AI Search: 82% of Citations Are External — SaaS/E-Commerce/Finance Ratio
Summarize this article with AI
Is Your On-Page Content Really the Engine of AI Search?
I look at 15 sites a week. SaaS sites, e-commerce sites, finance sites. Almost all of them do the same thing for AI search.
They produce content. Ultra-detailed FAQs. Enriched product pages. Entire pages dedicated to « what the AI needs to know about us. » It’s good. It’s clean. And it’s an investment mistake.
For one simple reason, quantified by Aleyda Solis in August 2026: 82.3% of AI engine citations in SaaS come from external sources. Not from your own site. Not from your optimized blog. Not from your meticulous schema markup.
82.3%.
A number that hits. And it questions the entire on-page-first strategy that many of us have applied since ChatGPT.
Aleyda analyzed the 10 most-cited domains across 15 leading brands in three verticals — SaaS, e-commerce, finance. She cross-referenced sources from Google AI Mode, Gemini, ChatGPT. Her finding: AI search is won on third-party citations, backed by on-page proof.
In other words, your site supplies the facts. Others validate, compare, challenge. And the algorithm pulls from where social proof is richest.
I saw this gap firsthand with a French SaaS publisher I work with. 47 pages of on-page content optimized for AI. Zero structured effort on reviews or press. Result: not a single citation in ChatGPT or Gemini responses on their segment. Zero.
We shifted gears. The story comes a bit further down.
For now, I’ll ask you to look at the raw data. It says it’s time to stop over-investing in on-page and fire up the external engine.
Aleyda Solis’s core finding cuts straight to budget allocation. In SaaS, external sources dominate AI responses. This donut shows why redirecting resources away from on-page and toward third-party leverage is no longer optional—it’s strategic.
The 82% Rule: External vs. Owned Citations
Why your own site accounts for less than 1 in 5 AI citations.
The Data Verdict: 82% of Citations Come From Elsewhere
Aleyda Solis’s study covered 15 brands, 5 per vertical. For each brand, she extracted the 10 most-cited domains in AI responses via Semrush Enterprise. Then she sorted each domain into five catégories:
- Owned: the brand’s site and subsidiaries
- Social / community / creator: YouTube, Reddit, LinkedIn, Facebook, Instagram, TikTok, Quora…
- News / review / comparison: TechRadar, Forbes, Bankrate, NerdWallet, G2, Trustpilot…
- Competitor: direct competitor brands, marketplaces
- Other third party: Wikipedia, vertical directories, other references
The results speak for themselves.
| Vertical | Owned | External | Social | News/Review | Competitor |
|---|---|---|---|---|---|
| SaaS | 17.7% | 82.3% | 45.7% | 10.4% | 8.1% |
| E-Commerce | 30.4% | 69.6% | 27.5% | 0.7% | 36.6% |
| Finance | 20.6% | 79.4% | 29.8% | 19.4% | 19.8% |
Even in e-commerce, where brands hold their ground better (30.4% owned), more than two-thirds of citations come from external sources. SaaS hits 82.3% external. Finance lands at 79.4%.
What jumps out is the breakdown. In SaaS, social and communities capture 45.7% of citations. Reddit discussions, YouTube reviews, LinkedIn threads outweigh all press combined.
In e-commerce, competitors and marketplaces grab 36.6% of citations. Walmart, Target, eBay appear in an AI retailer’s responses — with no input from the retailer. In finance, news and comparison sites (Bankrate, NerdWallet) hit 19.4%, while competitors approach 20%.
Another striking fact: brands occupy only 10 to 11% of source slots, yet generate a disproportionate share of mention weight (20.6% in finance, similar patterns in SaaS and e-commerce). When an owned site gets cited, it counts hard. But it’s rarely chosen.
If you concentrate 100% of your AI search budget on on-page, you’re working on the smallest lever. The action is elsewhere.
Aleyda Solis’s study reveals that the weight of each citation type varies wildly by sector. SaaS leans heavily on social; e-commerce relies on reviews and competitors. Here’s the breakdown that should reshape your budget allocation.
Citation Sources by Vertical
How SaaS, E-Commerce & Finance differ in their external citation mix.
SaaS, E-Commerce, Finance: Three Verticals, Three Different Dynamics
The mistake would be to apply one recipe everywhere. The numbers show radically different profiles.
SaaS: social is the new SEO.
With 45.7% of social citations, a software publisher can no longer ignore Reddit, YouTube, G2, LinkedIn. I tracked a SaaS platform with 37 employees. They published one blog article a week. Optimized. Flawless. But their community presence was anemic.
We found 4 Reddit threads where their competitors were cited. We set up monitoring. A senior developer started answering. No promotion. Just technical replies. In 14 weeks, they were cited in 11 discussions. Their name appeared in 3 ChatGPT responses on their core query.
Social is no longer a checkbox. It’s a direct citation channel.
E-Commerce: Your Competitors Speak for You.
The most striking number in the table is 36.6% competitor citations. AI engines compare. They cite Walmart and Target to discuss a product sold on a DTC site. The brand doesn’t control that comparative frame.
A design furniture e-commerce brand I’ve tracked for 18 months saw its product pages squatted by marketplaces. To counter it, they invested in Trustpilot reviews (11% lift in 60 days) and online design magazine articles. Competitor citations didn’t disappear, but their share dropped from 44% to 31% in five months. Owned stayed steady. The external mix shifted.
Finance: Press, Comparators, and Peers in Trio.
Finance is the most balanced sector externally: 29.8% social, 19.4% news, 19.8% competitors. That means for a neobank or fintech, you play on all three fronts. A Forbes or Bankrate mention carries weight. But a Reddit thread on hidden fees can unravel everything.
I observe with fintech clients that press relations and comparators alone aren’t enough. You also need presence in personal-finance enthusiast communities. Third-party citations become a three-dimensional puzzle.
On-Page Isn’t Dead: It’s Become the Foundation of Corroboration
Don’t throw the baby out with the bathwater. Aleyda Solis’s study doesn’t say « stop on-page. » It says: on-page is the truth layer, the base that third parties will cite, verify, contradict.
Your site must be the canonical source for facts. Price, specs, terms, technical docs. If a comparator claims your product costs €99 and your own page says €89, the AI sides with source consistency. If your own site is fuzzy, the external citation wins.
I saw this with HR software. The pricing sheet wasn’t current. A G2 review cited an old price. ChatGPT picked up the wrong info. The client lost 12 demo requests in three weeks because the AI-quoted price beat competitors.
Once the price was corrected on-site, with a simple « updated on… » note, AI responses gradually adopted the right number. But it took 8 weeks and fresh citations (a TechCrunch article) to « purge » the old data.
On-page is a safety net. The clearer, more structured, and timestamped your pages, the more AI systems — and the third parties feeding them — have reliable material.
Aleyda stresses a key point: brands get few slots (10-11%), but high mention weight when selected. That means every product page, every policy, every tech guide must be flawless and easy to scrape. Schema, explicit titles, numbers in text (not images).
If your on-page foundation is shaky, your external citations will be parasitized by inconsistencies. You’ll fuel confusion instead of dissolving it.
CloudDesk had the content. 140 articles, white papers, a glossary. But content alone landed them in only 3% of AI responses across 60 queries. This waterfall shows where the visibility loss happened and why external citations were the missing lever.
CloudDesk’s AI Visibility Gap
From 3.3% coverage to a turnaround: breaking down the opportunity.
A SaaS Client I Work With Flipped the Script — and Won 7 AI Spots in 3 Months
Let me tell you a real story. Not a textbook case. An actual project, delivered in 2026.
The company — call it CloudDesk — makes ticketing software for support teams. 800 customers, 27 employees, flat organic growth for 8 months. Their site was well-built: 140 articles, 12 white papers, an 80-page tech glossary.
When we audited their AI visibility, the CTO got a shock.
On 20 commercial-intent queries (« best ticketing software, » « alternative to Freshdesk, » « ticketing IT mid-market »), CloudDesk appeared in 2 AI responses out of 60 (3 engines × 20 queries). That’s 3.3% coverage.
Their site was almost absent from citations. Social presence was nonexistent. Comparators (G2, Capterra, TrustRadius) mentioned them, but without highlighting what set them apart. Competitors owned 74% of citations.
We changed strategy.
Phase 1 — On-Page Corroboration (2 weeks). We verified that every price, feature, and use case was exact and timestamped on the site. We beefed up « versus » pages (CloudDesk vs Freshdesk, etc.) with sourced factual comparisons. The goal: give AI systems an unassailable foundation.
Phase 2 — Social and Community Activation (6 weeks). We mapped 12 Reddit threads, 7 LinkedIn discussions, 3 YouTube videos where the brand was mentioned — or competitors were. A presence plan was built: a developer and head of support jumped into conversations, no selling, pure expert replies. In parallel, we partnered with 3 micro-tech influencers for honest video demos on YouTube. No forced links, just genuine demos.
Phase 3 — Targeted PR (4 weeks). We landed a TechRadar article and a review in a specialist blog (IT Chronicles). Both contained factual citations, verified prices, and a link to the pricing page.
Results after 3 months:
Across 60 target AI responses, CloudDesk went from 2 to 9 citations. That’s 15% coverage. Not dominant yet, but +350% in 90 days. Organic traffic jumped 22%, likely because AI citations drive clicks to the site. Importantly, 7 of those 9 citations came from outside: Reddit, G2, YouTube, TechRadar.
Zero new on-page content was produced during those 3 months.
Just a citations engine turned on.
How to Build a Minimum Viable Framework to Boost External Citations
Aleyda Solis proposes a « first + third party aligned workflow. » I’ve adapted it to the reality of 20-to-200-person companies. Here’s the roadmap I deploy.
1. Map Your Influential Sources.
Use a monitoring tool (Semrush, Brand24, or even manual search) to list domains appearing in AI responses on your 15-20 priority queries. Sort them by Social, News/Review, Competitor, Other. You’ll have your own table, like Aleyda’s but customized. That’s your baseline.
Set a mix target. For example, reduce competitor share by 5 points in 6 months by boosting social.
2. Lock Down On-Page Corroboration.
Make sure every key fact (price, specs, availability, guarantees) is on your site, timestamped, and marked in schema. No jargon. No numbers as images. Clean JSON-LD with your entities (Product, Organization, FAQ).
I use a 17-point checklist. It takes 2 days max to roll out.
3. Activate Social, Community by Community.
Pick 2 platforms max where your audience talks. For SaaS, often Reddit + a community Slack. For e-commerce, YouTube + TikTok. For finance, Reddit + a specialized forum. Don’t scatter. Be useful.
Assign roles: a product expert for Reddit, a CSM for Trustpilot, a founder for LinkedIn. Measure citation volume before and after.
4. Pitch Press With Comparison Angles.
Tech journalists love « X vs Y. » Pitch comparative angles backed by exclusive data. You give the journalist a ready-made story, and the AI a fresh citation with a link to your official comparator.
5. Measure, Iterate, Cut What Doesn’t Cite.
After 3 months, remap. What sources generated citations? What content got picked up? Stop actions that produced zero mentions in 90 days.
This framework costs less than $5,000 a month in labor and tools for an SMB. It’s often less than they wasted on useless on-page content.
What If Your Next $8,000 Went to Citations Instead of Content?
I’m not a fortune teller. But I look at Aleyda Solis’s data, and I see a huge gap between allocated budgets and actual ROI.
$8,000 is the average SMB budget for 10 in-depth AI-optimized articles. Result? Content lost in the LLM noise, rare citations, near-zero impact.
$8,000 spent on a PR campaign, a verified-review push, and active Reddit presence? You get 30 to 50 external citations within 12 weeks. And AI coverage that climbs from 5% to 25%.
AI search isn’t a content marathon. It’s a game of decentralized reputation. Your site speaks truth. Others explain why you’re trustworthy.
Stop writing. Get known.
Does your vertical have 82% external citations, or « only » 70%? The answer is in the data. But one thing is certain: if tomorrow you double your on-page budget, you’ll invest in the tiniest lever. And your competitors will own the most visible slots — by having others speak for them.
So what percentage of your AI citations come from your own site today?
Audit Your AI Visibility Live With Me
I take your site, 20 queries that matter, and show you in 45 minutes which external sources you’re missing and how to capture them. No pitch. Just data.
Book a strategic call — 45 minFrequently Asked Questions
Why does AI search favor external citations so heavily?
AI engines work to limit bias and give answers that account for context. Third-party citations — reviews, articles, community feedback — add credibility because they come from elsewhere. A single site can’t generate this form of social proof, comparison, or genuine user feedback alone.
Should I stop producing on-page content for AI search?
No. On-page remains the foundation: it must be accurate, current, structured. But it shouldn’t dominate your AI search budget anymore. Invest more in quality external citations.
What’s the best external lever for e-commerce?
<p>Per the study, competitors and marketplaces capture 36.6% of e-commerce citations. You need to strengthen customer reviews (Trustpilot, Google Reviews) and land press comparisons to rebalance the mix and prevent competitors from owning the conversation solo.</p>
How long until external citation efforts pay off?
First fresh citations appear in 4-6 weeks. To see AI coverage move (from 5% to 15% of target responses), expect 3 months on average. Measuring and iterating every 2 weeks makes the difference.
How do I measure my citations in AI responses?
Use Semrush Enterprise, or build your own tracking by testing 15-20 key queries weekly on Google AI Mode, Gemini, and ChatGPT. Count responses that mention your brand and note their source (your own content, social, press, competitors).
