Somantra's third AI Search Visibility audit, covering 20 Australian general insurance brands across ChatGPT and Google AI Overviews from May through August 2026, finds Budget Direct holding the overall lead with 9,960 observed mentions while NRMA overtook Allianz to lead on ChatGPT. The report, built on more than 34,000 consumer conversations, shows a widening gap between how often AI engines name a brand and how often they actually recommend it - a distinction that matters for insurers deciding where to invest in AI search optimization.
Budget Direct leads Google AI Overviews with 5,771 mentions, while NRMA's 5,649 ChatGPT mentions put it 83 ahead of Allianz. Month-over-month movement on a matched query cohort was mixed: Budget Direct fell 4.4%, Allianz fell 5.4%, AAMI fell 6.0%, and ING gained 4.2%.
Visibility rank and consideration rank are pulling apart
Somantra's Brand Consideration metric measures whether an AI answer frames a brand as the most-likely recommendation rather than simply listing it. The gap between raw visibility and consideration is stark in several cases. Shannons ranks #10 in raw visibility but #1 in most-likely framing - a nine-place premium in the elite, premium, and vintage motor vehicles segment. ING shows a similar pattern, ranking #11 in visibility but #3 in consideration.
At the other end, AAMI and NRMA are high-visibility brands that underperform on consideration, ranking five and four places lower on framing than on raw mentions, respectively.
"Most AI visibility tools stop at the mention count and call that the win," said Arun Prasad, Founder of Somantra. "Our data says the mention count is the least interesting number. Shannons is being actively recommended to prospective customers in the segments where the brand competes while NRMA is being named constantly on ChatGPT without a matching lift in how likely AI is to actually recommend it. That gap between being cited and being chosen is exactly what brands need to see and act on."
Prompt wording shifts AI answers
The audit introduces three new diagnostic layers. A Modifier Test found that a single word change in a prompt materially shifts an AI answer: "most trusted" produced the largest positive sentiment shift (+0.498) across tested brands, while "cheapest" produced the largest semantic distance from baseline (0.626), meaning it changes the substance of the answer, not just its tone.
A First Seen Audit found that Google AI Overviews places a tracked brand before the scroll line 85.6% of the time, compared with 34.0% for ChatGPT, whose longer-form answers push the first brand mention to a median of word 186. Citation-layer analysis of 7,945 domains found Canstar remains the top-cited domain across all three audits, and every one of July's top 100 cited domains remained present in August.
Most AI insurance answers still name no brand
The audit reinforces a finding from Somantra's May baseline: most AI-generated insurance advice still doesn't name a brand at all. In detailed ChatGPT conversational queries, 78.9% of category-level answers named no tracked insurer, up slightly from the roughly 82% brandless rate recorded in May and July. Life insurance is the least-branded category, with 96.9% of ChatGPT conversation answers naming no tracked brand.
That leaves significant whitespace for insurers willing to build content around unbranded, in-depth advice queries - a strategy that aligns with broader AI for Insurance applications where answering questions directly can build authority before a purchase decision happens.
Why this matters for insurance professionals
For insurers, the practical takeaway is that mention counts alone don't predict whether AI is driving customers to your brand. A brand that ranks high on raw visibility but low on consideration - like NRMA or AAMI - may be getting named without being recommended, which means the AI answer is informing the customer but not converting them. The reverse is true for Shannons and ING, which appear less often but are framed as the most-likely choice when they do appear.
Insurance marketers should track both metrics separately and treat the gap between them as a diagnostic. If your brand is highly visible but rarely recommended, the problem is likely in how AI engines assess your brand's fit for the query - not in how often you're mentioned. The full report, including the Somantra Visibility Model appendix and category-level breakdowns, is available at somantra.ai. For teams building out their approach, AI for Marketing resources can help frame how to act on these signals.
Your membership also unlocks:
