When AI tools like ChatGPT, Claude, Perplexity, or Google's AI search answer a question, they usually name the sources they pulled from.

Getting named (a "citation") is quickly becoming the new version of ranking high on Google. If AI recommends products and answers questions using certain sources, being one of those sources matters a lot.

Everyone in marketing claims creator content matters "more than ever" for this, but almost nobody has checked whether the data backs that up. This report does.

It pulls together and synthesizes more than twenty published studies from mid-2024 through mid-2026, covering hundreds of millions of individual AI citations across ChatGPT, Claude, Perplexity, Gemini, and Google's AI search features.

Three honest warnings before the numbers:

1. Most of this research comes from companies that sell AI-visibility tools. Striking numbers get them attention, so read everything with a grain of salt. Where studies disagree, both sides are shown.

2. Every study measures things a little differently. Don't treat numbers from different studies as if they're exact or directly comparable.

3. This world changes fast. Some findings reversed within a few months. Treat every number as a snapshot with a date on it, and check the source list at the end before repeating one in a meeting.

1. Most AI citations go to sources brands don't own. Three separate studies put it between 80% and 90%. When AI talks about a brand, it's mostly using what other people wrote, not the brand's own website.

2. A creator's popularity has almost nothing to do with getting cited. Views, likes, and subscriber counts show basically zero connection to how often AI cites a channel. AI picks sources that are useful and well-organized, not famous.

3. Two platforms dominate. YouTube and Reddit together make up about 78% of all social media citations across AI tools. LinkedIn comes third, and it leads for work-related questions.

4. Long videos win by a landslide. 94% of AI citations to YouTube point to long videos, not Shorts. The short clips most creator campaigns focus on are nearly invisible to AI.

5. The biggest creator platforms are the smallest AI surface. AI systems can't easily "read" TikTok videos, Instagram Reels, or podcast audio. A campaign that lives only on those platforms barely exists as far as AI is concerned.

6. Being talked about beats being linked to. How often a brand gets mentioned around the web predicts AI citations about three times better than how many links point to its site. Creators are part of this mix.

7. Showing up in several places has a multiplier effect. Brands mentioned positively on four or more independent platforms are 2.8 times more likely to show up in ChatGPT answers.

8. Citations stick around. Once a source earns citations, 96–100% of them carry over month to month. Old content keeps working; reviews from 18 months ago still drive citations today.

PART 01

Three research teams, using three different methods, landed on nearly the same answer: How you show up across platforms you don't own matters far more than what's on your company website.

Meltwater's ongoing tracking (5.35 million citations across eight AI models) adds a detail worth noticing: news and earned media make up about 39.5% of citations and growing, and social media's share jumped from 5.6% to 7.2% in a single month.

Social's slice sounds small, but it's growing fast, and it's exactly where creators live.

FIG. 01   Who owns the sources AI cites?

SOURCE  AIROPS; ACCELERATION PARTNERS (OCT 2025); 5W AI PLATFORM CITATION SOURCE INDEX (MAY 2026)

Ranking well on Google also no longer means AI will cite you.

Moz found 88% of Google AI Mode citations don't come from the top search results. The share of AI Overview citations coming from top-10 ranked pages collapsed from 76% in mid-2025 to 38% by early 2026 in one study, and as low as 17% in another.

Whatever earns AI citations, it's no longer the thing SEO was built to produce.

FIG. 02   The SEO signal is collapsing

SOURCE  AHREFS; BRIGHTEDGE; MOZ — MID-2025 THROUGH EARLY 2026

So what does earn them? The clearest answer comes from a study of 75,000 brands: how often a brand gets mentioned around the web predicts AI citations far better (a correlation of 0.664) than how many links point to its site (0.218).

In plain terms: AI trusts you when credible third parties talk about you, link or no link. And getting credible people to talk about you is what a creator program does.

FIG. 03   What predicts an AI citation — and what doesn’t

SOURCE  AHREFS CORRELATION STUDY, ~75,000 BRANDS (2026); OTTERLYAI (2026) FOR POPULARITY

PART 02

Where does creator content actually get cited? The platform-level data is now detailed enough to draw a real map — and it contains surprises in both directions.

YouTube shows up in roughly 16% of AI answers, ahead of Reddit's 10% in cross-company data. Google's own AI features love it most: over half of their social citations come from YouTube. Perplexity leans on video heavily too (38.7% of its social citations). ChatGPT barely uses it (9.5%), and Gemini and Copilot almost never do.

FIG. 04   One platform, five appetites

SOURCE  CROSS-FIRM ENGINE ANALYSES; MELTWATER, 2026

Three details:

FIG. 05   Long-form or invisible

SOURCE  OTTERLYAI YOUTUBE AI CITATION STUDY, 2026

Reddit is the most-cited social platform overall in most studies: it shows up in roughly 40% of AI answers, makes up 54–63% of ChatGPT's and Perplexity's social citations, and generated 2.5 times YouTube's citation volume in one 62-brand study.

FIG. 06   Reddit swings 9x across engines

SOURCE  TINUITI, 2026

That said: it can be volatile. One tracker measured Reddit's citation share growing 73% over a single quarter, while another measured it dropping by half over a similar period. It's a high-value, but highly monitored community that rewards genuine participation and punishes anything that smells like fake grassroots posting.

For work and business questions, LinkedIn is the most-cited site across ChatGPT, Google AI Mode, and Perplexity.

The detail that matters for creator strategy: 59% of the LinkedIn content AI cites comes from individual people's profiles, not company pages.

The cited content is consistently original (95%), medium-length (500–2,000-word posts make up 72–77% of citations), and comes from people who post at least five times a month (75% of cited authors).

In other words: the people AI trusts on LinkedIn behave like creators, whether they call themselves that or not.

The platforms with the biggest creator economies are nearly invisible to AI, for a mechanical reason: AI can't read them.

Videos and Reels can't be crawled, Stories disappear, captions are thin, and podcast audio is unreadable without a transcript.

A creator with 2 million TikTok followers and a 15,000-subscriber newsletter contributes more to AI visibility than one with 3 million followers and nothing AI can read.

"Can AI read this person's content?" has become a real selection question when brands are doing the work of Creator AEO.

PART 03

AI tools have preferences that don’t map cleanly to what’s ranking well. One 80-million-citation study even found that traditional "domain authority" (the classic SEO trust score) correlated negatively with citations.

What actually appears to get content cited:

Structure. Well-organized pages got cited across AI tools 78% of the time in one academic study. For creators, that means long-form beats short-form and transcripts and chapters make video more accessible to LLMs.

Specifics. AI rewards checkable facts. "This changed my life" adds nothing to an AI answer. “This [specific tool] works well for [specific use case]” is citation material. This is the biggest change AI demands of creator briefs (after a decade of training creators to share feelings instead of facts.)

Freshness. Perplexity cited content less than 30 days old 82% of the time. Recently updated content earned 3.2 times more citations. Even putting the year in a title helped by about 30%. Creators already publish on exactly this rhythm.

Corroboration. That 2.8x boost for being mentioned on four or more independent platforms explains why one creator partnership underperforms and a coordinated group doesn't. And once earned, citations hold: 96–100% carry over month to month. Creator content behaves like an asset that keeps paying, not a campaign that ends; which means brands treating creator fees as one-time campaign spend are underpricing what they're getting.

PART 04

01  Nobody has measured creators as a class.

Every citation study here measures domains and platforms. “YouTube’s citation share” conflates a solo educator, a brand channel, and a media company. The 59%-individual-profiles finding on LinkedIn is the closest thing to a human-level breakdown that exists.

No published study has coded a large citation sample by source type, which means the single most strategically important ratio in this field has never actually been measured.

02  Attribution remains immature.

Only 24% of marketers track AI visibility at all, and 27% of brands have already found AI describing them incorrectly—more brands than have a process for monitoring it.

03  Causality is mostly assumed.

The correlations are strong, but almost nobody has published a controlled experiment—add a creator program, change nothing else, measure the citation lift. The people running those experiments aren't sharing.

04  Stability is uncertain.

Reddit is rising and falling at the same time in different studies, and big month-to-month swings elsewhere suggest citation patterns can change quickly.

How well creator-earned authority survives AI model updates is an open question with real budget implications.

I’ll be monitoring this topic and reporting back on what unfolds!

Sauce recently rolled out MySauce, a hub where you can spotlight your citations and AI search visibility insights (as well as your hobbies, interests, and personality as a creator). Here’s a sneak peek at mine! Go grab yours :)