Ask a marketer what content earns citations in AI search and you will hear the same answers: publish more, build an FAQ page, keep a blog. We classified every one of the 8,493 citations in the August 2026 Cited Index by content format, across 339 Indian brands, 12 categories and five AI engines. The data reorders that advice.
Video and forum threads carry 15% of all citations, and nearly all of it sits on four platforms you do not own: YouTube, Instagram, Reddit and Quora. The rest is split across more than 2,300 other sites. That concentration, not the volume of pages you publish, is where a content plan should start.
The formats AI cites
Here is every format we classified. "Citations per site" is the average number of citations one website earned in one category for that format, so it shows how concentrated each format is.
| Content format | Share of all citations | Sites cited | Citations per site, per category |
|---|---|---|---|
| Video (YouTube, Instagram) | 8% | 4 | 27.5 |
| Forum thread (Reddit, Quora) | 7% | 12 | 19.1 |
| Social post | 1% | 4 | 5.0 |
| Product page | 9% | 108 | 4.7 |
| Category page | 2% | 51 | 3.7 |
| Listicle ("best of" roundups) | 12% | 294 | 3.1 |
| News article | 9% | 205 | 2.7 |
| Review | 4% | 98 | 2.5 |
| Landing page | 34% | 1,145 | 2.4 |
| Editorial article | 4% | 118 | 2.3 |
| Comparison ("X vs Y") | 5% | 185 | 2.0 |
| How-to guide | 2% | 73 | 1.6 |
| Original research | 1% | 59 | 1.5 |
| FAQ / Q&A page | 0% | 1 | 1.0 |
Shares are of all 8,493 citations. The 14 formats cover 99% of them, a 1% catch-all bucket is left off, and rounding makes the column add to 98%. Landing pages are a third of the total because so many different websites get cited for one, each earning about two citations in a category. A video site earns about 27, on average, in each category it shows up in.
Why four platforms carry so much weight
Reddit, YouTube and Instagram are the three most-cited websites in the whole August edition, ahead of every marketplace and publisher. Reddit alone earned 496 of the 592 forum-thread citations. YouTube and Instagram earned 713 of the 716 video citations. All three show up in every one of the 12 categories, from baby care to CRM software, so this is a habit of the engines across the market, not one vertical's quirk.
One limit on what this shows: we measure at the site level, so the data cannot tell you how many individual videos or threads carry those citations, or whether a few of them carry most of the weight.
The catch: the heaviest formats live off your own domain
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.
Look at what the top two formats have in common. You do not publish YouTube videos, Instagram reels, Reddit threads or Quora answers on your own site. You earn your way into them, through creators who cover your category and communities where your buyers already talk. This is the same off-site layer that ChatGPT leans on more than any other engine, and that Google AI Overviews loads up on through Instagram.
Your own website still matters, and for B2B software it carries most of the citations outright. But if you sell into a consumer category, the formats with the most weight per site are the ones you cannot write from your own domain. That moves the content plan from "publish more pages" to "earn a place in the few platforms engines read most."
Listicles are the part you can act on directly
Between the earned platforms and your own pages sits the listicle, the most useful line in the table for most teams. Roundups are 12% of all citations, the second-largest format, spread across about 290 sites. The "best [category] in India" articles that publishers write are content an engine reads and repeats, and getting your brand included is a concrete, repeatable pitch. On this data, getting into evergreen roundups deserves its own budget line, separate from launch PR.
The FAQ-page surprise
Pages classified as FAQ or Q&A earned exactly one citation out of 8,493. Be precise about what that number covers. It is about pages whose main format is an FAQ, and says nothing about FAQ schema, which marks up questions on any page. In classic search, Google limited FAQ rich results to well-known, authoritative government and health websites in August 2023. The one-citation figure does not mean FAQ content is wasted, because questions and answers usually live inside landing and product pages that do get cited. It means a standalone FAQ page, built in the hope that AI cites it on its own, is not where the citations are.
What this means for your GEO and AEO plan
GEO (generative engine optimization) and AEO (answer engine optimization) are two names for the same job: getting your brand into the answers AI engines give.
- Earn a place on video and community platforms. Creator reviews on YouTube and Instagram, an honest presence in the subreddits your buyers use. This is where the citation weight is densest.
- Chase the roundups. Getting into "best of" listicles is the high-volume format you can pursue directly.
- Keep your own pages extractable, and do not expect volume alone to carry them. Landing and product pages are real citations, just thin per site. Google says there are no special optimizations needed to appear in AI Overviews or AI Mode beyond being indexed and eligible to show with a snippet, so on Google the basics of a crawlable, clear page still do the work.
- Stop treating a standalone FAQ page as a citation play. Fold the questions into pages that already get read.
All of this sits in the Citability layer of the 3-Layer AI Visibility Stack: once an engine can find you, format is where the difference shows up.
How to run it on your own category
Take the prompts your buyers type, run them across the engines, and read the sources behind each answer. Note the format of each one: a video, a forum thread, a roundup, a brand page. You will see your category's own version of this table, and it will tell you whether your next unit of effort belongs on a page you own or a platform you have to earn into. That is the same category-first method behind where to spend on-site versus off-site.
Two honest caveats
First, the content format of each cited source is classified by an automated pass over the cited URLs, so a small share of pages sit in the wrong bucket. Second, this is the August 2026 edition, measured before we changed how we collect AI answers in late October, so read it as a snapshot of that month, not a trend. It is also the India market in aggregate, not a promise about your specific brand. What it does establish is where the citation weight sat by format in August: on four platforms you have to earn your way into.
Methodology. 8,493 citations in 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 website is classified by the content format of its cited pages. "Citations per site, per category" is total citations in a format divided by the number of website-and-category pairs in that format. Sites cited are distinct domains. Aggregate and market-wide: no individual brand is named, only public platforms and content types, and this is a different measurement from the per-brand Citation Mix in the Cited dashboard. Format classification is automated and carries a small error rate. This reads the same August citations as the source-type breakdown, cut by format instead of by domain type.
Want to see which content earns citations in your category? Get a free AI Visibility Report across ChatGPT, Gemini, and Perplexity.