Instagram and YouTube out-cite every carmaker in Indonesia's car AI Overviews.

Google cited Instagram in 76 of the 141 Indonesian car answers we recorded in early August, and YouTube in 68 of them. The most-cited carmaker-owned property, Toyota's own retail arm, showed up in fewer than one answer in five. If you sell or market cars in Indonesia and your entire visibility plan lives on your dealership website, the data says Google's AI is building its car recommendations mostly out of content you do not own and may never have looked at. We ran 149 buying-intent car queries through the SerpApi Google AI Overview API, saved every answer and every source in order, and the picture that came back is solid enough to publish, so here are the numbers.

How the study ran

We ran the queries against Google.co.id with an Indonesian locale (gl=id, hl=id), 134 of them in Bahasa Indonesia and 15 in English, phrased the way a shopper types rather than the way a marketer writes, so "mobil MPV terbaik 2026" and "apakah mobil listrik worth it di Indonesia" rather than a clean brand-and-model string. The set spans 34 category queries (SUV, MPV, city car, LCGC, sedan, pickup), 44 electric-vehicle queries, 15 head-to-head comparisons, 10 price-tier queries, 14 buyer-intent queries, 12 single-brand queries, 5 market queries, and 15 English mirrors. For every query we logged whether an AI Overview appeared, its full text, its structure, the brands and models it named in order, and every source it cited.

An AI Overview appeared for 141 of the 149 queries, a trigger rate of 94.6%. It barely moved by language, 94.8% in Bahasa Indonesia against 93.3% in English, so there is no meaningful "the AI stays quiet in Indonesian" effect to hide behind. Trigger rates did vary by intent, and the pattern is the one global research would predict.

The answers themselves follow one house style. Every AI Overview used a list format, the average answer stitched together 8 content blocks from roughly 8.1 cited sources, and it named 2.9 brands per answer. It hedged with "depends on your needs" language in 24.1% of answers and laid out explicit pros and cons in 23.4%. The typical response opens with a one or two sentence direct answer, then breaks into sub-headed lists by budget or body style, each item pairing a specific model with a one-line reason such as cabin space, fuel economy, or resale value. This behaves like a comparison-shopping surface rather than a single-recommendation one, so the two or three brands the AI decides to name are the only ones a shopper actually sees on that query.

Two caveats belong next to every number here. AI Overviews are personalized and they drift week to week, so the stable findings are the trigger rates, the answer structures, and the source patterns, while the exact brand sitting in each slot will shift by user and by month. And this is one snapshot from one locale on one day. The raw dataset, every query and every answer, is available if you ask us for it.

What three of these answers looked like

Numbers describe the pattern, but the raw answers show why it matters, so here are three of the 141, lightly translated from the Bahasa Indonesia they were served in, with the sources Google cited under each.

Ask for the best MPV of 2026 and the overview opens on a direct pick, then fans out into budget and body-style tiers with a one-line reason per model. Toyota leads it, and of the four sources feeding the answer, three are a marketplace, an aggregator, and Instagram, with a single Toyota dealer network the only manufacturer-linked property in the mix.

Prompt: mobil MPV terbaik 2026 (best MPV 2026)

Answer, translated excerpt: "The best MPVs for 2026 include popular options like the Toyota Kijang Innova Zenix, Mitsubishi Xpander, and Hyundai Stargazer. Mid-size family MPVs, the Toyota Kijang Innova Zenix with a roomy cabin, petrol and hybrid engines and high resale value, the Mitsubishi Xpander with a modern design and very comfortable suspension, the Hyundai Stargazer with a futuristic look and abundant safety tech. Budget LCGC MPVs, the Daihatsu Sigra and Toyota Calya. Premium and electric, the Toyota Alphard Hybrid, BYD M6 and Maxus MIFA 9."

Sources cited: moladin.com, oto.com, auto2000.co.id, instagram.com

On the cheapest electric car the overview does not generalize, it lists four specific, very new, non-Japanese models with prices and range attached, and it pulls that answer mostly from Instagram, YouTube, and a marketplace.

Prompt: mobil listrik termurah Indonesia (cheapest electric car in Indonesia)

Answer, translated excerpt: "The cheapest electric cars in Indonesia right now sit between Rp150 and Rp190 million. The main options are the VinFast VF 3, Seres E1, and Wuling Air ev Lite. VinFast VF 3, Rp152 to 192 million, range roughly 200 to 300 km. Seres E1, Rp189 million, roughly 200 km. Wuling Air ev Lite, Rp184 million, roughly 200 km. Changan Lumin, Rp199 to 215 million, roughly 301 km."

Sources cited: instagram.com, youtube.com, kabaroto.com, moladin.com

And on a straight head-to-head the overview refuses to crown a winner, handing back a balanced pros list for each brand, sourced from YouTube, TikTok, and a used-car marketplace rather than from Toyota or Honda.

Prompt: Toyota vs Honda mana lebih baik (Toyota vs Honda, which is better)

Answer, translated excerpt: "Toyota and Honda are both good but win at different things. Toyota wins on engine durability and ease of ownership, Honda on driving enjoyment and modern features, and the best choice depends on your needs. Toyota's strengths, a very durable long-lasting engine, soft suspension that is comfortable for family passengers, more stable resale value, and easy maintenance. Honda's strengths, more powerful performance at high revs, heavier and more stable steering at speed, a sportier cabin and exterior, and technology features updated more quickly."

Sources cited: youtube.com, caroline.id, tiktok.com, linkedin.com

Across all three the shape is the same, a list-first answer that names two or three brands and reaches for marketplaces, Instagram, and YouTube to back them, and in none of them does the winning brand's own website carry the recommendation.

Indonesia is the most AI-Overview-saturated car market on earth

The 94.6% figure is extreme even by Indonesia's own standards, and Indonesia's standards are already the highest in the world. A Brand Radar analysis of 108 million keywords found Indonesia leads every country measured, with 37.2% of keywords triggering an AI Overview, ahead of the Philippines and Mexico at 29.1% and India at 26.8%. Cars sit far above that national baseline because automotive queries are exactly the broad, comparative, non-branded questions the AI likes to answer, so "which MPV is best" gets summarized far more reliably than "book a test drive at dealer X." Content built around "terbaik," "vs," and "which type of car" framing is much more likely to be lifted into an answer than content built around a single showroom or a single VIN. The strategic reading is simple. In most markets a GEO practice still spends half its energy arguing that AI answers matter at all, and in Indonesian automotive that argument is over, because the answer layer is the first thing a car shopper sees on nearly every research query they type, and the only live question is who gets named inside it.

Toyota is the default answer in Indonesian car queries

Toyota is the runaway leader. It appears in 57.4% of all 141 answers and, more tellingly, it is the brand the AI reaches for first in 47 of them, more than double Honda's 21 opening mentions. Honda, Daihatsu, Suzuki, and Mitsubishi fill out the familiar Japanese tier, and then two results that would have looked strange two years ago. Wuling ties Mitsubishi on total visibility, and BYD sits a whisker behind on mentions while out-ranking almost everyone on first-position placements.

This hierarchy is not the AI inventing a world of its own. It tracks the physical market almost exactly. Gaikindo data has Indonesia's vehicle market up 16% year on year in the first half of 2026 to 436,564 units, with battery-electric passenger cars more than doubling to 70,860 units, and BYD becoming the only non-Japanese brand to break the top five, up 63% year on year to 23,257 units. Ask the AI directly for the best-selling cars of 2026 and it returns the Gaikindo June wholesale leaders, the Toyota Kijang Innova, the Daihatsu Gran Max pickup, and the Toyota Avanza, the same commercial-and-family roster that has topped the charts for years. The AI's mention share works as a real-time proxy for showroom momentum, which is good news and a warning in one, because Toyota's incumbency converts straight into answer dominance and a challenger that gains real ground gains AI ground with it, as BYD's 11 first-place mentions on a two-year tenure in-market already show.

Electric and petrol are two separate races in the answer

The most important structural finding in the dataset is that Indonesian car AI Overviews run two separate competitive sets, and a brand that treats them as one will waste its effort. Split the queries into electric intent and conventional intent and the leaderboard does not shrink or grow, it inverts.

On conventional questions the top of the table is almost entirely Japanese. On electric questions it is almost entirely Chinese, with VinFast carrying the Vietnamese flag, and Toyota collapses to 3 mentions with Honda on 2, mostly in a hybrid context rather than full electric. BYD and Wuling co-lead the EV conversation and are the two brands the AI opens with, and beneath them sits a genuinely crowded field of newcomers, DFSK, Seres, MG, Chery, Denza, Jaecoo, Changan, Geely, that barely register anywhere else in the data. No single brand has built Toyota-level dominance in the electric answer, and given that the Indonesian EV market is forecast to reach roughly USD 1.85 billion by the end of 2026 on around 155,000 battery-electric units, this is the contestable space. Independent trackers already show how fast it moves, with Focus2move putting BYD at 48.2% of the emerging EV segment through April while Geely surged 578.1% and Wuling slipped, and Samuel Sekuritas projecting EV share climbing to 30 to 35% of the total market this year from roughly 16%. Ask the AI for the best-selling EVs specifically and it does not return BYD's flagships at all, it names Jaecoo, Geely, and Wuling nameplates from the latest monthly data, a list volatile enough that a Chinese newcomer can top it in any given month.

The cheapest-EV answer names cars that did not exist last year

Nowhere is the electric market's openness clearer than the price questions. Ask for the cheapest electric car and the AI does not hedge or generalize, it lists specific, very new models with prices and range attached, naming the VinFast VF 3 at Rp 152 to 192 million, the Seres E1 at Rp 189 million, the Wuling Air ev Lite at Rp 184 million, and the Changan Lumin at Rp 199 to 215 million, each tagged with an approximate 200 to 300 kilometer range. None of those are Japanese, and several were barely on sale a year earlier. This is the one corner of the market where the AI is comfortable recommending an unfamiliar brand by name, which is precisely why a challenger's fastest route into an Indonesian car answer runs through the electric and value questions rather than a head-on fight for "best SUV."

On the running-cost comparison the AI commits to a clear position. Asked which is cheaper to run, it puts electric charging at roughly Rp 188 to 250 per kilometer, about Rp 60,000 to 80,000 for a 400 kilometer trip, against petrol at roughly Rp 1,000 to 2,000 per kilometer, then balances the verdict with the standard caveat that petrol cars carry a lower purchase price. On subsidies it is more careful than a marketer might be, describing the planned government-borne VAT scheme for 2026 and then stating plainly that its continuation and timing are still delayed and under discussion. A buyer reading that answer is told the incentive is uncertain rather than promised, which is an accurate reflection of a moving policy and a reminder that the AI will happily undercut an over-eager sales claim with a hedge of its own.

The segments where the ranking is still open

Toyota's overall lead looks insurmountable until you zoom into individual segments and comparisons, where the mention distributions flatten and the openings appear. In the SUV category Toyota and Honda tie at 4 mentions each with Mitsubishi and Hyundai just behind. On fuel-efficiency queries Honda, Toyota, and Daihatsu tie at 3. In a head-to-head like "Avanza vs Xpander" the AI splits evenly, and on "Brio vs Ayla" it divides cleanly between Honda and Daihatsu. The model-level data tells the same story of a market with room at the edges.

The Wuling Air ev tied the Toyota Kijang Innova at 24 mentions each, and that pairing is the one worth pausing on. A budget electric hatchback now appears in general-purpose "best car" answers as often as one of the most established family nameplates in the country, and the BYD M6 has become a default MPV answer, the one Chinese model that recurs across both the electric questions and the ordinary family-car questions. When the AI hedges on a comparison it also tips its hand about how the verdict was reached. Asked whether Toyota or Honda is better, it will not crown one, telling the reader both are strong but at different things, Toyota for engine durability and ease of ownership, Honda for driving feel and modern features, and that the right choice depends on the buyer. Comparison and segment content is where a challenger realistically displaces an incumbent, because these buckets stay contestable in a way the broad brand and market queries, where Toyota's lead is close to fixed, never are.

Google is not reading carmaker websites

This is the finding that should reshape how an Indonesian automotive brand spends its content budget. When Google's AI assembles a car answer, the sources it leans on are overwhelmingly not manufacturer or dealer domains.

Two patterns run through the whole dataset. First, automotive marketplaces and aggregators are the backbone of these answers, with oto.com, moladin, momobil, caroline.id, carsome, seva, jba, and setirkanan supplying roughly a third of all citations between them. Second, Instagram and YouTube are the two single most-cited domains in the entire study, ahead of every marketplace and every editorial site, and social and video together carry another 18% of citations. Recognizable news media barely clears 5%. Manufacturer and dealer sites together supply only about one citation in ten, and Toyota's own retail properties show the gap clearly, with astraotoshop.com cited in 18.4% of answers and toyota.astra.co.id in just 8.5%, both far behind the marketplaces and social platforms even as Toyota dominates the brand mentions. So a carmaker can own the answer while its own domains sit near the bottom of the sources feeding it.

This is the Indonesian, extra-concentrated version of a pattern GEO researchers have documented globally. Ahrefs Brand Radar's analysis of more than three million US queries found YouTube alone captures 20.9% of all AI Overview citations, that citation concentration on AI Overviews is more extreme than on any other AI surface, and that the top 1% of domains take 47% of all citations. A separate thousand-query study put YouTube at roughly 23.3% and found video dominating nearly every vertical, with the median cited page 14 months old and pages over 2,500 words earning 1.6 times more citations than short ones. The same pattern shows up outside cars too, since one study of 1,259 citations across commercial "best of" searches found third-party lists earned 63% of citations while the vendor's own site earned just 12%, a gap researchers named "the 12% Problem." The Indonesian car data is that problem in a local accent. Google is synthesizing its recommendations from social feeds, marketplace listings, and review videos ahead of the brands' own domains, and a beautifully optimized manufacturer website, on its own, is close to invisible in the answer.

Where your brand sits in the answer right now

Read the aggregate numbers back against your own position and the picture turns specific fast. If you are Toyota, Honda, or Daihatsu, you own the brand mentions and you will keep owning them on conventional and value queries for a while, but the sources feeding those answers are almost never yours. Toyota is named in 57.4% of all answers while toyota.astra.co.id is cited in 8.5% and astraotoshop.com in 18.4%, so the AI is recommending you on the strength of marketplace listings, review videos, and Instagram posts you neither control nor measure. That holds up right until a competitor learns to shape those same sources, and most incumbents have no early-warning system pointed at the surface actually doing the recommending.

If you are BYD, Wuling, VinFast, or one of the Chinese newcomers still fighting for a foothold, the electric and value questions are the one part of the market where the answer is still being written, and the data shows it moving month to month. BYD and Wuling already co-lead the EV answer, yet the best-selling-EV list rotates through Jaecoo, Geely, and Wuling nameplates depending on the month, and the cheapest-EV answer will name a brand that was not on sale a year ago. That window is open now and narrowing as the list settles, so the cost of being invisible in it this quarter is not the cost you will pay next year.

If you run a dealer group, the marketplaces are absorbing your visibility using your own inventory. oto.com, moladin, momobil, caroline.id, carsome, and seva are cited in roughly a third of all answers between them, your stock is what fills their listings, and your own site supplies about one citation in ten across the whole market. Every answer a shopper reads about a car you sell is being sourced from a platform that monetises your inventory rather than from you.

None of these positions is visible from inside your own analytics. Rank tracking and site traffic will not tell you how often the AI names you, which sources it pulled to do it, or where a rival is gaining on you inside the answer. It is a different measurement surface, and it is the one now sitting in front of the shopper first.

What this means for car brands and dealers in Indonesia

The instinct when you see these numbers is to treat AI visibility as a replacement for SEO, and that instinct is wrong. Google's own disclosed rationale for AI Overview citation still rests on conventional fundamentals, ranking well organically, building brand authority, earning mentions in the sources the AI already trusts, and publishing credible content that reads as experience and expertise. The newer, engineerable signals sit on top of that base rather than replacing it. They include structured content that leads with the answer, schema markup that has been associated with a 2.3 times lift in citation likelihood, naming your sources in the body copy for a further 2.1 times lift, and long-form depth, since pages over 2,500 words consistently out-cite thin ones. Ranking alone is not enough because only 17 to 54% of AI Overview citations trace back to the top ten organic results depending on query type, so first-page presence is necessary and not sufficient, and recency and deep-index authority carry real weight.

For an automotive brand or dealer group in Indonesia, that translates into a specific program. Put AutoDealer, Vehicle, and FAQ schema across the site so the machine can read your inventory and your answers without guessing. Rebuild vehicle detail pages and category pages as direct-answer content that states the recommendation, the price band in rupiah, and the reason in the first two sentences before expanding, because that is the exact shape the AI lifts. Keep name, address, phone, pricing, and inventory metadata consistent across every channel, since inconsistency is what turns your own content into the source of a wrong answer about you. And then spend where the AI is actually reading. The biggest win in this dataset has nothing to do with adding another page to your own domain. It is showing up where the citations already go, with structured comparison and walkaround videos on YouTube and information-dense Instagram posts whose captions carry specs, prices, and clear model comparisons rather than mood and hashtags. Those two platforms alone are cited in more than half of all Indonesian car AI Overviews, and the brands winning those citations today mostly did it by accident, which is exactly why doing it on purpose is an advantage.

One last point specific to this market. The Bahasa answers were consistently richer than the English mirrors, with more local price figures, more local reasoning such as ground clearance framed as safety on potholed roads, and heavier citation of the local marketplaces. For an Indonesian audience the Bahasa overview is the surface that matters, and it rewards local-language, locally-priced, locally-hosted content over anything translated in from a global template.

We ran the study above as a repeatable measurement, and we point the same method at a single brand or a single segment on request. Our Automotive AI Visibility read maps where your brand is named across the buying-intent queries in your category, which sources the answer trusts to recommend you, the gap between your own domains and the marketplaces reselling your inventory, and the exact creators and platforms you would need to win to move the answer in your favour. What you get back is your real standing in the surface Indonesian car shoppers now see first, not another crawl of your website. Tell us your brand and your segment and we will show you where you sit, and if you want the raw data behind this study, all 149 queries and every answer, ask.