The State of eCommerce Discoverability 2026
Where shoppers actually find products now, across Google and the AI answers that recommend, cite, and even check out for them.
The state of eCommerce discoverability
Momentic Research · May 2026
Where shoppers actually find products now, across Google and the AI answers that recommend, cite, and even check out for them, and what to do about it.
Shopping moved. The map you had is out of date.
What you'll know by the end
People still buy from Google in huge numbers, and they are also asking AI what to buy. Here is what holds up once you look at the data, and where the real opportunity sits.
- Where the money still is. Google organic drives most eCommerce revenue. AI sends a thin slice of traffic, and the size of that slice gets oversold.
- Why AI barely touches shopping. AI answers stay out of most shopping results, and there is a clear commercial reason.
- The checkout-in-the-answer shift. Agents can now buy inside the conversation. The plumbing is real; the volume is not there yet.
- What to actually do. The moves that earn product citations, drawn from data rather than hype.
A shopping question can end in an answer, not a results page
How AI shopping works
Ask an assistant for the best running shoes and it does not hand you ten links. It rewrites your question, pulls candidates from across the web, and writes a recommendation, sometimes with a checkout button attached.
Classic shopping SEO optimizes the page of links. The recommendation comes from the other path, where your product has to be chosen as evidence and trusted enough to name. A product can rank well and still never get recommended.
AI answers mostly stay out of shopping, on purpose
The shopping carve-out
Zoom out and AI Overviews are nearly everywhere: they now trigger on close to half of all tracked searches, up 58 percent in a year. Shopping is the deliberate exception. On shopping queries they appear on just 3 to 4 percent of searches, and over the 2025 holidays Google cut their shopping appearance by 57 percent. The reason is money. Product search is where Google sells shopping ads, and an AI answer eats that space.
This carve-out is worth defending, because where AI Overviews do appear they take the click. More than half of tracked keywords have already lost clicks to them, a figure some analysts expect to reach 75 to 80 percent. Shopping is, for now, the one surface holding that pressure back.
| Informational queries | 88% |
|---|---|
| Commercial queries | 7% |
| Shopping queries | 4% |
Source: AI Overview reach and click loss: Ilana Gershteyn and Kevin Indig, SEO Week 2026. Shopping appearance and the holiday pullback: BrightEdge and Semrush (SEO tool vendors); treat as directional.
Google organic pays the bills. AI is not a performance channel yet.
Where the revenue actually is
Google organic drives 80 percent or more of organic eCommerce revenue at a conversion rate you can measure. AI sends a fraction of that traffic, and the high conversion rates people quote for it do not survive scrutiny. The SEO chain has a known source for every input; the AI chain has a guess.
If you cannot define the math, you cannot promise the outcome. Treat AI presence as the top of the funnel, not the cash register. What converts above organic is the trust channels — direct, referral, and social — measured at 1.5 to 5 times the organic rate. Get seen in AI, believed through those channels, and the buying follows.
- SEO math (a chain you could brief a CFO on): search volume known from keyword tools; click-through known from the position curve; visits measured in analytics; conversion rate measured, about 5%; revenue attributable to the channel.
- GEO math (the same chain for AI, today): prompt volume not reported anywhere; personalization opaque, different per user; click-through largely unmeasurable; visits roughly a twentieth of organic; conversion unproven at real volume.
Source: Wil Reynolds, Seer Interactive, SEO Week 2026. SEO-math-versus-GEO-math framing and trust-channel conversion are from Seer's own analytics.
An agent can now buy inside the answer
The new checkout
In January 2026 Shopify and Google launched the Universal Commerce Protocol, an open standard that lets an AI agent browse a catalog, apply a code, and complete a purchase in the conversation. ChatGPT added Instant Checkout. The plumbing is real. The honest part: no platform has shown meaningful transaction volume yet.
- What is real: open checkout protocols, live integrations with Shopify and major retailers, agents that can transact.
- What is not proven: that shoppers are actually buying this way at scale. The volume data is absent from public disclosure.
- What it means now: get your product data clean and machine-readable so you are eligible when the volume arrives. Do not bet the year on it yet.
Your catalog is a math problem to the machine
The big idea · centroids
An AI does not read your brand name. It turns every product page, review, and mention into numbers and places them in a space where similar meanings sit together. Your brand is the average of all of it.
Distinct, original product content — real testing, photography, specs nobody else has — pulls your average into a spot only you hold. Scaling for its own sake, especially with templated or AI-spun copy, just smears it back into the crowd.
Your product pages and the wider web do different jobs
Where product mentions come from
Your product pages prove you are real and give the model clean specs to quote. Alongside them, most of what AI cites about a product is community discussion and third-party reviews. Around 40 percent of AI product answers reference Reddit. Momentic calls these outside surfaces the Influenceable Web: you can shape them, you cannot own them, and they carry the recommendation.
| Reddit & communities | 40% |
|---|---|
| Reviews & editorial | 33% |
| Your product pages | 12% |
| Other | 15% |
Source: Profound and MetaRouter analyses of AI product citations (AI-visibility vendors). Split is directional.
Most stores are unreadable to AI, and that is fixable
The readiness gap
Up to 90 percent of online stores are effectively invisible to AI shoppers. The platform can see them; their product data is just too thin for a machine to trust. Basic schema from a few years ago no longer clears the bar. This is the least glamorous work and the highest leverage, because almost everyone is failing it.
Source: eCommerce Fastlane, Mirakl readiness data (publisher and platform vendor).
Five moves for eCommerce brands
What to do Monday
Ordered by leverage, and grounded in the data rather than the hype. None of them require betting on agentic checkout before it is proven.
- Fix the product data first. Clean, complete, structured feeds and pages. This is what makes you eligible at all.
- Earn community presence. Show up honestly in the Reddit threads, reviews, and roundups AI actually cites. You cannot fake your way in.
- Publish things only you can. Real testing, original photography, specs and comparisons competitors cannot copy from a supplier sheet.
- Get ready for agentic checkout. Clean catalog data and the emerging checkout protocols, so you are there when the volume comes.
- Watch the trend, not a vanity number. Track whether you show up in AI answers over time. You will not trace a citation to a sale, so judge the program on the work.
Be the product worth recommending
Where this leaves you
Keep the organic base strong, make your product data clean enough for a machine to trust, and earn an honest presence where shoppers and AI actually look. The checkout button will follow the brands that did the unglamorous work first.
