Everyone has an opinion about what AI search rewards. We wanted measured facts instead. So on June 15, 2026 we ran a clean experiment: we took 30 real questions a 6-to-8-figure ecommerce operator would actually type, asked each one to four AI engines in their live-web-search mode, and recorded which domains those engines cited as sources. The result is this page — an original dataset showing exactly who AI search cites for ecommerce, and where the winnable positions are. Every number below comes from that run.
Methodology
- Run date
- June 15, 2026 (a single snapshot)
- Queries
- 30 real questions a 6-to-8-figure ecommerce operator would type
- AI engines
- ChatGPT (GPT-4o search), Claude, Perplexity, Gemini — each in live-web-search mode
- Cells
- 30 queries × 4 engines = 120 query-engine cells
- What we recorded
- Which domains appeared as cited sources in each answer
- Measured separately via organic SERP using DataForSEO
Honest limitations. This is a single snapshot from one day, not a trend yet — treat it as a baseline to re-measure against, not a moving average. And Gemini's per-engine citation counts were under-captured: its grounding citations did not parse consistently this run, so we exclude Gemini from the distinct-domains-per-engine numbers below rather than report a figure we cannot stand behind.
AI Search Is Citation-Driven
The first finding is the one that reframes everything else: AI answers cite sources, and they do it most of the time. Of the 120 query-engine cells we measured, 102 — that is 85 percent — returned at least one cited source. These engines are not pulling answers out of thin air; they are reading the live web, picking sources, and showing their work.
That single fact decides the whole strategy. If 85 percent of AI answers cite sources, then the game is not "rank a page" in the old blue-link sense — the game is being a citable source. An engine can only cite content it can find, crawl, parse, and trust enough to attribute. If your store is not in that pool, you are not losing the citation race; you are not in it at all.
AI search is citation-driven. 85 percent of the answers we measured cited at least one source — so being a citable source, not just a ranked page, is the game.
The Leaderboard: Who Actually Gets Cited
Across the 120 cells, 422 distinct domains were cited. That is a remarkably wide pool for just 30 queries — AI search is not winner-take-all. But the distribution is steep at the top. Three platforms — google.com, youtube.com, and shopify.com — sit far above everyone else, each appearing in roughly a fifth of all answers. Beneath them, the share drops sharply into a long tail of ordinary content sites and small SaaS blogs, where most domains appear in only a single-digit percentage of answers.
Here is the top of the index. "Share" is the percentage of the 120 AI answers that cited that domain.
| # | Domain | Share of answers | Role |
|---|---|---|---|
| 1 | google.com | 22% | Platform |
| 2 | youtube.com | 20% | Platform |
| 3 | shopify.com | 18% | Platform |
| 4 | ecomseo.co | 5% | Beatable |
| 5 | ahrefs.com | 4% | Beatable |
| 6 | seo.ai | 4% | Beatable |
| 7 | semrush.com | 3% | Beatable |
| 8 | techradar.com | 3% | Beatable |
| 9 | ringly.io | 3% | Beatable |
| 10 | get-ryze.ai | 3% | Beatable |
| — | Also at 3%: seoprofy.com, webfx.com, aioseo.com, eesel.ai, linkgraph.com | 3% | Beatable |
Platforms Dominate the Top, but the Mid-Tier Is Beatable
Look at the gap between rank 3 and rank 4. The three platforms — google.com (22%), youtube.com (20%), shopify.com (18%) — occupy a tier all their own. You are not going to out-cite Google with a product guide, and you do not need to try. They show up because AI engines lean on them as general references, not because they answered the operator's specific question better than you could.
The interesting part is everything below the platform tier. The next-most-cited domain, ecomseo.co at 5 percent, is an ordinary content site. Then ahrefs.com and seo.ai at 4 percent, and a cluster of sites at 3 percent — semrush.com, techradar.com, ringly.io, get-ryze.ai, plus seoprofy.com, webfx.com, aioseo.com, eesel.ai, and linkgraph.com. These are content sites and small SaaS blogs. None of them is a household name. This is the beatable mid-tier: a band of single-digit-share domains that earned their citations with content, not with platform gravity. That band is where a focused store can win citations.
The Engines Don't Agree
Here is the finding that quietly changes how you should think about "AI search optimization": the engines have wildly different appetites for sources. Counting the distinct domains each engine cited across the same 30 queries, Claude cited 213 distinct domains, Perplexity cited 166, and ChatGPT (GPT-4o search) cited 105. Claude pulled from roughly twice as many distinct sources as ChatGPT for the identical set of questions.
| Engine | Distinct domains cited |
|---|---|
| Claude | 213 |
| Perplexity | 166 |
| ChatGPT (GPT-4o search) | 105 |
| Gemini | Not reliably captured |
What this means in practice: "optimizing for AI search" is really optimizing for several different engines, each with a different willingness to reach into the long tail. A broad-appetite engine like Claude gives a wider field of sites a shot at a citation, while a narrower one like ChatGPT concentrates on fewer domains — so the bar to break in is higher there. There is no single "AI" to please. The deeper guides break down each engine's selection logic: how ChatGPT Search picks sources, how Perplexity decides citations, and how Claude decides citations.
Claude cited roughly twice as many distinct domains (213) as ChatGPT (105), with Perplexity in between (166). Different engines have different appetites, so AI search optimization is really multi-engine optimization — not one target.
The Category Is Winnable
It would be easy to read "platforms dominate the top" as "the giants own this and you cannot win." The data says the opposite. We also measured Google's organic results for the same 30 queries, and the result was unambiguous: in all 30 of 30 queries, at least one "beatable" non-giant content-site domain ranked in the top 3.
That is a clean signal. The category is not locked up by platforms. Every single operator query we tested had room in the top 3 for an ordinary content site — which means content sites rank, every time, in this space. Combine that with the wide citation pool (422 domains across 30 queries) and the beatable mid-tier on the AI leaderboard, and the picture is consistent across both surfaces: the giants take the very top, and the rest is open.
What This Means for Your Store
Everything below is grounded only in the run above — no assumptions, just what the numbers imply.
- Aim to be a citable source, not just a ranked page. 85 percent of AI answers cite sources, so the unit of victory is a citation. Build content an engine can find, parse, and attribute — that is the prerequisite for appearing at all.
- Target the beatable mid-tier, not the platforms. You will not displace google.com, youtube.com, or shopify.com, and you do not need to. The winnable positions are the single-digit-share content sites like the ones from ecomseo.co (5%) on down. Aim to be one of those.
- Optimize for several engines, because they disagree. Claude cited 213 distinct domains, ChatGPT only 105. A broad-appetite engine gives you more shots; a narrow one demands you clear a higher bar. Do not tune for a single "AI."
- Trust that the category is open. 30 of 30 queries had a beatable content site in Google's organic top 3, and the citation pool spanned 422 domains. Content sites rank and get cited here — the door is not closed.
- Re-measure before you conclude a trend. This is one snapshot. The honest move is to treat these numbers as a baseline and run the same 30 queries again later to see what moved.
For the step-by-step version of "become a citable source," read getting your store cited by AI search, and to set up the re-measurement loop, see measuring AI search visibility.