How 16 Indonesian Nutrition Brands Score on AI Search Authority.

We scored 16 of Indonesia's best known milk and nutrition brands on the authority they have built for AI search. We did not measure brand awareness, shelf share or ad spend. We measured the signals that decide whether Google's AI Overviews, ChatGPT, Gemini and Perplexity name a brand when a parent asks which milk to buy. The average across the category was 19.9 out of 100, the leader reached only 39.5, and no brand has come close to owning this yet.

Search in Indonesia is moving from a list of links to a single generated answer, and an answer engine names the brands it can read and trust, not the brands that shout loudest. That leaves the category wide open, because most nutrition brands have spent a decade optimising for human shoppers and almost nothing on the graph that answer engines read. We wanted to see how the whole category looks on that measure, so we built one.

Why we built a Total Graph Authority score

Total Graph Authority, or TGA, is our name for the machine-readable foundation underneath a brand. An answer engine recommending a milk brand does not look at its advertising. It looks at five connected graphs, and it recognises, trusts and cites a brand in proportion to how strong those graphs are. We turned each of the five into a signal we could measure and pulled the data live from Google Indonesia rather than estimating it.

What the five graphs measure

The Knowledge Graph is whether Google recognises the brand as a real, resolved entity with a knowledge panel and a Wikipedia or Wikidata record. The Entity Graph is how well the brand defines itself in machine-readable markup on its own site, including schema and the verified profiles it links to. The Topic Graph is how present the brand is across the questions people ask in the category. The Authority signal is the breadth of independent sources that reference the brand, which is what an answer engine leans on when it does not trust the brand's own words. The AI Answer Graph is the outcome, how often Google's AI Overviews name the brand for category questions today.

The 2026 index

Morinaga leads the category at 39.5, and it is the only brand that scores well across every graph. Prenagen and SGM follow, both strong on live visibility. After that the scores drop fast. Eleven of the sixteen brands sit below 20 out of 100, and the maternal and adult nutrition segments trail infant formula across the board.

Two of the brands in this index, Morinaga and Prenagen, are Search Agency clients, and they placed first and second. We scored every brand on the same public data and the same rules, and the full method is below so you can trace any score yourself, including theirs.

Visibility now and the foundation underneath it

The data splits into what a brand is winning right now and what it has built to keep winning. Plotting live AI answer presence against the Foundation score pulls three brands away from the rest.

SGM shows the sharpest split in the set. Google names it in more AI answers than any other brand, yet its Foundation score is just 10 out of 100. It has no knowledge panel, no Wikipedia entity and no schema on its homepage. It wins today on third party listicles that mention it, not on anything it owns. The moment those sources change its visibility can go with them, because SGM has built nothing of its own to hold the position.

Frisian Flag is the opposite case. It has one of the strongest foundations in the category, with schema, five linked profiles and a Wikipedia entity, yet it appeared in none of the twelve category answers we tested. The foundation is there, but Frisian Flag has not written the topic content that turns a foundation into answers.

Morinaga is the only brand with both, which is why it leads. It has the entity foundation and it shows up in the answers. Every other brand is missing one side of that.

The gaps we found are specific and fixable

Every failure in this study is concrete and has a specific fix. Only 5 of the 16 brands expose Organization schema that a non-rendering AI crawler can read, and they are Morinaga, Prenagen, Frisian Flag, Appeton and Entrasol. Three brands, Dancow, Lactogrow and S-26, served no readable page to our crawler at all. Two of them blocked the request with a 403, and the third did not respond, so a crawler that does not load the page in a full browser gets nothing where the brand should be.

Only 5 of the 16 have any Wikipedia or Wikidata entity, which is the single strongest signal that Google recognises a brand as a distinct thing rather than a generic product word. Only 3 brands declare their social profiles in schema, so most brands never tell a machine that the Instagram and YouTube accounts speaking for them are theirs. Just two brands, Morinaga and Prenagen, even mention AI crawlers in their robots file, and Prenagen's rules name GPTBot while leaving out ClaudeBot and PerplexityBot. Across the whole set, the average brand draws only about three independent editorial sources on its own branded search, which is why AI answers about these brands lean on health portals like Halodoc and Alodokter instead of the brands themselves.

What sets the leaders apart

Morinaga got there by doing the unglamorous work. Its homepage carries an Organization schema block with a description, an address, a contact point and five verified social profiles declared as its own. It holds an Indonesian Wikipedia entity, and it is one of only two brands whose robots file acknowledges AI crawlers. None of this is expensive, just a few weeks of structured work that most of the category has never prioritised, and it is the difference between a brand a machine can describe and a brand it has to guess at. The same pattern shows up in our work on getting cited by AI search engines, where the brands that win are the ones a machine can read without ambiguity.

Where your brand goes from here

If your brand is in this index, its position is a snapshot of a foundation you can still build faster than your competitors, not a verdict on the brand. The category average is 19.9, which means the whole field is open and the cost of moving is low. The brands that fix their schema, claim their entity and write for the questions their buyers ask will be the ones AI answers name for the next decade, and right now almost nobody is doing it. We run this exact scorecard as an AI Search Visibility Audit for individual brands, mapping every gap and the order to fix them. If you want to see where your brand sits against this index and what it would take to lead your segment, we can map it, and you can read more about owning your AI search visibility rather than renting it.

How the score is calculated

We selected 16 nutrition brands across the maternal, infant and growing up, and adult health segments. For each brand we pulled a live branded search and twelve category question searches from Google Indonesia in September 2026, capturing knowledge panels, organic results, AI Overview citations and the questions people ask. We then ran an entity extraction on each brand's homepage to read its schema, its declared profiles, its Wikidata presence and its robots rules.

From that data we score each of the five graphs from 0 to 100. The table below sets out what each metric measures, the signals that earn points, and the weight it carries in the final score.

The Total Graph Authority score is the weighted sum of the five metrics, tilted toward AI performance because this is an index of authority for AI search rather than classic ranking. The formula is TGA = (AI Answer × 0.35) + (Topic × 0.25) + (Entity × 0.15) + (Authority × 0.15) + (Knowledge Graph × 0.10). The Foundation score shown next to the ranking is the average of the Knowledge Graph and Entity scores, which is the part of the total a brand controls through its own site and entity work.

The Authority signal is a citation breadth measure built from search results rather than a full backlink dataset, so it reads the visible sources rather than every link. Where a homepage returned no readable HTML, we recorded its on-site entity data as absent, which is the right reading for a crawler that does not load a full browser, though it may understate schema that appears only after JavaScript runs. We also tested crawlability with bot user agents rather than the verified IP ranges the real crawlers use, so a site behind bot protection may still serve the genuine Googlebot or GPTBot even where it blocked us, which makes the blocked cases a flag to check in Search Console rather than a final verdict. None of this changes the shape of the finding, which is that the category as a whole has built very little of the foundation AI search now runs on.

Total Graph Authority is our own framework, not an industry standard, and the score is a directional read of how ready a brand is for AI search rather than a verified benchmark. It is a snapshot from September 2026, and because AI systems and the sources they draw on keep changing, the visibility numbers will move. The scope is 16 brands in Indonesian nutrition scored on local queries, so it does not carry over to other markets or categories without a fresh run. Read it for what it is built for, comparing AI readiness between brands and finding the specific gaps to fix, and not as a stand-in for sales, market share or revenue, because a brand that is strong on the shelf can still sit low here.