Ask five AI engines the same Indian buying question and you get five different answers. That much is obvious. What almost nobody measures is the sources underneath those answers, and there the engines agree on one thing and disagree on everything after it.
The agreement: every engine reaches for a brand's own website first. Across 8,493 citations in the August 2026 Cited Index, brand-owned sites are the single largest source for all five engines. The disagreement is how far past a brand's own site each engine travels, and that is where a source strategy is actually decided.
| Engine (citations) | Brand site | Community & social | Marketplace | Editorial | Review |
|---|---|---|---|---|---|
| ChatGPT (1,337) | 42% | 19% | 4% | 21% | 7% |
| Perplexity (1,465) | 50% | 14% | 8% | 18% | 7% |
| Gemini (1,614) | 72% | 2% | 9% | 9% | 3% |
| Google AI Overviews (3,159) | 54% | 22% | 11% | 6% | 3% |
| Google AI Mode (918) | 54% | 20% | 15% | 5% | 2% |
Three things in that table are worth sitting with.
Gemini barely leaves the brand's own website
Gemini cites brand-owned sites 72% of the time, higher than any other engine, and almost nothing else: community and social sources are just 2% of its citations. When Gemini answers a category question, it reads what brands themselves publish and largely stops there. If your own pages are thin, Gemini has little else of yours to reach for, and it will reach for a competitor's pages instead.
ChatGPT leaves it the most
ChatGPT is the mirror image. It is the least brand-site-heavy engine, at 42%, and the most likely to go outward: 21% of its citations are editorial (press, news, expert publications) and 19% are community and social. Marketplaces, the source ChatGPT is often assumed to run on, are just 4% of its citations, the lowest of any engine. When ChatGPT recommends a product, it is disproportionately reading what other people wrote about the brand, not the brand's own store listing.
The two Google platforms agree on what, and split on how many
Want to know how your brand scores on these same metrics?
We'll run 20 prompts across 3 AI platforms and send your AI Visibility Report within 24 hours.
The finding people repeat is that Google AI Overviews and AI Mode are opposites. In source type, they are not: both sit around 54% brand-owned and about 20% community. Where they genuinely diverge is density. AI Overviews lists roughly 10 sources per answer; AI Mode lists about 3. Same company, same kind of sources, very different number of slots. AI Overviews rewards being one of many present; AI Mode rewards being the one obvious answer. A brand comfortable in Overviews can still miss AI Mode, because there are only three seats.
Where earned media and marketplaces actually sit
Two corrections to the usual mental model. Marketplaces are a minority for every engine, from 4% on ChatGPT to 15% on AI Mode, not the dominant source they are assumed to be. Editorial, meanwhile, is stronger than expected on the two engines that reason over retrieved text: 21% of ChatGPT's citations and 18% of Perplexity's. A press hit is not dead weight for AI visibility. It is disproportionately useful on exactly the engines that read around a brand rather than straight at its store page.
What this means before you spend another rupee
The instinct is to treat "get cited by AI" as one task. The data says the first move is the same everywhere, publish content on your own domain worth citing, because every engine starts there. The second move is where the engines split, and it should follow the engine your buyers actually use.
If your buyers lean on Gemini, your own pages are doing almost all the citation work, so that is where the investment sits. If they lean on ChatGPT or Perplexity, editorial coverage and community presence are carrying a large share of your citations, and a thin off-site footprint is a real gap. This is the engine layer of the same split we mapped across categories in the August 2026 Cited Index: brand-owned share runs from 90% in B2B down to 27% in the most contested consumer categories. Where the 3-Layer AI Visibility Stack puts the pressure, Layer 1 on your own site or Layer 3 off it, shifts with the engine. Citation Source Mix, the sixth of the Cited 8 metrics, is the number that tells you which sources are carrying your citations on each engine.
One caution the data earns. This is a market-wide picture of which source types get cited, not proof of what causes a citation. It tells you where the engines are looking; it does not promise that flooding a channel lifts your visibility. The honest next step is to measure your own citations, per engine, per category, and find out which of these diets your buyers' engines are actually on.
Methodology. 8,493 citations from the August 2026 Cited Index, across 339 Indian brands in 12 categories and five engines (ChatGPT, Perplexity, Gemini, Google AI Overviews, Google AI Mode). Each cited source domain is classified by type and pooled market-wide by citation count. Because it is market-wide, there is no single "your brand" in it, so it uses no owned-versus-competitor split. Shares are summed citation counts, not averages of per-brand ratios; a small "other" bucket (4% to 7% per engine) is omitted from the table. The August edition runs on a rebuilt citation layer that reads the exact structured citations engines return, so its figures are not comparable to earlier editions.
Want to know how your own brand shows up in AI answers? Get a free AI Visibility Report across ChatGPT, Gemini, and Perplexity.