SaaS SEO Strategy & Keyword/Topical Foundations
The Websites That Influence AI Answers: What the Citation Data Actually Proves
On this page
Everyone in SEO now repeats some version of "be on Reddit, AI loves Reddit." Usually with no number attached. That bothers us, because the numbers exist: since 2025, multiple large-scale studies have measured exactly which domains AI engines cite, AI companies have signed content deals with disclosed dollar amounts, and correlation studies across tens of thousands of brands have ranked which presence signals actually predict AI visibility.
So this post is the sourced version of that conversation. Every claim below comes from a named study with a date and a sample size, or from a licensing deal with a public price tag. Where studies disagree - and they do - we show you both numbers instead of averaging away the truth.
Key takeaways
- The influence is real and measurable, but concentrated: analyses of consolidated citation data find the top 15 domains capture roughly two-thirds of AI citation share, with Reddit #1 across engines in the two largest studies (Peec AI, 30M citations; Semrush, 100M+ citations).
- The strongest proof isn't citation counting - it's money. Google pays Reddit $60M a year for content access; OpenAI's News Corp deal is $250M over five years. Companies do not pay nine figures for content that doesn't shape answers.
- Each platform has its own favorites: LinkedIn leads Google AI Mode (~15% of citations), YouTube leads AI Overviews (20.9%), Reddit leads Perplexity and ChatGPT (Semrush, Ahrefs). Only ~11% of cited domains overlap across platforms.
- For SaaS specifically, the picture is stranger: ChatGPT recommends tools but cites the vendor's own site only 11.6% of the time, and G2/Capterra received near-zero direct citations - the citations go to list-format comparison content.
- These numbers move fast. Reddit's ChatGPT citation share collapsed from ~60% to ~10% in a few weeks of September 2025. Treat every table below as a dated snapshot, not scripture.
Three tiers of proof
Claims about "what AI likes" come in three strengths. We built this post around the strongest available in each tier.
Tier 1 - Money. When an AI company signs a licensing contract with a website, influence stops being a theory. The deal sizes are the market's honest pricing of whose content shapes answers.
Tier 2 - Measured citations. Studies that ran hundreds of thousands of prompts through AI engines and counted which domains got cited, by platform, over time.
Tier 3 - Correlation. Studies that measured which brand signals (YouTube mentions, web mentions, backlinks) predict whether a brand shows up in AI answers at all.
Tier 1: The money proof
AI companies have paid for access to a short list of content sources. These are disclosed or well-reported figures:
| Content source | AI buyer | Reported value | Source |
|---|---|---|---|
| $60M / year (Feb 2024) | Columbia Journalism Review | ||
| OpenAI | ~$70M / year | Search Engine Land | |
| News Corp (WSJ, NY Post, Times UK) | OpenAI | $250M / 5 years - largest disclosed deal | Quartz |
| Axel Springer (Business Insider, Politico) | OpenAI | ~$13M / year | Quartz |
| Financial Times | OpenAI | $5-10M / year | Quartz |
| AP, Shutterstock, Dotdash Meredith, Informa | OpenAI / Google | $5M-$60M / year range | Quartz |
Read that table the way an investor would: Google's own search products cite Reddit constantly and Google pays Reddit for the privilege. The community platforms and news archives on this list are, by revealed preference, the highest-value answer-shaping real estate on the internet. You cannot buy your way onto News Corp's balance sheet, but you can absolutely have a presence where the deals point: Reddit threads about your category, and the trade press that licensing money keeps in the training data.
Tier 2: The citation leaderboards, platform by platform
The two largest measurement studies to date:
Semrush's 3-month study - 230,000+ prompts, 100M+ citations, July-October 2025:
| Platform | #1 cited | Next most cited | Notable |
|---|---|---|---|
| ChatGPT | Reddit (~10% in Oct, down from ~60% in Aug) | Wikipedia, Medium, Forbes, LinkedIn | September 2025 shift crushed both leaders' shares |
| Google AI Mode | LinkedIn (~15%) | YouTube, Reddit, Google properties | Wikipedia only ~2% here |
| Perplexity | LinkedIn, NIH, Microsoft, Google | Health/gov sources rank unusually high |
Ahrefs on Google AI Overviews (2026): YouTube is now the single most-cited domain in AI Overviews at 20.9% of all citations, up 34% in six months - helped, as Ahrefs notes, by Google owning both products and by GPT-4-class models having trained on over a million hours of YouTube transcripts. An earlier Semrush AI Overviews study of 10M+ keywords had Quora at #1 with Reddit holding ~21% - the difference between the two snapshots is itself the lesson: leaderboards shift quarter to quarter.
Profound's citation analysis - 1.4M citations, November 2025 to February 2026: LinkedIn is the most-cited domain for professional and B2B queries across all six major AI platforms, and climbed from ~#11 to #5 on ChatGPT in three months - the largest authority shift in the dataset.
The concentration effect: Peec AI's 30M-citation analysis found Reddit #1 across every major engine, while the top 15 domains combined capture ~68% of consolidated citation share. And a 118,000-answer audit found only 11% of cited domains appear on more than one platform - so "AI visibility" is really five separate games with five separate leaderboards.
The recurring names, with their own metrics
Cross-referencing the leaderboards above, these domains appear in the top tier of at least two independent studies. For scale, we've added each one's authority and traffic from our own dataset (Ahrefs DR, Moz DA, organic traffic - extracted July 11, 2026):
| Domain | Appears top-tier in | DR | DA | Organic traffic |
|---|---|---|---|---|
| reddit.com | Semrush, Peec, Ahrefs, Profound | 95 | 92 | 1.2B |
| linkedin.com | Semrush, Profound (all 6 platforms) | 99 | 99 | 180.3M |
| youtube.com | Ahrefs (AIO #1), Semrush (AI Mode) | — | — | — |
| wikipedia.org | Semrush (ChatGPT #2), Peec | — | — | — |
| quora.com | Semrush AIO (#1 in 2025 snapshot) | 92 | 93 | 101.3M |
| medium.com | Semrush (ChatGPT top-5) | 94 | 95 | 30.1M |
| forbes.com | Semrush (ChatGPT top-5) | — | — | — |
| github.com | Multiple dev-query studies | 97 | 97 | 49.0M |
| stackoverflow.com | Multiple dev-query studies | 92 | 92 | 8.2M |
Tier 3: What predicts whether AI mentions your brand
Ahrefs studied 75,000 brands to find which signals correlate with showing up in ChatGPT, AI Mode, and AI Overviews:
| Signal | Correlation with AI visibility |
|---|---|
| YouTube mentions | 0.737 - the strongest single factor |
| Branded web mentions | 0.664 |
| Branded anchor text | 0.527 |
| Brand search volume | 0.392 |
| Classic backlink metrics | Far weaker (DR correlates ~0.266 in related analyses) |
The top three factors are all off-site brand signals, not links. That matches everything else in this post: AI engines resolve entities through mentions across trusted surfaces, and being talked about beats being linked to - though the two compound, since referring domains remain the top predictor of getting cited once you are visible.
The SaaS-specific twist nobody talks about
If you sell software, one study should reframe your whole GEO strategy. Derivatex ran 40 B2B SaaS categories through ChatGPT ten times each - 233 recommendations across 219 tools - and traced every citation:
- ChatGPT cited the recommended tool's own website just 11.6% of the time.
- G2 and Capterra received zero direct citations across all 233 recommendations; all review aggregators combined got 0.9%.
- What got cited instead: vendors writing about themselves (51%), and small, often anonymous blogs (23%).
- The cited pages had a signature: 100% list-structured, 78% with a year in the title, 68% with a comparison table, 56% with an FAQ.
Two practical readings. First, review platforms still matter for entity trust (LLMs learned who you are partly from those profiles - see our foundational backlinks list), but the citation slot in a software recommendation goes to listicle-format comparison content. Second: half the citations went to vendors' own comparison pages. Publishing an honest "[X] vs [Y] vs [Z]" with a table, a year in the title, and an FAQ is currently the single most direct route into a ChatGPT software recommendation - which is why comparison pages and question-formatted content earn a disproportionate share of AI citations.
What to actually do with this
- Pick your platform battles. Only ~11% of cited domains overlap across engines. B2B SaaS: LinkedIn (leads AI Mode and professional queries everywhere) + Reddit (leads everything else) + YouTube (owns AI Overviews and the strongest brand-visibility correlation).
- Treat YouTube as a search asset, not a social channel. 20.9% of AIO citations and a 0.737 visibility correlation is not a "nice to have." One honest product-walkthrough video per major use case.
- Earn Reddit presence the only way that works - genuine participation in category threads. Google pays $60M/year for that content; a founder answering questions there is free.
- Ship list-format comparison content with tables, FAQs, and the year in the title. That's the measured citation signature, and it's also just useful content.
- Keep your registry profiles consistent (the foundational layer) so entity resolution works in your favor across all five engines.
- Re-check quarterly. Every leaderboard in this post shifted within months. We'll update these tables as the studies refresh; the sources are linked so you can verify us.
Frequently asked questions
Which website has the most influence on AI answers overall?
Reddit, by the weight of evidence: #1 across engines in the two largest citation studies (Peec AI's 30M and Semrush's 100M+ citation analyses), plus paid licensing deals with both Google ($60M/year) and OpenAI ($70M/year). The main caveat: its share on any single platform swings hard - it lost ~50 points of ChatGPT citation share in weeks during September 2025.
Do G2 and Capterra reviews help with AI visibility?
For entity trust and buyer research, yes - they remain core foundational profiles. But in the one study that traced ChatGPT's actual software-recommendation citations, review aggregators got 0.9% of citations combined. Presence there shapes what AI knows about you; it rarely earns the visible citation.
Is this the same as ranking in Google?
Decreasingly. Ahrefs found AI Overview citations drawn from top-10-ranking pages fell from 76% to 38% - AI engines increasingly cite pages classic search doesn't rank. That gap is the entire opportunity of GEO.
How often does this data change?
Fast enough that dates matter more than in any other corner of SEO. Every figure in this post carries its study period; the September 2025 ChatGPT shift reshuffled leaders in under a month. We re-verify this page as new study waves publish.
Buy vetted SaaS backlinks, simply.
Skip the outreach grind. Browse real-traffic sites, see every metric with its source, and track each link to indexed, with a 30-day guarantee.
Browse the marketplace
