Citation strategy - GEO

The AI citation supply chain you don't own

Whether your brand shows up in ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude comes down to a short list of sites you don't own: Reddit, Wikipedia, YouTube and a handful of publishers. The top 15 domains hold 68% of all consolidated AI citation share, so the job is getting onto the sources engines already trust, not polishing your own pages. That list is also unstable: a citation you win this week can be gone next month.

The premise

Your page is only part of what the engines cite

Most brands optimise their own site and wait to be cited. But engines build answers mostly from a handful of domains they already trust, and your site is only a sliver of that. 5W Public Relations' Citation Source Index pools 680 million citations from six published studies across ChatGPT, Google AI Overviews, Perplexity, Gemini and Claude (Aug 2024–Apr 2026). It puts the top 15 domains at 68% of consolidated citation share, a concentration it calls more extreme than Google PageRank ever produced.

Reddit is the most-cited source across the four engines Semrush measured, at 40.1% of citations (ahead of Wikipedia at 26.3% and YouTube at 23.5%). That figure comes from a June 2025 Semrush study of 150,000 citations across ChatGPT, Perplexity, Google AI Overviews and Google AI Mode, so read it as a four-engine aggregate rather than a flat rate everywhere. OtterlyAI's separate study reaches the same ranking on the engines it studied: Reddit was the most-cited domain overall and topped the list on every platform. If your brand isn't in those sources, no amount of on-page work puts you in the answer - you can rank #1 on Google and still be invisible to AI.

68%

of consolidated AI citation share sits in the top 15 domains

40.1%

of citations go to Reddit, the single most-cited source (four-engine aggregate, Semrush Jun 2025)

The map

Different engines, different suppliers

There is no universal top source. Each engine draws from its own mix, so plan one engine at a time. On ChatGPT in the U.S., Wikipedia (13.15%) and Reddit (11.97%) together account for over 25% of all citations, and outside those two no domain exceeds 3% of ChatGPT citations. Those figures come from 5W's Q1 2026 ChatGPT audit, which itself draws on Similarweb's analysis of nearly 600,000 primarily ChatGPT citation events (Jan–Feb 2026). Reuters lands at 2.27% (#7), Forbes at 1.38% (#18), and the WSJ, NYT, Bloomberg and FT don't appear in the top 20 at all. Profound reproduces the same pattern on a different dataset (~730,000 U.S. English ChatGPT conversations, Oct–Dec 2025): Wikipedia #1, Reddit #2, with Reuters the highest-placed traditional newswire.

Google AI Mode has a different top domain entirely: Fandom leads it at 7.16%, ahead of Wikipedia. YouTube is the strongest single predictor of AI visibility, correlating at 0.737 (an Ahrefs analysis of 75,000 brands, reported in 5W's audit). In Google AI Overviews specifically, YouTube is the top domain overall, ahead of Mayo Clinic. Among video sources the gap is stark: per BrightEdge, YouTube is cited roughly 200x more than any other video platform, with Vimeo and TikTok at 0.1% and Dailymotion and Twitch at 0%.

The reliance on user-generated versus authority sources splits hard by engine. Authority sources here means government, academic and major institutional content, in BrightEdge's definition, not news or brand domains. Ranked by how much each leans on them:

Gemini26%

Perplexity22%

ChatGPT18%

Google AI Mode14%

Google AI Overviews9.5%

Source: BrightEdge authority-source share by engine.

User-generated content splits the same way. AI Overviews pulls 17.5% of its citations from UGC platforms, against 0.5% for ChatGPT and 0.2% for Gemini, a roughly 35x gap over ChatGPT. The same forum thread that lands you in AI Overviews does almost nothing on Gemini. BrightEdge measures the spread directly: cross-engine source overlap runs just 16–59%, while brand-recommendation overlap holds steadier at 36–55%. So two engines can name the same brand while citing completely different pages - you have to win the source on each one separately.

The Reddit question

The concentration keeps shifting

That concentration would be easy to work with if it stayed put. It doesn't. In Semrush's primary study, 230,000+ prompts across ChatGPT Search, Google AI Mode and Perplexity over 13 weeks (Jul–Oct 2025, 100M+ citations), Reddit's ChatGPT share fell from close to 60% of prompt responses in early August to around 10% by mid-September, with Wikipedia dropping from roughly 55% to under 20% over the same window. 5W attributes the drop to a single Google parameter change. RocketBlue's eight-model tracker registered the same collapse at a different magnitude: Reddit at 14.29% of ChatGPT's cited sources in early August, down to 0.21% by mid-September.

~60%

of ChatGPT responses cited Reddit in early Aug 2025, down to ~10% by mid-September (Semrush)

how much more often ChatGPT cited Wikipedia and Reddit than the next domains, pre-September

The danger is where the value sat before the swing: ahead of the September shift, Semrush found ChatGPT cited Wikipedia and Reddit about five times as often as the next most-cited domains. The two domains ChatGPT leaned on hardest were the two that dropped. Other engines held steady: Wikipedia stayed in the low single digits on AI Mode and near 0.8% on Perplexity throughout. Citation share and referral traffic are different numbers and can move opposite ways - track whichever one your revenue actually rides on.

Reddit's weight also varies by engine right now. Tinuiti's Q1 2026 data has 24% of all Perplexity citations coming from Reddit alone, and Reddit making up 44% of Google AI Overviews' social-media citations versus 5% of Gemini's, a roughly 9x gap. On Perplexity, community forums account for 16.9% of citations.

Your own domain

Your own site still counts - but only on some engines

Your own site isn't dead weight. OtterlyAI's data has brand and owned domains taking 52.5% of all citations across ChatGPT, Perplexity and AI Overviews, the three engines it studied, with news at 20.3% and community forums at 5.9%. The average splits sharply by engine. Google AI Overviews sends the most to brand domains at 59.8%, ChatGPT sits in the middle at 44.7%, and Perplexity sends the least at 28.9%.

Your own page is a strong lever on AI Overviews and a weak one on Perplexity, where third-party sources carry the answer. News is the other source that spikes on the right queries: 27% of all AI citations, rising to 49% on time-sensitive ones. If you sell into anything newsy or seasonal, earned media is worth real budget here.

What you actually do

A source-by-source playbook

Stop treating GEO as page optimisation. Here is where the effort goes:

Reddit is tier-one distribution. It's #1 on every platform and the biggest single driver on Perplexity. Get your brand mentioned accurately in the subreddits your buyers actually read, through genuinely useful answers rather than planted threads. It moves more of your visibility than anything else and it is the most volatile, so keep a running check on it.

On Wikipedia, aim for a factually correct, well-sourced mention. Wikipedia is 26–48% of ChatGPT's top-10 share. You can't buy your way in, but you can make sure any entity or category page that mentions you is accurate and cited to independent sources.

YouTube has no substitute. A ~200x lead over the next video platform and the strongest single correlation with AI visibility make a competent channel the baseline for AI Mode and Overviews. TikTok and Vimeo sit at 0.1%, so don't split AI-reach effort across them.

For earned media, chase wire and trade. On ChatGPT, Reuters (2.27%, #7) outranks the WSJ, NYT, Bloomberg and FT, none of which crack the top 20.

Size owned-content effort to the engine. It's worth most on AI Overviews (59.8%) and least on Perplexity (28.9%), so weight the work toward the engines that reward it.

Assume nothing you win is permanent. A source can be dominant one month and marginal the next, as Reddit's 60%-to-10% swing on ChatGPT showed. Track your citations per engine every month so you catch a source cooling off before your traffic does.

Run a mention scan on your brand

FAQ

Common questions

Can I target one source and cover every engine?

No. Cross-engine source overlap runs only 16–59%, so a source that dominates ChatGPT can be marginal on Gemini or Perplexity. Map the engines your buyers use and invest per engine.

Is Reddit really worth the effort if it can collapse overnight?

Yes, but it is the highest-payoff source and the one most likely to vanish. It's the #1 domain on every platform and 24% of all Perplexity citations, yet Semrush recorded its ChatGPT share falling from ~60% to ~10% in six weeks. Track it and never bank on a single win.

Does my own website still matter for AI answers?

It depends on the engine. Owned domains take 59.8% of AI Overviews citations but only 28.9% on Perplexity. Keep your site strong for the engines that reward it, and lean on third-party presence where they don't.

How often should I re-check my AI citations?

Monthly, because the head of the distribution moves in weeks. Semrush watched ChatGPT's Wikipedia share fall from ~55% to under 20% inside a single 13-week study. Because the top of the list turns over in weeks, monthly per-engine checks are what catch a drop before your traffic does.

Sources: 5W Public Relations AI Platform Citation Source Index 2026 (PR Newswire); 5W's Q1 2026 U.S. ChatGPT audit (PR Newswire), whose figures draw on Similarweb's Jan–Feb 2026 analysis; Profound's independent ChatGPT citation study (Oct–Dec 2025); Semrush most-cited-domains study (Jul–Oct 2025), with corroborating mid-September Reddit data from RocketBlue; June 2025 Semrush four-engine study via Databoks/Statista; BrightEdge cross-engine analysis and YouTube data via Search Engine Land; Tinuiti Q1 2026 citation data via Search Engine Land; OtterlyAI AI Citations Report 2026. All statistics come from publicly available reports and studies from 2025–2026.