PRecious AI Authority Index · Final edition

Same question. Different AI. Different bank first.

Your brand can lead one answer and vanish from the next. We tested what changes across services, markets, questions and languages.

This is a point-in-time study of Southeast Asian banking. It shows what appeared in tested answers and what public proof sat behind them. It is not a league table or a measure of customer preference.

Download the 2-page brief Explore the full report

What changed when we changed the route?

There is no single AI view of your brand.

That is the real finding. The answer shifts with the service, the question, the market and the language. One dashboard view can miss the competitive field that matters.

43%

A different bank appeared first

In 706 of 1,628 eligible matched comparisons, two routes answering the same question put a different bank first.

4.73

Business questions named more banks

Business-banking answers named 4.73 banks on average, compared with 3.52 for consumer-banking questions.

Markets told different stories

Singapore, Malaysia and Vietnam had consistent leaders. Indonesia, the Philippines and Thailand split or shifted by question.

The final banking matrix showing answer presence and first-mentioned share for 13 priority banks

The Banking Matrix: 13 priority banksMove right for more answer presence. Move up for more first mentions. Presence and prominence are not the same thing.

Inside the report

The answers change. Here is the proof.

These are real pages from the final report, not campaign mock-ups.

Full report page with the central 43 percent finding

FULL REPORT · PAGE 4

One question. Three services. A different starting point.

See the recorded Indonesia example and why answer order needs to be checked route by route.

Full report page with language differences across six markets

FULL REPORT · PAGE 8

Language changes more than the translation.

See where the first bank changed and where local-language answers named fewer banks.

What communications teams should do

Stop treating every AI answer as the same answer.

Check consumer and workplace routes

Google AI Search and Copilot sit in different research settings. Test the routes your customers and decision-makers actually use.

Drop the “American versus Chinese” shortcut

Qwen, DeepSeek and Kimi did not move in one shared pattern. Neither did the other services. Check the product, language and source conditions, not the flag.

Build proof beyond your own website

Clear owned content matters. So do customer evidence, regulator material and credible earned coverage that gives people more than the brand's own account.

Tell one story, with local proof

Keep the core narrative consistent across web, social, media, spokesperson and sales content. Then make the examples, evidence and language fit the market.

One language example, not a regional rule: Thai and English produced a different first-mentioned bank in 15 of 25 eligible pairs. Malay answers showed the biggest average drop in banks named. Different languages can change different parts of the answer.

Method and scope

What we tested. What we can claim.

6markets

11test routes

8languages

13 + 3banks

396cross-route answers

2,740total records

We ran a point-in-time analysis in August 2026 across Indonesia, Malaysia, the Philippines, Singapore, Thailand and Vietnam. The broad banking study covered four question groups. The cross-route comparison used two languages and three questions per market across 11 service-labelled routes.

The 2,740 total includes original, repair and extension work. It does not mean 2,740 independent tests. The 43% finding uses matched pairs where both answers named a recognised bank. Those pairs share underlying answers.

Most comparisons used APIs. The Copilot route used an Azure model with web search, not Microsoft 365 Copilot. The Google route used a search-grounded Gemini API, not Google AI Overviews or AI Mode. Some source conditions differed by route.

First mention records answer order, not trust, endorsement or purchase preference. Results can change as products, models, search and public evidence change. Full limitations and route notes appear on pages 30 and 31 of the gated report.

Choose your depth

Start with the brief. Go deep with the full report.

Executive Brief

Two pages. The core finding, the final 13-bank matrix and three actions for communications leaders.

Download the brief

Full Report

32 pages. Market profiles, service and language differences, earned media, methodology and a practical action plan.

Get the full report

Get the full Banking report

Complete the short form to download the final 32-page report.

Your campaign source is retained for attribution. Personal form data is handled by HubSpot and is not sent to GA4.

From finding to action

Find the gap. Fix the story. Build proof that travels.

Map

Check the questions, services, markets and languages that shape real customer decisions.

Align

Remove conflicts across web, social, media, spokesperson and sales content.

Prove

Make claims specific, current and checkable. Add customer and earned evidence where it counts.

What is Pulse?

PRecious Pulse is the measurement, research and insights platform behind this work. It shows what AI says about a brand, which public evidence holds up, where the gaps are and what communications teams can change.

Discuss a focused answer review