AI Search: 82% of Your Mentions Come From Elsewhere – Your On-Page Is the Foundation
Summarize this article with AI
L’étude d’Aleyda révèle un déséquilibre brutal : en SaaS, 82.3% des citations IA proviennent de sources externes (Reddit, articles, avis). Votre propre domaine ne représente que 17.7% des sources mentionnées. Mais voici le twist : quand votre site est cité, c’est pour vérifier les faits. Vous n’êtes pas concurrent—vous êtes l’autorité.
The External Citation Reality: 82.3% Come From Outside (SaaS)
Your domain is outnumbered—but it’s your most trusted source for AI verification.
82.3% External Citations: Your Site Weighs 10% of Sources, But It’s What They Verify
I look at 15 SaaS sites per week. They all want to be cited by AI. None understand how it works.
Aleyda Solis analyzed 15 brands across Google AI Mode, Gemini, and ChatGPT. The numbers speak for themselves: in SaaS, 82.3% of citations come from outside. In finance, 79.4%. In e-commerce, 69.6%. Clear: your domain represents only 10% of top source slots. Yet paradoxically, it carries 20.6% of the weight of mentions in finance, and enormous proportions elsewhere. Your site doesn’t appear often, but when it does, it counts.
This changes everything. AI Search is not a ranking competition. It’s a matter of verification. Models seek social proof on Reddit, expertise on a comparison site, expérience on YouTube. But behind every external citation, they need to confirm it on your site. If your page is poorly written, the citation collapses. If it’s clear, structured, machine-readable, the external citation gains credibility.
« Brand domains are dense in mentions but rare in slots » — Aleyda Solis
In short: your site doesn’t rank in AI. It corroborates what others say about you.
Un client SaaS est passé de 3 mentions ChatGPT à 47 en optimisant uniquement l’on-page. Voici la décomposition : avant le diagnostic, le site était invisible aux IA. Après structuration des données et création d’une fact hub, les citations ont explosé. C’est la preuve que votre page n’a pas besoin de « ranker »—elle doit corroborer.
SaaS Case: From 3 AI Mentions to 47 Per Month
How structural on-page fixes unlocked a 4.7× increase in AI citations in 8 weeks.
The Case of a SaaS Publisher That Went From 3 to 47 AI Mentions Per Month
A SaaS client calls me on a Thursday. 2,200 organic sessions per month. A technical product, well-rated on G2 and Capterra. But in ChatGPT, their tool appeared only 3 times out of 50 test prompts. And those 3 mentions never cited their site—only a Reddit thread. Frustrating.
Diagnosis took 12 minutes. The on-page was decent, but invisible to AIs. No page with verifiable specs. No indexable technical FAQ. No Article or HowTo structured data. Pricing buried in a PDF. Security policy tucked in a tab.
We rebuilt 4 pages:
- A single product sheet with specs in marked-up tables, pricing clear, technical comparisons.
- A reference center with structured FAQ in HowTo and QAPage schema, addressing every frequent Reddit question.
- A security & compliance page accessible in one click, marked up in schema.org/WebPage with legal mentions.
- A JSON-LD glossary exported as .json for crawlers, with every vertical acronym.
Cost: 14 days of development. Impact: Four weeks later, the number of ChatGPT prompts citing the product jumped from 3 to 17. After 8 weeks, 47 mentions. Including 11 with direct links to the product sheet. Reddit citations enriched themselves with « according to official docs » and pointed to our hub. The loop was live.
Organic traffic didn’t explode. But credibility with AIs tripled. That’s what matters today.
Your On-Page Doesn’t Rank. It Corroborates.
The confusion is here: we treat AI Search like Google. Add content, optimize keywords, wait. Wrong. Generative AI doesn’t rank pages—it aggregates sources.
When ChatGPT cites you, it doesn’t « rank » your site. It picks a Reddit passage, a Trustpilot review, a TechRadar article. That passage must be verified elsewhere. Where? On your domain. Your on-page becomes the landing spot where AI cross-checks facts. If your site doesn’t contain the exact information, accessible in JSON-LD and clean HTML, the citation has no foundation and AI can reject it.
Aleyda’s data shows it: owned domains are rare in slots (10-11%), yet overweighted in mention value (20%+). This imbalance reveals the corroboration role. You’re cited infrequently, but when you are, it’s powerful because you validate others.
This yields one simple rule: everything a third party can say about you must have its exact equivalent on your site. Review? You must host the details of tested features. Comparison? You must publish your benchmarks. Reddit thread? Your FAQ must answer the exact question, with the same terms. On-page is no longer a storefront—it’s the reference.
Build a Fact Hub That Runs Without You
I call it the « fact hub. » An architecture of purely factual, non-promotional pages that centralize every verifiable piece of company data: pricing, versions, compatibilities, policies, patents, awards, case studies, glossaries.
The rules are strict:
- Every data point structured (JSON-LD): Product, FAQ, HowTo, Article, Dataset if you have figures.
- Stable URLs, no parameters: so crawlers like Common Crawl can freeze them.
- Factual content, no storytelling: AI seeks verifiable claims, not metaphors.
- Dated updates: schema.org/dateModified on every page to signal freshness.
- HTML tables marked with row span / col span compatible with Knowledge Graph extraction.
Result at another SaaS client: after 6 months, the fact hub became the reference source for 23 Reddit communities and 4 vertical newsletters. Zero outreach. Just because the information was cleaner, more readable by bots and humans.
You win by stopping writing to please. Write to be reusable.
SEO, Digital PR, and Social: The Trio That Powers the Loop
Aleyda’s study confirms it with numbers: depending on sector, the nature of external citations varies. In SaaS, 45.7% come from social and communities (Reddit, LinkedIn, YouTube). In e-commerce, 36.6% come from competitors and marketplaces. In finance, 19.8% from competitors, 19.4% from news/review sites. Impossible to manage everything from one silo.
I learned this with Guillaume Attias and the DOSE framework (BMO Academy). The O for « Orchestration » takes full meaning here. SEO can’t stay isolated. It must brief digital PR on topics to push in press, align social media with emerging AI queries, and produce the on-page content that will be cited.
Concretely:
- SEO identifiés AI prompts where your brand is absent, then lists missing facts.
- Digital PR places those facts in articles, podcasts, interviews.
- Social media brings those facts to life in Reddit, Quora, LinkedIn conversations.
- On-page centralizes and marks up the « single source of truth » version that everyone else points to.
Once this circuit is established, every external effort strengthens the on-page hub, and every on-page improvement makes future citations easier. It’s a virtuous cycle.
Aleyda Solis a analysé 15 marques majeures et découvert que le poids des citations externes varie fortement selon votre secteur. Voici où vos concurrents, les réseaux sociaux et la presse puisent leurs citations—c’est là que vous devez être présent pour être cité.
Citation Sources by Vertical: Where Your Mentions Really Come From
External citations dominate across SaaS, e-commerce, and finance—but the source mix varies dramatically.
What Vertical Numbers Change in Your Priorities
Back to the numbers. In e-commerce, competitors and marketplaces weigh heavily: 36.6%. Your on-page must plan for comparisons. Publish objective comparison tables on your product pages. Specify dimensions, materials, certifications. Make it easy for Walmart, Amazon, eBay to cite your products.
In finance, news and reviews count for 19.4%. Digital PR is priority. But on-page must host your regulatory ratios, your annual reports marked up in Dataset schema, your compliance records. Every Forbes or Bankrate article will verify these figures. If they find only a PDF, the citation will be fuzzy.
In SaaS, social and communities represent 45.7%. Reddit is the playground. Your on-page must answer Reddit questions with the same words, the same objections, the same use cases. Build a living FAQ that evolves with threads. I saw a SaaS client gain 13 extra AI citations in 3 weeks just by adding a « Reddit section » to their FAQ, where every question repeated the exact title of a popular thread.
Adapt your on-page structure to your sector.
Your Corroboration Loop, Now
Here’s the minimum viable to start:
- Audit your 20 priority AI prompts. Note the cited sources, the mentioned facts.
- Verify the on-page presence of each fact. Is it on your site? Is it in HTML? In JSON-LD? Is it findable in 2 clicks?
- Complete the fact hub. Add missing pages, mark them in structured data, get them indexed.
- Feed social via SEO. Brief your community manager on uncovered Reddit questions. Have them answer with links to your fact pages.
- Place digital PR on news and comparison sites missing from your citations. Provide them factual sheets, not press releases.
- Measure source evolution. Each month, reanalyze your top 10 citations on your prompts. Your domain should gain in weight, even slowly.
The loop doesn’t demand more budget. It demands new coordination. On-page, PR, and social are no longer separate channels. They’re three pillars of the same corroboration system.
And once this system is tuned, it runs without you.
Audit Your AI Foundations in 25 Minutes
I review your 10 critical prompts, on-page foundation, and current third-party sources. In 25 minutes, you know exactly which 3 pages to build to enter the corroboration loop. No sales pitch. Just diagnosis.
Book a strategic call — 45 minFrequently Asked Questions
How long before I see the first AI citations thanks to on-page?
<p>First signs arrive in 3 to 6 weeks if AI crawlers reindex quickly. I’ve observed measurable gains by week 4 at several SaaS clients. The key factor is crawl freshness: use dateModified in JSON-LD and submit the hub to Common Crawl.</p>
Should I abandon traditional SEO for AI Search?
<p>Not at all. On-page that serves corroboration also strengthens classic ranking. Fact pages attract natural backlinks. Classic SEO feeds AI’s foundation. Both reinforce each other.</p>
Which schema.org tags should I use for a fact hub?
<p>Product, FAQ, HowTo, Article, Dataset (for figures), Organization with legalName and description, and QAPage for FAQ. What matters is including dateModified everywhere. Centralize critical facts first.</p>
How do I measure the impact of AI citations on my traffic?
Traffic from AI is hard to trace since links are often removed or set to ‘no referrer’. Instead, measure the increase in your brand mentions in AI responses (via tools like Semrush or AI SERP trackers), and correlate with the evolution of branded search volume in Search Console.
My site is small. Is a fact hub worth it?
Absolutely. Small brands have an edge: they start from scratch and gain authority faster. A 5-page fact hub well-marked up can capture first niche citations. I have several clients with fewer than 50 pages dominating their AI verticals because their facts are cleanest.
