Indonesia's biggest brands are nearly invisible to AI search — I scored 9 of them

Ask ChatGPT "best marketplace in Indonesia" or "which bank has the best savings account" and it has to quote some website. So I checked whose site it actually can quote — and the household names are losing.

Background — why I built this

I built a GEO audit agent that scores any URL 0–100 on how ready it is to be cited inside AI answers (ChatGPT, Claude, Perplexity). I'd already run it on fried-chicken chains. This time I pointed it at two sectors every Indonesian touches daily: e-commerce marketplaces and banks. The bigger the brand, I assumed, the better the setup. Wrong.

What

A head-to-head GEO audit of nine of Indonesia's most-used sites — five marketplaces (Tokopedia, Shopee, Bukalapak, Blibli, Lazada) and four banks (BCA, BRI, Bank Mandiri, BNI) — scoring each on whether an AI assistant can crawl, parse, and quote it.

Why it matters

AI search doesn't show ten links — it reads a few sources and quotes them. When a model answers "where should I shop" or "which bank," it pulls from the sites it can actually read. A market leader with a JavaScript-locked, schema-less site simply isn't in that answer — and a smaller, well-built competitor takes the slot. Market share doesn't transfer to AI citations automatically; readability does.

Who it's for

Brand and growth teams who assume "we're the biggest, of course AI mentions us," and the smaller players who can quietly out-optimise them. Fame is not the same as being citable.

When & where

Run live against the public Indonesian sites, June 2026. The audit checks each homepage plus the domain root (robots, llms.txt, sitemap) in one pass.

How — the marketplaces

Same seven weighted checks: AI-bot access, llms.txt, sitemap, structured data, crawlable content, answer structure, meta. Here's where each store wins or leaks:

• Shopee — Indonesia's most-used marketplace — is Invisible (14). A bot reaches the door and finds an almost-empty JavaScript shell: ~10 words of real HTML, no schema, no sitemap. The biggest store is the hardest to quote.
• Tokopedia is Weak (37) and partly self-inflicted. Its robots.txt scores 10 — it's actively discouraging crawlers — and the homepage is JS-rendered. Good schema can't save a page a bot can't read.
• Underdogs lead. Blibli (70) and Bukalapak (61) are the most AI-ready — crawlable HTML, a sitemap, clean structure. The smaller names did the boring things right.
• The whole column is red on one file. Not one of the five has an llms.txt. That's an open lane: the first marketplace to ship a clean LLM map becomes the easy source to cite.

How — the banks

Same audit, even starker result:

• Two of the four are Invisible (0). The bot couldn't even fetch bri.co.id or bni.co.id — blocked or JS-locked at the front door. If a crawler can't load your page, you don't exist to AI search, full stop. (These are two of the largest banks in the country.)
• BCA leads (54) and is the rare exception — it actually ships an llms.txt and a sitemap, the only brand across both sectors to do so. Still no structured data, so it's "Weak," but it's ahead by simply trying.
• Bank Mandiri (53) is crawlable with decent answer structure but has no sitemap and no schema — findable, not optimised.

The pattern across both sectors

Nine of Indonesia's biggest digital brands, and the scoreboard tells one story:

• Size is anti-correlated with readiness here. The most-used marketplace (Shopee) and two giant banks (BRI, BNI) score at the very bottom — their apps are great, their websites are invisible to AI.
• llms.txt is basically unclaimed. 8 of 9 sites score 0 on it. The single cheapest GEO win in the country is sitting untouched.
• JavaScript is the silent killer. Most low scores trace back to one thing: the real content lives behind JS that AI crawlers don't run. Server-rendered HTML alone would lift several of these by 30+ points.

The takeaway

In AI search, the question isn't "who's the biggest brand" — it's "whose website can the model actually read." Right now in Indonesia, the answer is: mostly the underdogs. The giants are spending on apps and ads while leaving the AI answer slot wide open. For a smaller player, that gap is the whole opportunity — be the clean, readable, llms.txt-equipped source, and you get quoted where the market leader can't.

→ Run the free GEO audit on your own site — paste a URL, get the 0–100 score and the exact fixes in one pass. No sign-up.

---

Honest limits: a low score means low readiness, not that the brand never appears — AI also pulls from third-party reviews, maps, news, and aggregators. This audits each brand's own site, live, and configs change. Two banks scored 0 because the homepage was unreachable to the crawler at audit time; that's a fetch/config issue, not a statement about the institution.