
The same research also reinforces another warning: AI search is not a single channel. The NetContentSEO article reports that Budget Direct’s strongest platform in August was Google AI Overviews, while NRMA led ChatGPT for the second consecutive audit. Earlier reporting on Somantra’s May work found that Allianz had led combined visibility at that time, and that Budget Direct was already strongest inside Google AI Overviews specifically, with platform-level differences shaping the ranking picture.
Those differences are not cosmetic. Google AI Overviews appear inside a search-results environment designed for rapid scanning, while ChatGPT often generates longer explanatory answers before introducing commercial options. In Somantra’s August data as cited by NetContentSEO, Google AI Overviews placed a tracked brand before the scroll line 85.6% of the time, compared with 34.0% for ChatGPT; in ChatGPT responses, the first brand appeared at a median position around word 186. A brand technically present in both systems may therefore receive very different practical exposure.
Independent benchmark work points in the same general direction. Conductor’s 2026 insurance AI search benchmarks found that answer engines and AI Overviews are becoming a new first impression for insurance shoppers, with citation patterns differing across engines and with citation visibility often diverging from traditional organic rankings. The common theme is that AI-search performance cannot be reduced to one blended number without losing important competitive detail.
The source layer is part of the brand layer
AI recommendation does not emerge from brand awareness alone. Retrieval-based systems draw on a wider information environment that may include a company’s own site, comparison pages, regulator information, reviews, forums, editorial coverage and specialist publications. The NetContentSEO article notes that Somantra’s August audit analyzed 7,945 cited domains across ChatGPT and Google AI Overviews and found a stable core of repeatedly cited sources, including comparison publisher Canstar.
That matters because the competitive unit is larger than the brand website. A company may describe its product clearly on its own pages, but AI systems may also rely on how independent sources describe the same product category, which exclusions are highlighted, how comparison sites frame trade-offs and whether third-party content consistently associates the brand with a particular customer need. In that environment, AI optimization becomes less like keyword targeting and more like reputation engineering.
The practical implication is not that brands should try to manipulate third-party sources. It is that their positioning needs to be legible, consistent and evidence-backed across the web. Product pages, help centers, structured data, comparison content, expert explanations and public-facing policy details all become raw material for AI systems attempting to build a useful answer. When that material is incomplete or contradictory, the model may still mention the brand but fail to understand where it fits.
Brandless answers are still a major opportunity
One of the most important findings is that many detailed AI insurance conversations still do not name any tracked brand. The NetContentSEO article reports that 78.9% of detailed ChatGPT category-level conversations in Somantra’s monitored set named none of the 20 tracked insurers, with life insurance particularly sparse at 96.9% brandless answers. Earlier coverage by Asian Business Review also highlighted the high share of unbranded responses in detailed insurance questions.
For insurers, that is both a warning and an opening. If AI assistants are still answering many consumer questions at the level of principles, exclusions and decision criteria, brands have room to become credible sources before the recommendation layer hardens. The opportunity is not to flood the web with thin pages aimed at every possible prompt. It is to publish clear, trustworthy, technically accurate information that helps a system explain a real decision responsibly.
What marketers should measure next
The lesson from the 20-brand insurance audit is not that any current ranking is permanent. AI models, retrieval systems, prompt behavior and source corpora change quickly. The durable lesson is about measurement. AI visibility, AI understanding and AI preference-like framing are different outcomes, and they require different interventions.
A mature AI-search dashboard should measure mention frequency, first appearance, source citations, placement inside the answer, sentiment or framing, category fit, competitive co-occurrence and downstream user behavior where available. It should also separate platforms rather than averaging away engine-level differences. A brand that wins in Google AI Overviews but disappears inside ChatGPT does not have one AI-search problem. It has two different distribution problems.
For now, the safest language is careful language. Somantra’s Brand Consideration metric, as described in the source article, is a model-output measure. It does not prove consumer preference, policy purchases or revenue impact. But it does show that AI-generated answers can position brands differently even when raw visibility suggests a simpler story. In search, being seen was never the same as being chosen. AI is making that distinction harder to ignore.
