AI Search: the 5 key data points for SaaS and e-commerce brands

37,000 organic sessions, and the hidden truth behind AI

37,000 organic sessions. Fourteen months ago, that same site was doing 4,000. Yet 84% of the citations that fueled this growth didn’t come from the site itself.

I saw this on a SaaS project in 2025. Their traffic was surging thanks to ChatGPT and Google AI Mode responses. I dug into the data and understood immediately: AI engines were pulling mostly from external sources. Forums, videos, comparisons—very little from the site’s own pages.

Aleyda Solis published an analysis confirming this trend with precise numbers. By studying 15 leading SaaS brands in the US, she isolated 5 key data points that shift visibility in AI Search. And these findings apply just as much to e-commerce brands.

84 to 93% of AI citations don’t come from your site

Here’s the statistic that changes everything. According to Aleyda Solis, 84 to 93% of citation weight comes from external sources for the SaaS brands studied: third-party sites, publishers, communities, videos.

Your site, even if perfect, controls only a small slice of your presence in AI responses. The rest plays out across comparison tools, forums, YouTube, third-party docs, press.

For an e-commerce brand, that means your optimized product pages alone aren’t enough. Consumer reviews, unboxing videos, blogger comparisons matter just as much when it comes to AI deciding to recommend your product.

What to do: map out the proof ecosystem of your category. Who talks about you? On which sites? In what tone? Treat these external sources as visibility channels just as important as your own pages.

84 to 93% of AI citation weight rests on third-party sites. Your site can’t do it all alone.

Le tableau ci-dessous illustre comment ChatGPT et AI Mode puisent dans des sources très différentes. Le graphique interactif permet de visualiser l’écart pour chaque type de source.

ChatGPT and AI Mode don’t value the same external proof

Two platforms, two very different ecosystems. Aleyda Solis’s analysis shows that ChatGPT relies on structured written sources, while Google AI Mode weights videos and social networks more heavily.

Video accounts for nearly one-quarter of AI Mode sources. Almost zero in ChatGPT. That’s a significant gap.

For a SaaS selling project management software, it means a demo hosted on YouTube can land a citation in AI Mode, while ChatGPT will almost ignore it. Conversely, well-indexed technical docs get more weight in ChatGPT.

For e-commerce, think about your product videos. On AI Mode, they become content that generates citations. Not ads. A concrete AI SEO lever.

My field expérience: I helped an e-commerce client structure tutorial videos on YouTube. Three months later, AI Mode citations on their products spiked. Order of magnitude: +30% of mentions by our tracking. Without increasing the number of product pages.

ChatGPT and AI Mode don’t cite the same pages on your site

Another finding from the study: the two AIs don’t pull from the same places on your site. Aleyda Solis observes that ChatGPT mainly cites support pages, documentation, or deep-dive articles. AI Mode prefers product pages or landing pages.

This gap shifts your content strategy. A SaaS brand optimizing only commercial pages will miss ChatGPT citations. One polishing just technical resources loses visibility on AI Mode.

Example: a financial solutions client saw their tax help pages heavily cited by ChatGPT, but AI Mode cited their feature comparison pages. Two content types, two AI channels.

My advice: audit the pages cited by each platform (Semrush provides this data). Spot the gaps. Create internal linking that gives AI crawlers access to all content types. Think « dual channel. »

The cited page isn’t the one that gets the traffic

Here’s what surprises most. In the study, the page cited in the AI response (the one that explains, recommends, or compares) is often different from the page users (or AI agents) actually visit.

In SaaS: the citation is about an integration docs page, but the click lands on a homepage or trial form. In e-commerce: the AI cites an external comparison, the user lands on a generic product page.

Aleyda explains this through « task completion »: the AI answers a question (évaluation, setup), the user has a different goal (access the product, buy, compare). AI traffic follows its own path, distinct from the citation circuit.

To do: measure both citations and destination pages from AI clicks (via Similarweb AI referral). Adjust your landing pages based on actual task profiles. The cited page acts as relay, not endpoint.

No one-size-fits-all recipe: one strategy per market category

The fifth data point is a humility check. Aleyda Solis reminds us that each sub-vertical (CRM, collaboration, finance) has very different source patterns. For example, in accounting software, tax help sites account for over 16% of external citations, while in collaboration tools, communities and content creators dominate.

In plain terms, what works for a CRM won’t work for a fintech solution. Each category has its own « evidence mirror. »

For e-commerce brands, the parallel is immediate: the sources that shape AI recommendations for a mattress won’t be the same as for cybersecurity software. In bedding, it’s consumer reviews, specialist media comparisons, and test videos. In cybersecurity, technical reports and certifications carry more weight.

My approach: I deconstruct each client’s citation ecosystem across 20 to 30 target queries. I identify which source types matter. Then I activate specific levers: tech PR, publisher partnerships, video production, or community contribution. No template. Pure custom fit.

Your checklist to win in AI Search visibility

With these 5 data points, a method emerges. Not just on-page optimization. Full ecosystem work. Here are the actions I use in my engagements, within Guillaume Attias’s DOSE framework (Design, Optimize, Scale, Evolve) from BMO Academy:

• Design: map out your external citation universe. List the sources that matter to AI in your category.
• Optimize: structure your internal content to serve both AI channels (support pages vs product pages). Audit and correct information about you on third-party sites.
• Scale: deploy video presence on YouTube (AI Mode) and strengthen your contributions to technical communities and forums (ChatGPT).
• Evolve: continuously measure AI citations and traffic. Adjust your destination pages based on actual user journeys.

The AI Search landscape is just beginning. Those treating their external ecosystem as a dedicated SEO channel gain a real edge.

I don’t guess. I read the cold numbers. And you—is your external ecosystem already a structured SEO channel?

Stéphane Jambu

SEO & AI Engineer

I build growth systems / AI / Neuroscience | 650+ clients · 80 LinkedIn testimonials · 30 years of expertise · 15 years of systems running without me.

Follow on LinkedIn